Subj_static.log

Namespace(batch_size=50, data_name='Subj', dropout=0.5, epochs=200, gpu=0, log_interval=30, model_mode='static')
Use gpu0
maximum length (in tokens):  120
Done! Tokenizing Time=0.24s, #Sentences=10000
SentimentNet(
  (embedding): Embedding(21326 -> 300, float32)
  (encoder): ConvolutionalEncoder(
    (_convs): HybridConcurrent(
      (0): HybridSequential(
        (0): Conv1D(300 -> 100, kernel_size=(3,), stride=(1,))
        (1): HybridLambda(<lambda>)
        (2): Activation(relu)
      )
      (1): HybridSequential(
        (0): Conv1D(300 -> 100, kernel_size=(4,), stride=(1,))
        (1): HybridLambda(<lambda>)
        (2): Activation(relu)
      )
      (2): HybridSequential(
        (0): Conv1D(300 -> 100, kernel_size=(5,), stride=(1,))
        (1): HybridLambda(<lambda>)
        (2): Activation(relu)
      )
    )
  )
  (output): HybridSequential(
    (0): Dropout(p = 0.5, axes=())
    (1): Dense(None -> 2, linear)
  )
)
[Epoch 0 Batch 30/162] avg loss 0.0139927, throughput 0.563051K wps
[Epoch 0 Batch 60/162] avg loss 0.0138302, throughput 9.52509K wps
[Epoch 0 Batch 90/162] avg loss 0.0137475, throughput 9.45732K wps
[Epoch 0 Batch 120/162] avg loss 0.0136195, throughput 9.51905K wps
[Epoch 0 Batch 150/162] avg loss 0.0135821, throughput 9.46978K wps
Begin Testing...
[Epoch 0] train avg loss 0.0137352, dev acc 0.7733, dev avg loss 0.662958, throughput 2.41023K wps
Observed Improvement.
Begin Testing...
[Epoch 1 Batch 30/162] avg loss 0.0133297, throughput 9.67788K wps
[Epoch 1 Batch 60/162] avg loss 0.0131657, throughput 9.37649K wps
[Epoch 1 Batch 90/162] avg loss 0.0131255, throughput 9.49583K wps
[Epoch 1 Batch 120/162] avg loss 0.0129709, throughput 9.35736K wps
[Epoch 1 Batch 150/162] avg loss 0.0129759, throughput 9.46346K wps
Begin Testing...
[Epoch 1] train avg loss 0.013072, dev acc 0.6622, dev avg loss 0.63981, throughput 9.48049K wps
[Epoch 2 Batch 30/162] avg loss 0.0128046, throughput 9.62158K wps
[Epoch 2 Batch 60/162] avg loss 0.0125604, throughput 9.39331K wps
[Epoch 2 Batch 90/162] avg loss 0.0124079, throughput 9.55848K wps
[Epoch 2 Batch 120/162] avg loss 0.0122335, throughput 9.54988K wps
[Epoch 2 Batch 150/162] avg loss 0.0122206, throughput 9.68006K wps
Begin Testing...
[Epoch 2] train avg loss 0.0124286, dev acc 0.8478, dev avg loss 0.600085, throughput 9.55656K wps
Observed Improvement.
Begin Testing...
[Epoch 3 Batch 30/162] avg loss 0.0119629, throughput 9.74152K wps
[Epoch 3 Batch 60/162] avg loss 0.0119284, throughput 9.48386K wps
[Epoch 3 Batch 90/162] avg loss 0.0117038, throughput 9.53708K wps
[Epoch 3 Batch 120/162] avg loss 0.0116698, throughput 9.333K wps
[Epoch 3 Batch 150/162] avg loss 0.0113905, throughput 9.67526K wps
Begin Testing...
[Epoch 3] train avg loss 0.011693, dev acc 0.8567, dev avg loss 0.564077, throughput 9.52981K wps
Observed Improvement.
Begin Testing...
[Epoch 4 Batch 30/162] avg loss 0.0111505, throughput 9.79962K wps
[Epoch 4 Batch 60/162] avg loss 0.0109733, throughput 9.69988K wps
[Epoch 4 Batch 90/162] avg loss 0.0109122, throughput 9.54538K wps
[Epoch 4 Batch 120/162] avg loss 0.01089, throughput 9.43827K wps
[Epoch 4 Batch 150/162] avg loss 0.0107541, throughput 9.62589K wps
Begin Testing...
[Epoch 4] train avg loss 0.0108868, dev acc 0.8678, dev avg loss 0.522418, throughput 9.6274K wps
Observed Improvement.
Begin Testing...
[Epoch 5 Batch 30/162] avg loss 0.0104188, throughput 9.71097K wps
[Epoch 5 Batch 60/162] avg loss 0.0102718, throughput 9.58992K wps
[Epoch 5 Batch 90/162] avg loss 0.00994886, throughput 9.49034K wps
[Epoch 5 Batch 120/162] avg loss 0.00980264, throughput 9.46891K wps
[Epoch 5 Batch 150/162] avg loss 0.00998378, throughput 9.70823K wps
Begin Testing...
[Epoch 5] train avg loss 0.0100641, dev acc 0.8722, dev avg loss 0.482372, throughput 9.60097K wps
Observed Improvement.
Begin Testing...
[Epoch 6 Batch 30/162] avg loss 0.00960078, throughput 9.66465K wps
[Epoch 6 Batch 60/162] avg loss 0.00951817, throughput 9.49316K wps
[Epoch 6 Batch 90/162] avg loss 0.00915172, throughput 9.61593K wps
[Epoch 6 Batch 120/162] avg loss 0.00909617, throughput 9.456K wps
[Epoch 6 Batch 150/162] avg loss 0.00920177, throughput 9.35415K wps
Begin Testing...
[Epoch 6] train avg loss 0.00926776, dev acc 0.8689, dev avg loss 0.447398, throughput 9.5075K wps
[Epoch 7 Batch 30/162] avg loss 0.00898816, throughput 9.71807K wps
[Epoch 7 Batch 60/162] avg loss 0.00859521, throughput 9.54446K wps
[Epoch 7 Batch 90/162] avg loss 0.00854635, throughput 9.38923K wps
[Epoch 7 Batch 120/162] avg loss 0.0082965, throughput 9.61768K wps
[Epoch 7 Batch 150/162] avg loss 0.00855405, throughput 9.64452K wps
Begin Testing...
[Epoch 7] train avg loss 0.00859653, dev acc 0.8722, dev avg loss 0.417487, throughput 9.58623K wps
Observed Improvement.
Begin Testing...
[Epoch 8 Batch 30/162] avg loss 0.00807569, throughput 9.63364K wps
[Epoch 8 Batch 60/162] avg loss 0.0081229, throughput 9.54556K wps
[Epoch 8 Batch 90/162] avg loss 0.00795082, throughput 9.47762K wps
[Epoch 8 Batch 120/162] avg loss 0.00809858, throughput 9.34018K wps
[Epoch 8 Batch 150/162] avg loss 0.00807176, throughput 9.64029K wps
Begin Testing...
[Epoch 8] train avg loss 0.00802045, dev acc 0.8700, dev avg loss 0.394164, throughput 9.51211K wps
[Epoch 9 Batch 30/162] avg loss 0.00765048, throughput 9.68748K wps
[Epoch 9 Batch 60/162] avg loss 0.00766964, throughput 9.60219K wps
[Epoch 9 Batch 90/162] avg loss 0.00778438, throughput 9.46143K wps
[Epoch 9 Batch 120/162] avg loss 0.00752026, throughput 9.51964K wps
[Epoch 9 Batch 150/162] avg loss 0.0073391, throughput 9.57476K wps
Begin Testing...
[Epoch 9] train avg loss 0.0075777, dev acc 0.8733, dev avg loss 0.376142, throughput 9.548K wps
Observed Improvement.
Begin Testing...
[Epoch 10 Batch 30/162] avg loss 0.00717406, throughput 9.68558K wps
[Epoch 10 Batch 60/162] avg loss 0.00721908, throughput 9.46155K wps
[Epoch 10 Batch 90/162] avg loss 0.00730357, throughput 9.58996K wps
[Epoch 10 Batch 120/162] avg loss 0.00733003, throughput 9.44458K wps
[Epoch 10 Batch 150/162] avg loss 0.00701399, throughput 9.40509K wps
Begin Testing...
[Epoch 10] train avg loss 0.00718916, dev acc 0.8756, dev avg loss 0.363214, throughput 9.51757K wps
Observed Improvement.
Begin Testing...
[Epoch 11 Batch 30/162] avg loss 0.00678275, throughput 9.54591K wps
[Epoch 11 Batch 60/162] avg loss 0.00700366, throughput 9.4893K wps
[Epoch 11 Batch 90/162] avg loss 0.00689843, throughput 9.4057K wps
[Epoch 11 Batch 120/162] avg loss 0.0070885, throughput 9.4621K wps
[Epoch 11 Batch 150/162] avg loss 0.0067608, throughput 9.6145K wps
Begin Testing...
[Epoch 11] train avg loss 0.00691345, dev acc 0.8778, dev avg loss 0.352661, throughput 9.48861K wps
Observed Improvement.
Begin Testing...
[Epoch 12 Batch 30/162] avg loss 0.00683382, throughput 9.78665K wps
[Epoch 12 Batch 60/162] avg loss 0.00662973, throughput 9.41849K wps
[Epoch 12 Batch 90/162] avg loss 0.0062095, throughput 9.52567K wps
[Epoch 12 Batch 120/162] avg loss 0.00622745, throughput 9.63293K wps
[Epoch 12 Batch 150/162] avg loss 0.00668349, throughput 9.51729K wps
Begin Testing...
[Epoch 12] train avg loss 0.00650781, dev acc 0.8811, dev avg loss 0.341088, throughput 9.57071K wps
Observed Improvement.
Begin Testing...
[Epoch 13 Batch 30/162] avg loss 0.0066457, throughput 9.55024K wps
[Epoch 13 Batch 60/162] avg loss 0.0066435, throughput 9.52019K wps
[Epoch 13 Batch 90/162] avg loss 0.00622139, throughput 9.42255K wps
[Epoch 13 Batch 120/162] avg loss 0.00649197, throughput 9.44282K wps
[Epoch 13 Batch 150/162] avg loss 0.00611179, throughput 9.57489K wps
Begin Testing...
[Epoch 13] train avg loss 0.00644013, dev acc 0.8856, dev avg loss 0.333625, throughput 9.49058K wps
Observed Improvement.
Begin Testing...
[Epoch 14 Batch 30/162] avg loss 0.00632254, throughput 9.62778K wps
[Epoch 14 Batch 60/162] avg loss 0.00630582, throughput 9.36483K wps
[Epoch 14 Batch 90/162] avg loss 0.00641791, throughput 9.35305K wps
[Epoch 14 Batch 120/162] avg loss 0.00604265, throughput 9.41917K wps
[Epoch 14 Batch 150/162] avg loss 0.00612689, throughput 9.57848K wps
Begin Testing...
[Epoch 14] train avg loss 0.00624127, dev acc 0.8867, dev avg loss 0.326977, throughput 9.44655K wps
Observed Improvement.
Begin Testing...
[Epoch 15 Batch 30/162] avg loss 0.00631406, throughput 9.60864K wps
[Epoch 15 Batch 60/162] avg loss 0.00592309, throughput 9.57579K wps
[Epoch 15 Batch 90/162] avg loss 0.00617731, throughput 9.50213K wps
[Epoch 15 Batch 120/162] avg loss 0.00596338, throughput 9.37959K wps
[Epoch 15 Batch 150/162] avg loss 0.00599038, throughput 9.4467K wps
Begin Testing...
[Epoch 15] train avg loss 0.00605979, dev acc 0.8889, dev avg loss 0.320968, throughput 9.48432K wps
Observed Improvement.
Begin Testing...
[Epoch 16 Batch 30/162] avg loss 0.00569059, throughput 9.56589K wps
[Epoch 16 Batch 60/162] avg loss 0.00582947, throughput 9.40569K wps
[Epoch 16 Batch 90/162] avg loss 0.0062145, throughput 9.26223K wps
[Epoch 16 Batch 120/162] avg loss 0.00600952, throughput 9.43629K wps
[Epoch 16 Batch 150/162] avg loss 0.00606738, throughput 9.49933K wps
Begin Testing...
[Epoch 16] train avg loss 0.00593747, dev acc 0.8867, dev avg loss 0.315545, throughput 9.42478K wps
[Epoch 17 Batch 30/162] avg loss 0.00557341, throughput 9.58483K wps
[Epoch 17 Batch 60/162] avg loss 0.00592358, throughput 9.39695K wps
[Epoch 17 Batch 90/162] avg loss 0.00622066, throughput 9.32288K wps
[Epoch 17 Batch 120/162] avg loss 0.00554141, throughput 9.35937K wps
[Epoch 17 Batch 150/162] avg loss 0.00607266, throughput 9.64441K wps
Begin Testing...
[Epoch 17] train avg loss 0.00583486, dev acc 0.8889, dev avg loss 0.311482, throughput 9.46071K wps
Observed Improvement.
Begin Testing...
[Epoch 18 Batch 30/162] avg loss 0.00559593, throughput 9.73618K wps
[Epoch 18 Batch 60/162] avg loss 0.00565354, throughput 9.46467K wps
[Epoch 18 Batch 90/162] avg loss 0.00555657, throughput 9.26849K wps
[Epoch 18 Batch 120/162] avg loss 0.00615792, throughput 9.34718K wps
[Epoch 18 Batch 150/162] avg loss 0.00531747, throughput 9.45421K wps
Begin Testing...
[Epoch 18] train avg loss 0.0056676, dev acc 0.8889, dev avg loss 0.306577, throughput 9.44157K wps
Observed Improvement.
Begin Testing...
[Epoch 19 Batch 30/162] avg loss 0.00553821, throughput 9.64454K wps
[Epoch 19 Batch 60/162] avg loss 0.00530412, throughput 9.39186K wps
[Epoch 19 Batch 90/162] avg loss 0.00555156, throughput 9.39406K wps
[Epoch 19 Batch 120/162] avg loss 0.00581841, throughput 9.43238K wps
[Epoch 19 Batch 150/162] avg loss 0.00549919, throughput 9.34169K wps
Begin Testing...
[Epoch 19] train avg loss 0.00553728, dev acc 0.8911, dev avg loss 0.302825, throughput 9.43016K wps
Observed Improvement.
Begin Testing...
[Epoch 20 Batch 30/162] avg loss 0.00521793, throughput 9.78528K wps
[Epoch 20 Batch 60/162] avg loss 0.00523199, throughput 9.59378K wps
[Epoch 20 Batch 90/162] avg loss 0.00544018, throughput 9.55586K wps
[Epoch 20 Batch 120/162] avg loss 0.00572085, throughput 9.62187K wps
[Epoch 20 Batch 150/162] avg loss 0.00584005, throughput 9.55164K wps
Begin Testing...
[Epoch 20] train avg loss 0.00545546, dev acc 0.8922, dev avg loss 0.298785, throughput 9.61922K wps
Observed Improvement.
Begin Testing...
[Epoch 21 Batch 30/162] avg loss 0.00563956, throughput 9.55574K wps
[Epoch 21 Batch 60/162] avg loss 0.00519013, throughput 9.38313K wps
[Epoch 21 Batch 90/162] avg loss 0.0051913, throughput 9.47522K wps
[Epoch 21 Batch 120/162] avg loss 0.00510079, throughput 9.50785K wps
[Epoch 21 Batch 150/162] avg loss 0.00502946, throughput 9.27265K wps
Begin Testing...
[Epoch 21] train avg loss 0.00524858, dev acc 0.8956, dev avg loss 0.295013, throughput 9.43116K wps
Observed Improvement.
Begin Testing...
[Epoch 22 Batch 30/162] avg loss 0.0052942, throughput 9.69388K wps
[Epoch 22 Batch 60/162] avg loss 0.00522257, throughput 9.48309K wps
[Epoch 22 Batch 90/162] avg loss 0.00504285, throughput 9.44366K wps
[Epoch 22 Batch 120/162] avg loss 0.00495531, throughput 9.39026K wps
[Epoch 22 Batch 150/162] avg loss 0.00539518, throughput 9.52251K wps
Begin Testing...
[Epoch 22] train avg loss 0.00516369, dev acc 0.8967, dev avg loss 0.291609, throughput 9.51514K wps
Observed Improvement.
Begin Testing...
[Epoch 23 Batch 30/162] avg loss 0.00514018, throughput 9.64495K wps
[Epoch 23 Batch 60/162] avg loss 0.00501121, throughput 9.4804K wps
[Epoch 23 Batch 90/162] avg loss 0.00491868, throughput 9.47652K wps
[Epoch 23 Batch 120/162] avg loss 0.00524275, throughput 9.41957K wps
[Epoch 23 Batch 150/162] avg loss 0.0052641, throughput 9.45964K wps
Begin Testing...
[Epoch 23] train avg loss 0.00508836, dev acc 0.8922, dev avg loss 0.289848, throughput 9.48341K wps
[Epoch 24 Batch 30/162] avg loss 0.00510846, throughput 9.47826K wps
[Epoch 24 Batch 60/162] avg loss 0.00495213, throughput 9.58132K wps
[Epoch 24 Batch 90/162] avg loss 0.00466404, throughput 9.34193K wps
[Epoch 24 Batch 120/162] avg loss 0.00530204, throughput 9.60358K wps
[Epoch 24 Batch 150/162] avg loss 0.0047967, throughput 9.47435K wps
Begin Testing...
[Epoch 24] train avg loss 0.00494935, dev acc 0.8956, dev avg loss 0.285739, throughput 9.50178K wps
[Epoch 25 Batch 30/162] avg loss 0.00484253, throughput 9.66595K wps
[Epoch 25 Batch 60/162] avg loss 0.00514943, throughput 9.43817K wps
[Epoch 25 Batch 90/162] avg loss 0.00455663, throughput 9.34473K wps
[Epoch 25 Batch 120/162] avg loss 0.00507725, throughput 9.44696K wps
[Epoch 25 Batch 150/162] avg loss 0.00469302, throughput 9.48272K wps
Begin Testing...
[Epoch 25] train avg loss 0.00487417, dev acc 0.8956, dev avg loss 0.283306, throughput 9.47791K wps
[Epoch 26 Batch 30/162] avg loss 0.00503558, throughput 9.49072K wps
[Epoch 26 Batch 60/162] avg loss 0.00440408, throughput 9.44584K wps
[Epoch 26 Batch 90/162] avg loss 0.00464679, throughput 9.52314K wps
[Epoch 26 Batch 120/162] avg loss 0.00515276, throughput 9.33639K wps
[Epoch 26 Batch 150/162] avg loss 0.00482135, throughput 9.45824K wps
Begin Testing...
[Epoch 26] train avg loss 0.00482186, dev acc 0.8978, dev avg loss 0.280569, throughput 9.45502K wps
Observed Improvement.
Begin Testing...
[Epoch 27 Batch 30/162] avg loss 0.0045356, throughput 9.64995K wps
[Epoch 27 Batch 60/162] avg loss 0.00455278, throughput 9.36593K wps
[Epoch 27 Batch 90/162] avg loss 0.00482514, throughput 9.35764K wps
[Epoch 27 Batch 120/162] avg loss 0.00492322, throughput 9.25941K wps
[Epoch 27 Batch 150/162] avg loss 0.00494774, throughput 9.46653K wps
Begin Testing...
[Epoch 27] train avg loss 0.00475806, dev acc 0.8967, dev avg loss 0.2784, throughput 9.40425K wps
[Epoch 28 Batch 30/162] avg loss 0.00434146, throughput 9.67811K wps
[Epoch 28 Batch 60/162] avg loss 0.00488754, throughput 9.60404K wps
[Epoch 28 Batch 90/162] avg loss 0.00423977, throughput 9.34693K wps
[Epoch 28 Batch 120/162] avg loss 0.00503577, throughput 9.59636K wps
[Epoch 28 Batch 150/162] avg loss 0.00460634, throughput 9.33307K wps
Begin Testing...
[Epoch 28] train avg loss 0.00464542, dev acc 0.8978, dev avg loss 0.275934, throughput 9.50684K wps
Observed Improvement.
Begin Testing...
[Epoch 29 Batch 30/162] avg loss 0.00482399, throughput 9.56692K wps
[Epoch 29 Batch 60/162] avg loss 0.00454589, throughput 9.25683K wps
[Epoch 29 Batch 90/162] avg loss 0.00435749, throughput 9.57799K wps
[Epoch 29 Batch 120/162] avg loss 0.00478881, throughput 9.4965K wps
[Epoch 29 Batch 150/162] avg loss 0.00409712, throughput 9.51701K wps
Begin Testing...
[Epoch 29] train avg loss 0.00448683, dev acc 0.8978, dev avg loss 0.273418, throughput 9.4819K wps
Observed Improvement.
Begin Testing...
[Epoch 30 Batch 30/162] avg loss 0.00478067, throughput 9.80287K wps
[Epoch 30 Batch 60/162] avg loss 0.00419504, throughput 9.3498K wps
[Epoch 30 Batch 90/162] avg loss 0.00431353, throughput 9.42033K wps
[Epoch 30 Batch 120/162] avg loss 0.00470617, throughput 9.29221K wps
[Epoch 30 Batch 150/162] avg loss 0.00430479, throughput 9.39394K wps
Begin Testing...
[Epoch 30] train avg loss 0.00446317, dev acc 0.9000, dev avg loss 0.272378, throughput 9.45538K wps
Observed Improvement.
Begin Testing...
[Epoch 31 Batch 30/162] avg loss 0.00446604, throughput 9.61072K wps
[Epoch 31 Batch 60/162] avg loss 0.00452568, throughput 9.51161K wps
[Epoch 31 Batch 90/162] avg loss 0.00415589, throughput 9.45126K wps
[Epoch 31 Batch 120/162] avg loss 0.00450758, throughput 9.564K wps
[Epoch 31 Batch 150/162] avg loss 0.00432665, throughput 9.42179K wps
Begin Testing...
[Epoch 31] train avg loss 0.00440475, dev acc 0.8967, dev avg loss 0.269901, throughput 9.51569K wps
[Epoch 32 Batch 30/162] avg loss 0.00433878, throughput 9.51472K wps
[Epoch 32 Batch 60/162] avg loss 0.00443535, throughput 9.47296K wps
[Epoch 32 Batch 90/162] avg loss 0.00449091, throughput 9.45848K wps
[Epoch 32 Batch 120/162] avg loss 0.00428892, throughput 9.28994K wps
[Epoch 32 Batch 150/162] avg loss 0.00416887, throughput 9.31816K wps
Begin Testing...
[Epoch 32] train avg loss 0.00433306, dev acc 0.8989, dev avg loss 0.268951, throughput 9.4196K wps
[Epoch 33 Batch 30/162] avg loss 0.00421211, throughput 9.58092K wps
[Epoch 33 Batch 60/162] avg loss 0.00408083, throughput 9.63121K wps
[Epoch 33 Batch 90/162] avg loss 0.00463901, throughput 9.34245K wps
[Epoch 33 Batch 120/162] avg loss 0.0044146, throughput 9.38966K wps
[Epoch 33 Batch 150/162] avg loss 0.00431222, throughput 9.31115K wps
Begin Testing...
[Epoch 33] train avg loss 0.00428337, dev acc 0.8989, dev avg loss 0.268036, throughput 9.46413K wps
[Epoch 34 Batch 30/162] avg loss 0.00408815, throughput 9.47671K wps
[Epoch 34 Batch 60/162] avg loss 0.00419916, throughput 9.39808K wps
[Epoch 34 Batch 90/162] avg loss 0.00407953, throughput 9.35001K wps
[Epoch 34 Batch 120/162] avg loss 0.0037462, throughput 9.5343K wps
[Epoch 34 Batch 150/162] avg loss 0.00441781, throughput 9.38551K wps
Begin Testing...
[Epoch 34] train avg loss 0.00414092, dev acc 0.8989, dev avg loss 0.264133, throughput 9.41777K wps
[Epoch 35 Batch 30/162] avg loss 0.00430011, throughput 9.60308K wps
[Epoch 35 Batch 60/162] avg loss 0.0040314, throughput 9.43085K wps
[Epoch 35 Batch 90/162] avg loss 0.00414802, throughput 9.59724K wps
[Epoch 35 Batch 120/162] avg loss 0.00419202, throughput 9.41857K wps
[Epoch 35 Batch 150/162] avg loss 0.00377472, throughput 9.55167K wps
Begin Testing...
[Epoch 35] train avg loss 0.00408505, dev acc 0.8978, dev avg loss 0.262154, throughput 9.52085K wps
[Epoch 36 Batch 30/162] avg loss 0.00398365, throughput 9.4616K wps
[Epoch 36 Batch 60/162] avg loss 0.00406087, throughput 9.57624K wps
[Epoch 36 Batch 90/162] avg loss 0.00419579, throughput 9.36475K wps
[Epoch 36 Batch 120/162] avg loss 0.00370225, throughput 9.47195K wps
[Epoch 36 Batch 150/162] avg loss 0.00422434, throughput 9.56823K wps
Begin Testing...
[Epoch 36] train avg loss 0.00403643, dev acc 0.9022, dev avg loss 0.26199, throughput 9.48986K wps
Observed Improvement.
Begin Testing...
[Epoch 37 Batch 30/162] avg loss 0.00369248, throughput 9.70319K wps
[Epoch 37 Batch 60/162] avg loss 0.00399778, throughput 9.23667K wps
[Epoch 37 Batch 90/162] avg loss 0.00388887, throughput 9.34555K wps
[Epoch 37 Batch 120/162] avg loss 0.0043826, throughput 9.49703K wps
[Epoch 37 Batch 150/162] avg loss 0.00427731, throughput 9.60465K wps
Begin Testing...
[Epoch 37] train avg loss 0.00402125, dev acc 0.9000, dev avg loss 0.259542, throughput 9.4719K wps
[Epoch 38 Batch 30/162] avg loss 0.00380891, throughput 9.49633K wps
[Epoch 38 Batch 60/162] avg loss 0.0040139, throughput 9.49164K wps
[Epoch 38 Batch 90/162] avg loss 0.00386191, throughput 9.35476K wps
[Epoch 38 Batch 120/162] avg loss 0.00370861, throughput 9.45103K wps
[Epoch 38 Batch 150/162] avg loss 0.0042393, throughput 9.26727K wps
Begin Testing...
[Epoch 38] train avg loss 0.0039613, dev acc 0.9011, dev avg loss 0.258285, throughput 9.40156K wps
[Epoch 39 Batch 30/162] avg loss 0.00401124, throughput 9.54127K wps
[Epoch 39 Batch 60/162] avg loss 0.00390607, throughput 9.31581K wps
[Epoch 39 Batch 90/162] avg loss 0.00373968, throughput 9.27547K wps
[Epoch 39 Batch 120/162] avg loss 0.00396615, throughput 9.41871K wps
[Epoch 39 Batch 150/162] avg loss 0.00379264, throughput 9.39576K wps
Begin Testing...
[Epoch 39] train avg loss 0.00390702, dev acc 0.8989, dev avg loss 0.256657, throughput 9.39427K wps
[Epoch 40 Batch 30/162] avg loss 0.00393162, throughput 9.67434K wps
[Epoch 40 Batch 60/162] avg loss 0.0037819, throughput 9.57699K wps
[Epoch 40 Batch 90/162] avg loss 0.00377915, throughput 9.46876K wps
[Epoch 40 Batch 120/162] avg loss 0.00373343, throughput 9.45007K wps
[Epoch 40 Batch 150/162] avg loss 0.00380252, throughput 9.26224K wps
Begin Testing...
[Epoch 40] train avg loss 0.00381492, dev acc 0.8989, dev avg loss 0.255187, throughput 9.4706K wps
[Epoch 41 Batch 30/162] avg loss 0.00399802, throughput 9.49814K wps
[Epoch 41 Batch 60/162] avg loss 0.00345699, throughput 9.29504K wps
[Epoch 41 Batch 90/162] avg loss 0.00360227, throughput 9.45748K wps
[Epoch 41 Batch 120/162] avg loss 0.00398935, throughput 9.41777K wps
[Epoch 41 Batch 150/162] avg loss 0.00378132, throughput 9.27861K wps
Begin Testing...
[Epoch 41] train avg loss 0.00374704, dev acc 0.8989, dev avg loss 0.25391, throughput 9.38096K wps
[Epoch 42 Batch 30/162] avg loss 0.00357563, throughput 9.54487K wps
[Epoch 42 Batch 60/162] avg loss 0.00361566, throughput 9.62968K wps
[Epoch 42 Batch 90/162] avg loss 0.00354063, throughput 9.49293K wps
[Epoch 42 Batch 120/162] avg loss 0.00361762, throughput 9.44148K wps
[Epoch 42 Batch 150/162] avg loss 0.00418969, throughput 9.36544K wps
Begin Testing...
[Epoch 42] train avg loss 0.00373444, dev acc 0.9000, dev avg loss 0.25275, throughput 9.50581K wps
[Epoch 43 Batch 30/162] avg loss 0.00370573, throughput 9.5242K wps
[Epoch 43 Batch 60/162] avg loss 0.00373772, throughput 9.43488K wps
[Epoch 43 Batch 90/162] avg loss 0.00357915, throughput 9.30858K wps
[Epoch 43 Batch 120/162] avg loss 0.00372048, throughput 9.3571K wps
[Epoch 43 Batch 150/162] avg loss 0.00356147, throughput 9.36751K wps
Begin Testing...
[Epoch 43] train avg loss 0.00365763, dev acc 0.9011, dev avg loss 0.251234, throughput 9.39331K wps
[Epoch 44 Batch 30/162] avg loss 0.00378978, throughput 9.61609K wps
[Epoch 44 Batch 60/162] avg loss 0.00358534, throughput 9.39665K wps
[Epoch 44 Batch 90/162] avg loss 0.00380747, throughput 9.63629K wps
[Epoch 44 Batch 120/162] avg loss 0.00341006, throughput 9.35279K wps
[Epoch 44 Batch 150/162] avg loss 0.00383862, throughput 9.451K wps
Begin Testing...
[Epoch 44] train avg loss 0.00362973, dev acc 0.9011, dev avg loss 0.250194, throughput 9.48177K wps
[Epoch 45 Batch 30/162] avg loss 0.00359368, throughput 9.65537K wps
[Epoch 45 Batch 60/162] avg loss 0.00378111, throughput 9.51237K wps
[Epoch 45 Batch 90/162] avg loss 0.00363177, throughput 9.39398K wps
[Epoch 45 Batch 120/162] avg loss 0.00371031, throughput 9.26322K wps
[Epoch 45 Batch 150/162] avg loss 0.00336906, throughput 9.49175K wps
Begin Testing...
[Epoch 45] train avg loss 0.00359808, dev acc 0.9011, dev avg loss 0.249553, throughput 9.46465K wps
[Epoch 46 Batch 30/162] avg loss 0.00344127, throughput 9.72143K wps
[Epoch 46 Batch 60/162] avg loss 0.00328195, throughput 9.34182K wps
[Epoch 46 Batch 90/162] avg loss 0.00362182, throughput 9.33831K wps
[Epoch 46 Batch 120/162] avg loss 0.00373991, throughput 9.35089K wps
[Epoch 46 Batch 150/162] avg loss 0.00354795, throughput 9.44608K wps
Begin Testing...
[Epoch 46] train avg loss 0.00350159, dev acc 0.9056, dev avg loss 0.249423, throughput 9.43889K wps
Observed Improvement.
Begin Testing...
[Epoch 47 Batch 30/162] avg loss 0.00342204, throughput 9.64541K wps
[Epoch 47 Batch 60/162] avg loss 0.00347048, throughput 9.31912K wps
[Epoch 47 Batch 90/162] avg loss 0.00336565, throughput 9.37817K wps
[Epoch 47 Batch 120/162] avg loss 0.00349036, throughput 9.57605K wps
[Epoch 47 Batch 150/162] avg loss 0.00336753, throughput 9.44081K wps
Begin Testing...
[Epoch 47] train avg loss 0.00339945, dev acc 0.9067, dev avg loss 0.248338, throughput 9.44925K wps
Observed Improvement.
Begin Testing...
[Epoch 48 Batch 30/162] avg loss 0.00350012, throughput 9.44065K wps
[Epoch 48 Batch 60/162] avg loss 0.00330553, throughput 9.36003K wps
[Epoch 48 Batch 90/162] avg loss 0.00328175, throughput 9.31135K wps
[Epoch 48 Batch 120/162] avg loss 0.00379125, throughput 9.6338K wps
[Epoch 48 Batch 150/162] avg loss 0.00353947, throughput 9.38186K wps
Begin Testing...
[Epoch 48] train avg loss 0.00342202, dev acc 0.9067, dev avg loss 0.247824, throughput 9.43725K wps
Observed Improvement.
Begin Testing...
[Epoch 49 Batch 30/162] avg loss 0.00321639, throughput 9.54198K wps
[Epoch 49 Batch 60/162] avg loss 0.00353585, throughput 9.35214K wps
[Epoch 49 Batch 90/162] avg loss 0.00326472, throughput 9.29211K wps
[Epoch 49 Batch 120/162] avg loss 0.00307165, throughput 9.37366K wps
[Epoch 49 Batch 150/162] avg loss 0.00346831, throughput 9.27105K wps
Begin Testing...
[Epoch 49] train avg loss 0.00331722, dev acc 0.9022, dev avg loss 0.245726, throughput 9.35504K wps
[Epoch 50 Batch 30/162] avg loss 0.00318394, throughput 9.53669K wps
[Epoch 50 Batch 60/162] avg loss 0.00343808, throughput 9.49831K wps
[Epoch 50 Batch 90/162] avg loss 0.00327675, throughput 9.54207K wps
[Epoch 50 Batch 120/162] avg loss 0.00381823, throughput 9.49602K wps
[Epoch 50 Batch 150/162] avg loss 0.00297331, throughput 9.33471K wps
Begin Testing...
[Epoch 50] train avg loss 0.00333092, dev acc 0.9022, dev avg loss 0.244905, throughput 9.47841K wps
[Epoch 51 Batch 30/162] avg loss 0.00321521, throughput 9.52019K wps
[Epoch 51 Batch 60/162] avg loss 0.0032855, throughput 9.41971K wps
[Epoch 51 Batch 90/162] avg loss 0.00326702, throughput 9.27154K wps
[Epoch 51 Batch 120/162] avg loss 0.00322066, throughput 9.26474K wps
[Epoch 51 Batch 150/162] avg loss 0.00306519, throughput 9.47363K wps
Begin Testing...
[Epoch 51] train avg loss 0.00322283, dev acc 0.9022, dev avg loss 0.244081, throughput 9.39416K wps
[Epoch 52 Batch 30/162] avg loss 0.00318501, throughput 9.45487K wps
[Epoch 52 Batch 60/162] avg loss 0.00311068, throughput 9.43871K wps
[Epoch 52 Batch 90/162] avg loss 0.00349898, throughput 9.52401K wps
[Epoch 52 Batch 120/162] avg loss 0.00307708, throughput 9.49713K wps
[Epoch 52 Batch 150/162] avg loss 0.00310367, throughput 9.58282K wps
Begin Testing...
[Epoch 52] train avg loss 0.00319324, dev acc 0.9078, dev avg loss 0.244634, throughput 9.4922K wps
Observed Improvement.
Begin Testing...
[Epoch 53 Batch 30/162] avg loss 0.00330412, throughput 9.5568K wps
[Epoch 53 Batch 60/162] avg loss 0.00340498, throughput 9.38188K wps
[Epoch 53 Batch 90/162] avg loss 0.002726, throughput 9.33004K wps
[Epoch 53 Batch 120/162] avg loss 0.00325923, throughput 9.30984K wps
[Epoch 53 Batch 150/162] avg loss 0.00311163, throughput 9.50999K wps
Begin Testing...
[Epoch 53] train avg loss 0.00313863, dev acc 0.9078, dev avg loss 0.24387, throughput 9.40989K wps
Observed Improvement.
Begin Testing...
[Epoch 54 Batch 30/162] avg loss 0.00280522, throughput 9.5136K wps
[Epoch 54 Batch 60/162] avg loss 0.00312256, throughput 9.34957K wps
[Epoch 54 Batch 90/162] avg loss 0.0032516, throughput 9.49484K wps
[Epoch 54 Batch 120/162] avg loss 0.00296608, throughput 9.6302K wps
[Epoch 54 Batch 150/162] avg loss 0.00336393, throughput 9.57051K wps
Begin Testing...
[Epoch 54] train avg loss 0.00309793, dev acc 0.9067, dev avg loss 0.24246, throughput 9.5052K wps
[Epoch 55 Batch 30/162] avg loss 0.00311743, throughput 9.41864K wps
[Epoch 55 Batch 60/162] avg loss 0.00315787, throughput 9.32091K wps
[Epoch 55 Batch 90/162] avg loss 0.00304349, throughput 9.47142K wps
[Epoch 55 Batch 120/162] avg loss 0.00312005, throughput 9.26735K wps
[Epoch 55 Batch 150/162] avg loss 0.00288105, throughput 9.34897K wps
Begin Testing...
[Epoch 55] train avg loss 0.00304497, dev acc 0.9033, dev avg loss 0.240875, throughput 9.3566K wps
[Epoch 56 Batch 30/162] avg loss 0.00289679, throughput 9.51227K wps
[Epoch 56 Batch 60/162] avg loss 0.00314012, throughput 9.31222K wps
[Epoch 56 Batch 90/162] avg loss 0.00300704, throughput 9.40977K wps
[Epoch 56 Batch 120/162] avg loss 0.00313969, throughput 9.2384K wps
[Epoch 56 Batch 150/162] avg loss 0.00277625, throughput 9.54468K wps
Begin Testing...
[Epoch 56] train avg loss 0.00299008, dev acc 0.9067, dev avg loss 0.240352, throughput 9.39314K wps
[Epoch 57 Batch 30/162] avg loss 0.00299225, throughput 9.46794K wps
[Epoch 57 Batch 60/162] avg loss 0.00281025, throughput 9.35416K wps
[Epoch 57 Batch 90/162] avg loss 0.00297896, throughput 9.40174K wps
[Epoch 57 Batch 120/162] avg loss 0.00293832, throughput 9.25101K wps
[Epoch 57 Batch 150/162] avg loss 0.00276375, throughput 9.43863K wps
Begin Testing...
[Epoch 57] train avg loss 0.0029357, dev acc 0.9078, dev avg loss 0.240183, throughput 9.39874K wps
Observed Improvement.
Begin Testing...
[Epoch 58 Batch 30/162] avg loss 0.00280653, throughput 9.64921K wps
[Epoch 58 Batch 60/162] avg loss 0.00301661, throughput 9.3229K wps
[Epoch 58 Batch 90/162] avg loss 0.00294062, throughput 9.38517K wps
[Epoch 58 Batch 120/162] avg loss 0.00316029, throughput 9.40372K wps
[Epoch 58 Batch 150/162] avg loss 0.00312705, throughput 9.37382K wps
Begin Testing...
[Epoch 58] train avg loss 0.00300378, dev acc 0.9100, dev avg loss 0.240917, throughput 9.41193K wps
Observed Improvement.
Begin Testing...
[Epoch 59 Batch 30/162] avg loss 0.00293071, throughput 9.59655K wps
[Epoch 59 Batch 60/162] avg loss 0.00278885, throughput 9.3282K wps
[Epoch 59 Batch 90/162] avg loss 0.00260123, throughput 9.34104K wps
[Epoch 59 Batch 120/162] avg loss 0.00304476, throughput 9.48451K wps
[Epoch 59 Batch 150/162] avg loss 0.00294758, throughput 9.38572K wps
Begin Testing...
[Epoch 59] train avg loss 0.00288044, dev acc 0.9044, dev avg loss 0.238944, throughput 9.42144K wps
[Epoch 60 Batch 30/162] avg loss 0.00300888, throughput 9.568K wps
[Epoch 60 Batch 60/162] avg loss 0.00291366, throughput 9.44866K wps
[Epoch 60 Batch 90/162] avg loss 0.00267981, throughput 9.24282K wps
[Epoch 60 Batch 120/162] avg loss 0.00292821, throughput 9.48347K wps
[Epoch 60 Batch 150/162] avg loss 0.00276907, throughput 9.45153K wps
Begin Testing...
[Epoch 60] train avg loss 0.00283028, dev acc 0.9056, dev avg loss 0.239464, throughput 9.42948K wps
[Epoch 61 Batch 30/162] avg loss 0.00260855, throughput 9.67308K wps
[Epoch 61 Batch 60/162] avg loss 0.00278626, throughput 9.50827K wps
[Epoch 61 Batch 90/162] avg loss 0.00290986, throughput 9.53516K wps
[Epoch 61 Batch 120/162] avg loss 0.00299438, throughput 9.45582K wps
[Epoch 61 Batch 150/162] avg loss 0.00274816, throughput 9.48674K wps
Begin Testing...
[Epoch 61] train avg loss 0.00281143, dev acc 0.9056, dev avg loss 0.237615, throughput 9.5077K wps
[Epoch 62 Batch 30/162] avg loss 0.00288795, throughput 9.4864K wps
[Epoch 62 Batch 60/162] avg loss 0.00276271, throughput 9.33604K wps
[Epoch 62 Batch 90/162] avg loss 0.00266529, throughput 9.41457K wps
[Epoch 62 Batch 120/162] avg loss 0.00267132, throughput 9.4805K wps
[Epoch 62 Batch 150/162] avg loss 0.00278233, throughput 9.42662K wps
Begin Testing...
[Epoch 62] train avg loss 0.002759, dev acc 0.9056, dev avg loss 0.23689, throughput 9.42278K wps
[Epoch 63 Batch 30/162] avg loss 0.00265385, throughput 9.74434K wps
[Epoch 63 Batch 60/162] avg loss 0.00286414, throughput 9.40379K wps
[Epoch 63 Batch 90/162] avg loss 0.0028863, throughput 9.39386K wps
[Epoch 63 Batch 120/162] avg loss 0.00272298, throughput 9.50042K wps
[Epoch 63 Batch 150/162] avg loss 0.00257107, throughput 9.36452K wps
Begin Testing...
[Epoch 63] train avg loss 0.00273231, dev acc 0.9078, dev avg loss 0.23665, throughput 9.49075K wps
[Epoch 64 Batch 30/162] avg loss 0.00287369, throughput 9.70174K wps
[Epoch 64 Batch 60/162] avg loss 0.00268595, throughput 9.32407K wps
[Epoch 64 Batch 90/162] avg loss 0.00249368, throughput 9.40545K wps
[Epoch 64 Batch 120/162] avg loss 0.00279411, throughput 9.55831K wps
[Epoch 64 Batch 150/162] avg loss 0.00262947, throughput 9.50336K wps
Begin Testing...
[Epoch 64] train avg loss 0.00268018, dev acc 0.9067, dev avg loss 0.236, throughput 9.49381K wps
[Epoch 65 Batch 30/162] avg loss 0.00239457, throughput 9.43358K wps
[Epoch 65 Batch 60/162] avg loss 0.00267318, throughput 9.31637K wps
[Epoch 65 Batch 90/162] avg loss 0.00275123, throughput 9.47293K wps
[Epoch 65 Batch 120/162] avg loss 0.00294261, throughput 9.32146K wps
[Epoch 65 Batch 150/162] avg loss 0.00245131, throughput 9.52185K wps
Begin Testing...
[Epoch 65] train avg loss 0.00264365, dev acc 0.9089, dev avg loss 0.235717, throughput 9.40323K wps
[Epoch 66 Batch 30/162] avg loss 0.00268589, throughput 9.51885K wps
[Epoch 66 Batch 60/162] avg loss 0.00272469, throughput 9.6092K wps
[Epoch 66 Batch 90/162] avg loss 0.00248978, throughput 9.60937K wps
[Epoch 66 Batch 120/162] avg loss 0.00285202, throughput 9.5301K wps
[Epoch 66 Batch 150/162] avg loss 0.00249202, throughput 9.27504K wps
Begin Testing...
[Epoch 66] train avg loss 0.00264761, dev acc 0.9078, dev avg loss 0.235078, throughput 9.4942K wps
[Epoch 67 Batch 30/162] avg loss 0.00227765, throughput 9.60786K wps
[Epoch 67 Batch 60/162] avg loss 0.00280586, throughput 9.40882K wps
[Epoch 67 Batch 90/162] avg loss 0.00261633, throughput 9.45542K wps
[Epoch 67 Batch 120/162] avg loss 0.00236493, throughput 9.3603K wps
[Epoch 67 Batch 150/162] avg loss 0.00251714, throughput 9.48641K wps
Begin Testing...
[Epoch 67] train avg loss 0.0025287, dev acc 0.9078, dev avg loss 0.235218, throughput 9.47176K wps
[Epoch 68 Batch 30/162] avg loss 0.00247929, throughput 9.50213K wps
[Epoch 68 Batch 60/162] avg loss 0.00237766, throughput 9.40113K wps
[Epoch 68 Batch 90/162] avg loss 0.00258673, throughput 9.31912K wps
[Epoch 68 Batch 120/162] avg loss 0.00243504, throughput 9.55616K wps
[Epoch 68 Batch 150/162] avg loss 0.00259164, throughput 9.27384K wps
Begin Testing...
[Epoch 68] train avg loss 0.00252898, dev acc 0.9078, dev avg loss 0.235461, throughput 9.42179K wps
[Epoch 69 Batch 30/162] avg loss 0.00265411, throughput 9.56154K wps
[Epoch 69 Batch 60/162] avg loss 0.00241417, throughput 9.462K wps
[Epoch 69 Batch 90/162] avg loss 0.00254495, throughput 9.43332K wps
[Epoch 69 Batch 120/162] avg loss 0.0021794, throughput 9.3068K wps
[Epoch 69 Batch 150/162] avg loss 0.00265025, throughput 9.39508K wps
Begin Testing...
[Epoch 69] train avg loss 0.00246926, dev acc 0.9078, dev avg loss 0.235717, throughput 9.42618K wps
[Epoch 70 Batch 30/162] avg loss 0.00257705, throughput 9.51677K wps
[Epoch 70 Batch 60/162] avg loss 0.00226928, throughput 9.36987K wps
[Epoch 70 Batch 90/162] avg loss 0.00249652, throughput 9.38456K wps
[Epoch 70 Batch 120/162] avg loss 0.00211324, throughput 9.50327K wps
[Epoch 70 Batch 150/162] avg loss 0.00271965, throughput 9.46819K wps
Begin Testing...
[Epoch 70] train avg loss 0.00242111, dev acc 0.9089, dev avg loss 0.233756, throughput 9.43801K wps
[Epoch 71 Batch 30/162] avg loss 0.00235651, throughput 9.51452K wps
[Epoch 71 Batch 60/162] avg loss 0.00242878, throughput 9.47533K wps
[Epoch 71 Batch 90/162] avg loss 0.00259031, throughput 9.4441K wps
[Epoch 71 Batch 120/162] avg loss 0.00239291, throughput 9.36571K wps
[Epoch 71 Batch 150/162] avg loss 0.00234223, throughput 9.32055K wps
Begin Testing...
[Epoch 71] train avg loss 0.00241264, dev acc 0.9089, dev avg loss 0.233406, throughput 9.4133K wps
[Epoch 72 Batch 30/162] avg loss 0.00242557, throughput 9.51617K wps
[Epoch 72 Batch 60/162] avg loss 0.00242044, throughput 9.2519K wps
[Epoch 72 Batch 90/162] avg loss 0.00240604, throughput 9.48976K wps
[Epoch 72 Batch 120/162] avg loss 0.00222254, throughput 9.34074K wps
[Epoch 72 Batch 150/162] avg loss 0.00241634, throughput 9.24993K wps
Begin Testing...
[Epoch 72] train avg loss 0.00238551, dev acc 0.9078, dev avg loss 0.232928, throughput 9.37433K wps
[Epoch 73 Batch 30/162] avg loss 0.00225104, throughput 9.69523K wps
[Epoch 73 Batch 60/162] avg loss 0.00251166, throughput 9.39546K wps
[Epoch 73 Batch 90/162] avg loss 0.00224245, throughput 9.55207K wps
[Epoch 73 Batch 120/162] avg loss 0.00242628, throughput 9.45621K wps
[Epoch 73 Batch 150/162] avg loss 0.00239749, throughput 9.45189K wps
Begin Testing...
[Epoch 73] train avg loss 0.00236795, dev acc 0.9089, dev avg loss 0.232871, throughput 9.4951K wps
[Epoch 74 Batch 30/162] avg loss 0.00213699, throughput 9.43431K wps
[Epoch 74 Batch 60/162] avg loss 0.00242813, throughput 9.41929K wps
[Epoch 74 Batch 90/162] avg loss 0.00227872, throughput 9.30512K wps
[Epoch 74 Batch 120/162] avg loss 0.00230027, throughput 9.32665K wps
[Epoch 74 Batch 150/162] avg loss 0.00216548, throughput 9.6628K wps
Begin Testing...
[Epoch 74] train avg loss 0.00230945, dev acc 0.9100, dev avg loss 0.233672, throughput 9.44452K wps
Observed Improvement.
Begin Testing...
[Epoch 75 Batch 30/162] avg loss 0.00229099, throughput 9.66157K wps
[Epoch 75 Batch 60/162] avg loss 0.00231805, throughput 9.19063K wps
[Epoch 75 Batch 90/162] avg loss 0.00238184, throughput 9.45356K wps
[Epoch 75 Batch 120/162] avg loss 0.00212968, throughput 9.50277K wps
[Epoch 75 Batch 150/162] avg loss 0.00231665, throughput 9.42112K wps
Begin Testing...
[Epoch 75] train avg loss 0.00234109, dev acc 0.9067, dev avg loss 0.232223, throughput 9.43704K wps
[Epoch 76 Batch 30/162] avg loss 0.00228757, throughput 9.60872K wps
[Epoch 76 Batch 60/162] avg loss 0.00238633, throughput 9.20488K wps
[Epoch 76 Batch 90/162] avg loss 0.00226496, throughput 9.28966K wps
[Epoch 76 Batch 120/162] avg loss 0.00231158, throughput 9.40825K wps
[Epoch 76 Batch 150/162] avg loss 0.0021501, throughput 9.45735K wps
Begin Testing...
[Epoch 76] train avg loss 0.00224102, dev acc 0.9067, dev avg loss 0.232594, throughput 9.37968K wps
[Epoch 77 Batch 30/162] avg loss 0.00203727, throughput 9.46889K wps
[Epoch 77 Batch 60/162] avg loss 0.00236498, throughput 9.41467K wps
[Epoch 77 Batch 90/162] avg loss 0.00234807, throughput 9.31376K wps
[Epoch 77 Batch 120/162] avg loss 0.00238359, throughput 9.66378K wps
[Epoch 77 Batch 150/162] avg loss 0.00216933, throughput 9.54635K wps
Begin Testing...
[Epoch 77] train avg loss 0.00222111, dev acc 0.9078, dev avg loss 0.232175, throughput 9.46046K wps
[Epoch 78 Batch 30/162] avg loss 0.00237146, throughput 9.48115K wps
[Epoch 78 Batch 60/162] avg loss 0.00226466, throughput 9.40173K wps
[Epoch 78 Batch 90/162] avg loss 0.00214546, throughput 9.49443K wps
[Epoch 78 Batch 120/162] avg loss 0.00216933, throughput 9.43839K wps
[Epoch 78 Batch 150/162] avg loss 0.00218924, throughput 9.4294K wps
Begin Testing...
[Epoch 78] train avg loss 0.0022003, dev acc 0.9078, dev avg loss 0.231781, throughput 9.44886K wps
[Epoch 79 Batch 30/162] avg loss 0.00214779, throughput 9.63114K wps
[Epoch 79 Batch 60/162] avg loss 0.00208279, throughput 9.5244K wps
[Epoch 79 Batch 90/162] avg loss 0.00217244, throughput 9.30055K wps
[Epoch 79 Batch 120/162] avg loss 0.00213707, throughput 9.37776K wps
[Epoch 79 Batch 150/162] avg loss 0.00215754, throughput 9.31297K wps
Begin Testing...
[Epoch 79] train avg loss 0.00213948, dev acc 0.9078, dev avg loss 0.231746, throughput 9.43039K wps
[Epoch 80 Batch 30/162] avg loss 0.00212842, throughput 9.63843K wps
[Epoch 80 Batch 60/162] avg loss 0.00195101, throughput 9.35867K wps
[Epoch 80 Batch 90/162] avg loss 0.00209711, throughput 9.50732K wps
[Epoch 80 Batch 120/162] avg loss 0.00241555, throughput 9.50327K wps
[Epoch 80 Batch 150/162] avg loss 0.00220332, throughput 9.46487K wps
Begin Testing...
[Epoch 80] train avg loss 0.00215957, dev acc 0.9089, dev avg loss 0.231133, throughput 9.49374K wps
[Epoch 81 Batch 30/162] avg loss 0.0018672, throughput 9.54823K wps
[Epoch 81 Batch 60/162] avg loss 0.00219145, throughput 9.32757K wps
[Epoch 81 Batch 90/162] avg loss 0.00221833, throughput 9.41622K wps
[Epoch 81 Batch 120/162] avg loss 0.00197305, throughput 9.29481K wps
[Epoch 81 Batch 150/162] avg loss 0.00201746, throughput 9.45812K wps
Begin Testing...
[Epoch 81] train avg loss 0.00207677, dev acc 0.9100, dev avg loss 0.230953, throughput 9.37993K wps
Observed Improvement.
Begin Testing...
[Epoch 82 Batch 30/162] avg loss 0.00188539, throughput 9.65986K wps
[Epoch 82 Batch 60/162] avg loss 0.00204478, throughput 9.46403K wps
[Epoch 82 Batch 90/162] avg loss 0.00201831, throughput 9.2557K wps
[Epoch 82 Batch 120/162] avg loss 0.00241526, throughput 9.48381K wps
[Epoch 82 Batch 150/162] avg loss 0.0020828, throughput 9.29595K wps
Begin Testing...
[Epoch 82] train avg loss 0.00207981, dev acc 0.9078, dev avg loss 0.230809, throughput 9.44472K wps
[Epoch 83 Batch 30/162] avg loss 0.00223152, throughput 9.45692K wps
[Epoch 83 Batch 60/162] avg loss 0.00222273, throughput 9.41858K wps
[Epoch 83 Batch 90/162] avg loss 0.00200768, throughput 9.38583K wps
[Epoch 83 Batch 120/162] avg loss 0.00199637, throughput 9.35457K wps
[Epoch 83 Batch 150/162] avg loss 0.00221681, throughput 9.31573K wps
Begin Testing...
[Epoch 83] train avg loss 0.00212103, dev acc 0.9100, dev avg loss 0.230458, throughput 9.40186K wps
Observed Improvement.
Begin Testing...
[Epoch 84 Batch 30/162] avg loss 0.00215412, throughput 9.72015K wps
[Epoch 84 Batch 60/162] avg loss 0.00193827, throughput 9.26361K wps
[Epoch 84 Batch 90/162] avg loss 0.0021938, throughput 9.34604K wps
[Epoch 84 Batch 120/162] avg loss 0.00206668, throughput 9.28541K wps
[Epoch 84 Batch 150/162] avg loss 0.00189408, throughput 9.25179K wps
Begin Testing...
[Epoch 84] train avg loss 0.00204826, dev acc 0.9089, dev avg loss 0.230426, throughput 9.36956K wps
[Epoch 85 Batch 30/162] avg loss 0.00185962, throughput 9.52811K wps
[Epoch 85 Batch 60/162] avg loss 0.0020107, throughput 9.42722K wps
[Epoch 85 Batch 90/162] avg loss 0.00203444, throughput 9.41173K wps
[Epoch 85 Batch 120/162] avg loss 0.00194089, throughput 9.52288K wps
[Epoch 85 Batch 150/162] avg loss 0.00204689, throughput 9.42042K wps
Begin Testing...
[Epoch 85] train avg loss 0.00200129, dev acc 0.9067, dev avg loss 0.230463, throughput 9.47302K wps
[Epoch 86 Batch 30/162] avg loss 0.00177675, throughput 9.49244K wps
[Epoch 86 Batch 60/162] avg loss 0.00202632, throughput 9.66008K wps
[Epoch 86 Batch 90/162] avg loss 0.00204808, throughput 9.38044K wps
[Epoch 86 Batch 120/162] avg loss 0.00215758, throughput 9.52754K wps
[Epoch 86 Batch 150/162] avg loss 0.00180453, throughput 9.42865K wps
Begin Testing...
[Epoch 86] train avg loss 0.00196139, dev acc 0.9078, dev avg loss 0.231065, throughput 9.48334K wps
[Epoch 87 Batch 30/162] avg loss 0.0019585, throughput 9.55588K wps
[Epoch 87 Batch 60/162] avg loss 0.00199308, throughput 9.46191K wps
[Epoch 87 Batch 90/162] avg loss 0.00180912, throughput 9.33256K wps
[Epoch 87 Batch 120/162] avg loss 0.00202829, throughput 9.53227K wps
[Epoch 87 Batch 150/162] avg loss 0.00181426, throughput 9.34087K wps
Begin Testing...
[Epoch 87] train avg loss 0.00192785, dev acc 0.9067, dev avg loss 0.230277, throughput 9.44801K wps
[Epoch 88 Batch 30/162] avg loss 0.00176082, throughput 9.42677K wps
[Epoch 88 Batch 60/162] avg loss 0.0020671, throughput 9.46952K wps
[Epoch 88 Batch 90/162] avg loss 0.0018393, throughput 9.4635K wps
[Epoch 88 Batch 120/162] avg loss 0.00189293, throughput 9.41574K wps
[Epoch 88 Batch 150/162] avg loss 0.00200162, throughput 9.41319K wps
Begin Testing...
[Epoch 88] train avg loss 0.00192076, dev acc 0.9067, dev avg loss 0.229971, throughput 9.44874K wps
[Epoch 89 Batch 30/162] avg loss 0.00196706, throughput 9.57919K wps
[Epoch 89 Batch 60/162] avg loss 0.00193885, throughput 9.35798K wps
[Epoch 89 Batch 90/162] avg loss 0.00202345, throughput 9.34419K wps
[Epoch 89 Batch 120/162] avg loss 0.00190312, throughput 9.41812K wps
[Epoch 89 Batch 150/162] avg loss 0.00173288, throughput 9.36746K wps
Begin Testing...
[Epoch 89] train avg loss 0.00191078, dev acc 0.9089, dev avg loss 0.231062, throughput 9.43163K wps
[Epoch 90 Batch 30/162] avg loss 0.001871, throughput 9.48289K wps
[Epoch 90 Batch 60/162] avg loss 0.00174085, throughput 9.46199K wps
[Epoch 90 Batch 90/162] avg loss 0.00205183, throughput 9.26592K wps
[Epoch 90 Batch 120/162] avg loss 0.0018345, throughput 9.47404K wps
[Epoch 90 Batch 150/162] avg loss 0.00207165, throughput 9.42786K wps
Begin Testing...
[Epoch 90] train avg loss 0.00190886, dev acc 0.9089, dev avg loss 0.229688, throughput 9.42216K wps
[Epoch 91 Batch 30/162] avg loss 0.00185121, throughput 9.64327K wps
[Epoch 91 Batch 60/162] avg loss 0.00188709, throughput 9.33688K wps
[Epoch 91 Batch 90/162] avg loss 0.00186984, throughput 9.43082K wps
[Epoch 91 Batch 120/162] avg loss 0.00178008, throughput 9.49693K wps
[Epoch 91 Batch 150/162] avg loss 0.00195783, throughput 9.25683K wps
Begin Testing...
[Epoch 91] train avg loss 0.00187586, dev acc 0.9100, dev avg loss 0.229561, throughput 9.43077K wps
Observed Improvement.
Begin Testing...
[Epoch 92 Batch 30/162] avg loss 0.00191878, throughput 9.58135K wps
[Epoch 92 Batch 60/162] avg loss 0.00185139, throughput 9.45728K wps
[Epoch 92 Batch 90/162] avg loss 0.00199974, throughput 9.49063K wps
[Epoch 92 Batch 120/162] avg loss 0.00185051, throughput 9.25664K wps
[Epoch 92 Batch 150/162] avg loss 0.00176807, throughput 9.41778K wps
Begin Testing...
[Epoch 92] train avg loss 0.00186376, dev acc 0.9089, dev avg loss 0.229451, throughput 9.45148K wps
[Epoch 93 Batch 30/162] avg loss 0.00169152, throughput 9.58804K wps
[Epoch 93 Batch 60/162] avg loss 0.00184662, throughput 9.37001K wps
[Epoch 93 Batch 90/162] avg loss 0.00181727, throughput 9.30156K wps
[Epoch 93 Batch 120/162] avg loss 0.00197526, throughput 9.57869K wps
[Epoch 93 Batch 150/162] avg loss 0.00155701, throughput 9.44153K wps
Begin Testing...
[Epoch 93] train avg loss 0.00177294, dev acc 0.9067, dev avg loss 0.229503, throughput 9.44153K wps
[Epoch 94 Batch 30/162] avg loss 0.0017501, throughput 9.54151K wps
[Epoch 94 Batch 60/162] avg loss 0.00189529, throughput 9.4493K wps
[Epoch 94 Batch 90/162] avg loss 0.00171395, throughput 9.34279K wps
[Epoch 94 Batch 120/162] avg loss 0.00168188, throughput 9.3006K wps
[Epoch 94 Batch 150/162] avg loss 0.00176516, throughput 9.26229K wps
Begin Testing...
[Epoch 94] train avg loss 0.00177128, dev acc 0.9089, dev avg loss 0.229222, throughput 9.3895K wps
[Epoch 95 Batch 30/162] avg loss 0.00192507, throughput 9.48517K wps
[Epoch 95 Batch 60/162] avg loss 0.00171763, throughput 9.40488K wps
[Epoch 95 Batch 90/162] avg loss 0.00158727, throughput 9.49213K wps
[Epoch 95 Batch 120/162] avg loss 0.00190603, throughput 9.37772K wps
[Epoch 95 Batch 150/162] avg loss 0.00168699, throughput 9.34463K wps
Begin Testing...
[Epoch 95] train avg loss 0.00175758, dev acc 0.9056, dev avg loss 0.229732, throughput 9.41329K wps
[Epoch 96 Batch 30/162] avg loss 0.00164219, throughput 9.64598K wps
[Epoch 96 Batch 60/162] avg loss 0.00180407, throughput 9.26244K wps
[Epoch 96 Batch 90/162] avg loss 0.00168376, throughput 9.47289K wps
[Epoch 96 Batch 120/162] avg loss 0.00174691, throughput 9.56701K wps
[Epoch 96 Batch 150/162] avg loss 0.00146591, throughput 9.55153K wps
Begin Testing...
[Epoch 96] train avg loss 0.00166984, dev acc 0.9056, dev avg loss 0.230337, throughput 9.4896K wps
[Epoch 97 Batch 30/162] avg loss 0.00164687, throughput 9.45882K wps
[Epoch 97 Batch 60/162] avg loss 0.00149098, throughput 9.34822K wps
[Epoch 97 Batch 90/162] avg loss 0.00176893, throughput 9.42753K wps
[Epoch 97 Batch 120/162] avg loss 0.00181979, throughput 9.41794K wps
[Epoch 97 Batch 150/162] avg loss 0.00175131, throughput 9.41042K wps
Begin Testing...
[Epoch 97] train avg loss 0.00171622, dev acc 0.9056, dev avg loss 0.22993, throughput 9.42032K wps
[Epoch 98 Batch 30/162] avg loss 0.00152995, throughput 9.69674K wps
[Epoch 98 Batch 60/162] avg loss 0.00145983, throughput 9.30966K wps
[Epoch 98 Batch 90/162] avg loss 0.0015705, throughput 9.51451K wps
[Epoch 98 Batch 120/162] avg loss 0.00182273, throughput 9.4183K wps
[Epoch 98 Batch 150/162] avg loss 0.00193128, throughput 9.33212K wps
Begin Testing...
[Epoch 98] train avg loss 0.00168408, dev acc 0.9067, dev avg loss 0.230773, throughput 9.46114K wps
[Epoch 99 Batch 30/162] avg loss 0.00166608, throughput 9.61557K wps
[Epoch 99 Batch 60/162] avg loss 0.00174995, throughput 9.18671K wps
[Epoch 99 Batch 90/162] avg loss 0.0017058, throughput 9.48183K wps
[Epoch 99 Batch 120/162] avg loss 0.00178574, throughput 9.26609K wps
[Epoch 99 Batch 150/162] avg loss 0.00156201, throughput 9.28227K wps
Begin Testing...
[Epoch 99] train avg loss 0.0016845, dev acc 0.9044, dev avg loss 0.229811, throughput 9.35877K wps
[Epoch 100 Batch 30/162] avg loss 0.00161489, throughput 9.53247K wps
[Epoch 100 Batch 60/162] avg loss 0.00151833, throughput 9.30704K wps
[Epoch 100 Batch 90/162] avg loss 0.00163977, throughput 9.63101K wps
[Epoch 100 Batch 120/162] avg loss 0.00157494, throughput 9.45964K wps
[Epoch 100 Batch 150/162] avg loss 0.00160811, throughput 9.34586K wps
Begin Testing...
[Epoch 100] train avg loss 0.00158965, dev acc 0.9067, dev avg loss 0.229617, throughput 9.4506K wps
[Epoch 101 Batch 30/162] avg loss 0.00155059, throughput 9.58724K wps
[Epoch 101 Batch 60/162] avg loss 0.00159646, throughput 9.43018K wps
[Epoch 101 Batch 90/162] avg loss 0.00153429, throughput 9.39148K wps
[Epoch 101 Batch 120/162] avg loss 0.0017968, throughput 9.41788K wps
[Epoch 101 Batch 150/162] avg loss 0.00141439, throughput 9.53901K wps
Begin Testing...
[Epoch 101] train avg loss 0.00158273, dev acc 0.9056, dev avg loss 0.229857, throughput 9.48433K wps
[Epoch 102 Batch 30/162] avg loss 0.00145153, throughput 9.51218K wps
[Epoch 102 Batch 60/162] avg loss 0.00146349, throughput 9.39322K wps
[Epoch 102 Batch 90/162] avg loss 0.00155615, throughput 9.55787K wps
[Epoch 102 Batch 120/162] avg loss 0.00171872, throughput 9.48334K wps
[Epoch 102 Batch 150/162] avg loss 0.00179406, throughput 9.36459K wps
Begin Testing...
[Epoch 102] train avg loss 0.00158761, dev acc 0.9056, dev avg loss 0.229883, throughput 9.4716K wps
[Epoch 103 Batch 30/162] avg loss 0.0016011, throughput 9.51903K wps
[Epoch 103 Batch 60/162] avg loss 0.00142531, throughput 9.58996K wps
[Epoch 103 Batch 90/162] avg loss 0.00175576, throughput 9.39976K wps
[Epoch 103 Batch 120/162] avg loss 0.00163905, throughput 9.37047K wps
[Epoch 103 Batch 150/162] avg loss 0.00150459, throughput 9.43596K wps
Begin Testing...
[Epoch 103] train avg loss 0.00157957, dev acc 0.9089, dev avg loss 0.229327, throughput 9.47277K wps
[Epoch 104 Batch 30/162] avg loss 0.00153513, throughput 9.33868K wps
[Epoch 104 Batch 60/162] avg loss 0.00149964, throughput 9.5506K wps
[Epoch 104 Batch 90/162] avg loss 0.00160712, throughput 9.2441K wps
[Epoch 104 Batch 120/162] avg loss 0.00144635, throughput 9.53K wps
[Epoch 104 Batch 150/162] avg loss 0.00153667, throughput 9.47824K wps
Begin Testing...
[Epoch 104] train avg loss 0.00155162, dev acc 0.9078, dev avg loss 0.229468, throughput 9.40776K wps
[Epoch 105 Batch 30/162] avg loss 0.00141183, throughput 9.51289K wps
[Epoch 105 Batch 60/162] avg loss 0.00151852, throughput 9.49043K wps
[Epoch 105 Batch 90/162] avg loss 0.00152356, throughput 9.39677K wps
[Epoch 105 Batch 120/162] avg loss 0.0015766, throughput 9.27471K wps
[Epoch 105 Batch 150/162] avg loss 0.00143873, throughput 9.32436K wps
Begin Testing...
[Epoch 105] train avg loss 0.00149833, dev acc 0.9067, dev avg loss 0.229239, throughput 9.40824K wps
[Epoch 106 Batch 30/162] avg loss 0.00154192, throughput 9.75021K wps
[Epoch 106 Batch 60/162] avg loss 0.0015088, throughput 9.44832K wps
[Epoch 106 Batch 90/162] avg loss 0.00148075, throughput 9.30569K wps
[Epoch 106 Batch 120/162] avg loss 0.00147681, throughput 9.3575K wps
[Epoch 106 Batch 150/162] avg loss 0.00143831, throughput 9.3803K wps
Begin Testing...
[Epoch 106] train avg loss 0.00147954, dev acc 0.9078, dev avg loss 0.229283, throughput 9.42957K wps
[Epoch 107 Batch 30/162] avg loss 0.00161458, throughput 9.68302K wps
[Epoch 107 Batch 60/162] avg loss 0.00136929, throughput 9.38921K wps
[Epoch 107 Batch 90/162] avg loss 0.00169635, throughput 9.27688K wps
[Epoch 107 Batch 120/162] avg loss 0.00150827, throughput 9.4801K wps
[Epoch 107 Batch 150/162] avg loss 0.00157872, throughput 9.39013K wps
Begin Testing...
[Epoch 107] train avg loss 0.00155469, dev acc 0.9078, dev avg loss 0.229541, throughput 9.42395K wps
[Epoch 108 Batch 30/162] avg loss 0.00155633, throughput 9.56858K wps
[Epoch 108 Batch 60/162] avg loss 0.00148016, throughput 9.56082K wps
[Epoch 108 Batch 90/162] avg loss 0.0015239, throughput 9.29913K wps
[Epoch 108 Batch 120/162] avg loss 0.00149799, throughput 9.50152K wps
[Epoch 108 Batch 150/162] avg loss 0.00138366, throughput 9.34164K wps
Begin Testing...
[Epoch 108] train avg loss 0.00148478, dev acc 0.9067, dev avg loss 0.23007, throughput 9.44638K wps
[Epoch 109 Batch 30/162] avg loss 0.00148439, throughput 9.60358K wps
[Epoch 109 Batch 60/162] avg loss 0.00142033, throughput 9.33026K wps
[Epoch 109 Batch 90/162] avg loss 0.00131167, throughput 9.40558K wps
[Epoch 109 Batch 120/162] avg loss 0.00148522, throughput 9.38532K wps
[Epoch 109 Batch 150/162] avg loss 0.0016369, throughput 9.36925K wps
Begin Testing...
[Epoch 109] train avg loss 0.00146949, dev acc 0.9044, dev avg loss 0.230278, throughput 9.41857K wps
[Epoch 110 Batch 30/162] avg loss 0.00143108, throughput 9.57213K wps
[Epoch 110 Batch 60/162] avg loss 0.00141821, throughput 9.59217K wps
[Epoch 110 Batch 90/162] avg loss 0.00134611, throughput 9.18034K wps
[Epoch 110 Batch 120/162] avg loss 0.00144927, throughput 9.48332K wps
[Epoch 110 Batch 150/162] avg loss 0.00164894, throughput 9.54565K wps
Begin Testing...
[Epoch 110] train avg loss 0.00145218, dev acc 0.9044, dev avg loss 0.229918, throughput 9.46688K wps
[Epoch 111 Batch 30/162] avg loss 0.00155226, throughput 9.71186K wps
[Epoch 111 Batch 60/162] avg loss 0.00130029, throughput 9.2966K wps
[Epoch 111 Batch 90/162] avg loss 0.00141153, throughput 9.27986K wps
[Epoch 111 Batch 120/162] avg loss 0.00144165, throughput 9.43018K wps
[Epoch 111 Batch 150/162] avg loss 0.00145312, throughput 9.34458K wps
Begin Testing...
[Epoch 111] train avg loss 0.00143556, dev acc 0.9067, dev avg loss 0.230067, throughput 9.42379K wps
[Epoch 112 Batch 30/162] avg loss 0.00139717, throughput 9.45331K wps
[Epoch 112 Batch 60/162] avg loss 0.00138277, throughput 9.41432K wps
[Epoch 112 Batch 90/162] avg loss 0.00147046, throughput 9.40112K wps
[Epoch 112 Batch 120/162] avg loss 0.0013802, throughput 9.26478K wps
[Epoch 112 Batch 150/162] avg loss 0.00127427, throughput 9.24456K wps
Begin Testing...
[Epoch 112] train avg loss 0.00136954, dev acc 0.9067, dev avg loss 0.229929, throughput 9.34168K wps
[Epoch 113 Batch 30/162] avg loss 0.00143926, throughput 9.31985K wps
[Epoch 113 Batch 60/162] avg loss 0.00141941, throughput 9.49312K wps
[Epoch 113 Batch 90/162] avg loss 0.00138592, throughput 9.25285K wps
[Epoch 113 Batch 120/162] avg loss 0.0014533, throughput 9.49174K wps
[Epoch 113 Batch 150/162] avg loss 0.00131121, throughput 9.22303K wps
Begin Testing...
[Epoch 113] train avg loss 0.00139108, dev acc 0.9056, dev avg loss 0.230654, throughput 9.36915K wps
[Epoch 114 Batch 30/162] avg loss 0.00132429, throughput 9.55196K wps
[Epoch 114 Batch 60/162] avg loss 0.00138075, throughput 9.25876K wps
[Epoch 114 Batch 90/162] avg loss 0.00130877, throughput 9.51793K wps
[Epoch 114 Batch 120/162] avg loss 0.00149331, throughput 9.36408K wps
[Epoch 114 Batch 150/162] avg loss 0.00127485, throughput 9.24873K wps
Begin Testing...
[Epoch 114] train avg loss 0.00134654, dev acc 0.9089, dev avg loss 0.229764, throughput 9.39565K wps
[Epoch 115 Batch 30/162] avg loss 0.00143874, throughput 9.50552K wps
[Epoch 115 Batch 60/162] avg loss 0.0012415, throughput 9.275K wps
[Epoch 115 Batch 90/162] avg loss 0.00140917, throughput 9.57682K wps
[Epoch 115 Batch 120/162] avg loss 0.00133977, throughput 9.35983K wps
[Epoch 115 Batch 150/162] avg loss 0.00131923, throughput 9.24978K wps
Begin Testing...
[Epoch 115] train avg loss 0.0013441, dev acc 0.9078, dev avg loss 0.229782, throughput 9.3838K wps
[Epoch 116 Batch 30/162] avg loss 0.00130383, throughput 9.59202K wps
[Epoch 116 Batch 60/162] avg loss 0.00136091, throughput 9.53288K wps
[Epoch 116 Batch 90/162] avg loss 0.00137945, throughput 9.48212K wps
[Epoch 116 Batch 120/162] avg loss 0.00136721, throughput 9.42407K wps
[Epoch 116 Batch 150/162] avg loss 0.0013071, throughput 9.43917K wps
Begin Testing...
[Epoch 116] train avg loss 0.00134737, dev acc 0.9089, dev avg loss 0.229896, throughput 9.50402K wps
[Epoch 117 Batch 30/162] avg loss 0.00134371, throughput 9.66092K wps
[Epoch 117 Batch 60/162] avg loss 0.00131858, throughput 9.41767K wps
[Epoch 117 Batch 90/162] avg loss 0.001278, throughput 9.58842K wps
[Epoch 117 Batch 120/162] avg loss 0.00138105, throughput 9.36733K wps
[Epoch 117 Batch 150/162] avg loss 0.00132221, throughput 9.27931K wps
Begin Testing...
[Epoch 117] train avg loss 0.00132176, dev acc 0.9089, dev avg loss 0.230519, throughput 9.46554K wps
[Epoch 118 Batch 30/162] avg loss 0.00149979, throughput 9.60908K wps
[Epoch 118 Batch 60/162] avg loss 0.00116583, throughput 9.47154K wps
[Epoch 118 Batch 90/162] avg loss 0.00124398, throughput 9.3079K wps
[Epoch 118 Batch 120/162] avg loss 0.00132941, throughput 9.37475K wps
[Epoch 118 Batch 150/162] avg loss 0.00113277, throughput 9.42613K wps
Begin Testing...
[Epoch 118] train avg loss 0.00127486, dev acc 0.9089, dev avg loss 0.230407, throughput 9.43966K wps
[Epoch 119 Batch 30/162] avg loss 0.00133807, throughput 9.57203K wps
[Epoch 119 Batch 60/162] avg loss 0.00126535, throughput 9.41099K wps
[Epoch 119 Batch 90/162] avg loss 0.00132665, throughput 9.52672K wps
[Epoch 119 Batch 120/162] avg loss 0.00127105, throughput 9.32167K wps
[Epoch 119 Batch 150/162] avg loss 0.00130381, throughput 9.5156K wps
Begin Testing...
[Epoch 119] train avg loss 0.00129623, dev acc 0.9078, dev avg loss 0.230656, throughput 9.45224K wps
[Epoch 120 Batch 30/162] avg loss 0.00119283, throughput 9.74089K wps
[Epoch 120 Batch 60/162] avg loss 0.00130535, throughput 9.2668K wps
[Epoch 120 Batch 90/162] avg loss 0.00129203, throughput 9.41908K wps
[Epoch 120 Batch 120/162] avg loss 0.00122921, throughput 9.2502K wps
[Epoch 120 Batch 150/162] avg loss 0.00121516, throughput 9.36331K wps
Begin Testing...
[Epoch 120] train avg loss 0.00124534, dev acc 0.9067, dev avg loss 0.232324, throughput 9.39858K wps
[Epoch 121 Batch 30/162] avg loss 0.00130413, throughput 9.51718K wps
[Epoch 121 Batch 60/162] avg loss 0.00123464, throughput 9.31816K wps
[Epoch 121 Batch 90/162] avg loss 0.00131969, throughput 9.32898K wps
[Epoch 121 Batch 120/162] avg loss 0.00112522, throughput 9.52111K wps
[Epoch 121 Batch 150/162] avg loss 0.00135551, throughput 9.27221K wps
Begin Testing...
[Epoch 121] train avg loss 0.00126869, dev acc 0.9089, dev avg loss 0.232824, throughput 9.38386K wps
[Epoch 122 Batch 30/162] avg loss 0.00121979, throughput 9.61797K wps
[Epoch 122 Batch 60/162] avg loss 0.00128912, throughput 9.48748K wps
[Epoch 122 Batch 90/162] avg loss 0.00114383, throughput 9.51794K wps
[Epoch 122 Batch 120/162] avg loss 0.00123551, throughput 9.40637K wps
[Epoch 122 Batch 150/162] avg loss 0.0013794, throughput 9.27151K wps
Begin Testing...
[Epoch 122] train avg loss 0.00125345, dev acc 0.9089, dev avg loss 0.231178, throughput 9.47191K wps
[Epoch 123 Batch 30/162] avg loss 0.00123153, throughput 9.44986K wps
[Epoch 123 Batch 60/162] avg loss 0.00117243, throughput 9.37411K wps
[Epoch 123 Batch 90/162] avg loss 0.00115469, throughput 9.23474K wps
[Epoch 123 Batch 120/162] avg loss 0.00120026, throughput 9.60462K wps
[Epoch 123 Batch 150/162] avg loss 0.00130324, throughput 9.31313K wps
Begin Testing...
[Epoch 123] train avg loss 0.00120856, dev acc 0.9100, dev avg loss 0.230575, throughput 9.38487K wps
Observed Improvement.
Begin Testing...
[Epoch 124 Batch 30/162] avg loss 0.00131587, throughput 9.45153K wps
[Epoch 124 Batch 60/162] avg loss 0.00107548, throughput 9.32783K wps
[Epoch 124 Batch 90/162] avg loss 0.00116448, throughput 9.38583K wps
[Epoch 124 Batch 120/162] avg loss 0.00123663, throughput 9.40002K wps
[Epoch 124 Batch 150/162] avg loss 0.00116129, throughput 9.54536K wps
Begin Testing...
[Epoch 124] train avg loss 0.00119748, dev acc 0.9078, dev avg loss 0.231368, throughput 9.4279K wps
[Epoch 125 Batch 30/162] avg loss 0.00122607, throughput 9.60759K wps
[Epoch 125 Batch 60/162] avg loss 0.000989195, throughput 9.34235K wps
[Epoch 125 Batch 90/162] avg loss 0.00120134, throughput 9.42609K wps
[Epoch 125 Batch 120/162] avg loss 0.00116429, throughput 9.30946K wps
[Epoch 125 Batch 150/162] avg loss 0.00120988, throughput 9.45256K wps
Begin Testing...
[Epoch 125] train avg loss 0.00117088, dev acc 0.9067, dev avg loss 0.232322, throughput 9.43699K wps
[Epoch 126 Batch 30/162] avg loss 0.00125266, throughput 9.48973K wps
[Epoch 126 Batch 60/162] avg loss 0.00116606, throughput 9.36464K wps
[Epoch 126 Batch 90/162] avg loss 0.00106504, throughput 9.47364K wps
[Epoch 126 Batch 120/162] avg loss 0.00133055, throughput 9.3321K wps
[Epoch 126 Batch 150/162] avg loss 0.00117007, throughput 9.45828K wps
Begin Testing...
[Epoch 126] train avg loss 0.00119145, dev acc 0.9089, dev avg loss 0.230981, throughput 9.41892K wps
[Epoch 127 Batch 30/162] avg loss 0.00125357, throughput 9.55533K wps
[Epoch 127 Batch 60/162] avg loss 0.00118331, throughput 9.53164K wps
[Epoch 127 Batch 90/162] avg loss 0.00122439, throughput 9.4177K wps
[Epoch 127 Batch 120/162] avg loss 0.0011428, throughput 9.35033K wps
[Epoch 127 Batch 150/162] avg loss 0.00119592, throughput 9.29073K wps
Begin Testing...
[Epoch 127] train avg loss 0.00118048, dev acc 0.9089, dev avg loss 0.233368, throughput 9.43154K wps
[Epoch 128 Batch 30/162] avg loss 0.00127954, throughput 9.52522K wps
[Epoch 128 Batch 60/162] avg loss 0.00116916, throughput 9.47814K wps
[Epoch 128 Batch 90/162] avg loss 0.00116029, throughput 9.42214K wps
[Epoch 128 Batch 120/162] avg loss 0.00101199, throughput 9.50624K wps
[Epoch 128 Batch 150/162] avg loss 0.00120605, throughput 9.48094K wps
Begin Testing...
[Epoch 128] train avg loss 0.00116465, dev acc 0.9089, dev avg loss 0.231094, throughput 9.48963K wps
[Epoch 129 Batch 30/162] avg loss 0.0010908, throughput 9.40183K wps
[Epoch 129 Batch 60/162] avg loss 0.00121261, throughput 9.32102K wps
[Epoch 129 Batch 90/162] avg loss 0.00120329, throughput 9.32543K wps
[Epoch 129 Batch 120/162] avg loss 0.00103374, throughput 9.40656K wps
[Epoch 129 Batch 150/162] avg loss 0.00116011, throughput 9.30561K wps
Begin Testing...
[Epoch 129] train avg loss 0.00112974, dev acc 0.9078, dev avg loss 0.231639, throughput 9.36664K wps
[Epoch 130 Batch 30/162] avg loss 0.00127606, throughput 9.51606K wps
[Epoch 130 Batch 60/162] avg loss 0.00115915, throughput 9.39679K wps
[Epoch 130 Batch 90/162] avg loss 0.00118054, throughput 9.40936K wps
[Epoch 130 Batch 120/162] avg loss 0.00105857, throughput 9.5402K wps
[Epoch 130 Batch 150/162] avg loss 0.00117376, throughput 9.30031K wps
Begin Testing...
[Epoch 130] train avg loss 0.00116596, dev acc 0.9089, dev avg loss 0.231688, throughput 9.43072K wps
[Epoch 131 Batch 30/162] avg loss 0.0011466, throughput 9.78201K wps
[Epoch 131 Batch 60/162] avg loss 0.00112706, throughput 9.31586K wps
[Epoch 131 Batch 90/162] avg loss 0.00104819, throughput 9.3102K wps
[Epoch 131 Batch 120/162] avg loss 0.00114394, throughput 9.37388K wps
[Epoch 131 Batch 150/162] avg loss 0.00117354, throughput 9.33197K wps
Begin Testing...
[Epoch 131] train avg loss 0.00111917, dev acc 0.9078, dev avg loss 0.231436, throughput 9.42101K wps
[Epoch 132 Batch 30/162] avg loss 0.00113781, throughput 9.52037K wps
[Epoch 132 Batch 60/162] avg loss 0.00101034, throughput 9.2867K wps
[Epoch 132 Batch 90/162] avg loss 0.00120764, throughput 9.32912K wps
[Epoch 132 Batch 120/162] avg loss 0.00107254, throughput 9.3168K wps
[Epoch 132 Batch 150/162] avg loss 0.00117417, throughput 9.35506K wps
Begin Testing...
[Epoch 132] train avg loss 0.00111857, dev acc 0.9089, dev avg loss 0.23163, throughput 9.35942K wps
[Epoch 133 Batch 30/162] avg loss 0.00116766, throughput 9.51698K wps
[Epoch 133 Batch 60/162] avg loss 0.00117161, throughput 9.35977K wps
[Epoch 133 Batch 90/162] avg loss 0.00105401, throughput 9.31941K wps
[Epoch 133 Batch 120/162] avg loss 0.00102707, throughput 9.23051K wps
[Epoch 133 Batch 150/162] avg loss 0.00107245, throughput 9.30163K wps
Begin Testing...
[Epoch 133] train avg loss 0.00108673, dev acc 0.9089, dev avg loss 0.232259, throughput 9.36327K wps
[Epoch 134 Batch 30/162] avg loss 0.00110652, throughput 9.5603K wps
[Epoch 134 Batch 60/162] avg loss 0.00104329, throughput 9.52606K wps
[Epoch 134 Batch 90/162] avg loss 0.00115807, throughput 9.19236K wps
[Epoch 134 Batch 120/162] avg loss 0.00112703, throughput 9.39201K wps
[Epoch 134 Batch 150/162] avg loss 0.000917882, throughput 9.33058K wps
Begin Testing...
[Epoch 134] train avg loss 0.00106053, dev acc 0.9089, dev avg loss 0.232083, throughput 9.40031K wps
[Epoch 135 Batch 30/162] avg loss 0.00120027, throughput 9.6566K wps
[Epoch 135 Batch 60/162] avg loss 0.000972883, throughput 9.36635K wps
[Epoch 135 Batch 90/162] avg loss 0.0010226, throughput 9.50579K wps
[Epoch 135 Batch 120/162] avg loss 0.00107877, throughput 9.57004K wps
[Epoch 135 Batch 150/162] avg loss 0.00104918, throughput 9.43732K wps
Begin Testing...
[Epoch 135] train avg loss 0.00106597, dev acc 0.9100, dev avg loss 0.233435, throughput 9.4907K wps
Observed Improvement.
Begin Testing...
[Epoch 136 Batch 30/162] avg loss 0.00104226, throughput 9.49632K wps
[Epoch 136 Batch 60/162] avg loss 0.0011248, throughput 9.42263K wps
[Epoch 136 Batch 90/162] avg loss 0.000979714, throughput 9.39932K wps
[Epoch 136 Batch 120/162] avg loss 0.00112235, throughput 9.53759K wps
[Epoch 136 Batch 150/162] avg loss 0.00106865, throughput 9.28172K wps
Begin Testing...
[Epoch 136] train avg loss 0.00107747, dev acc 0.9089, dev avg loss 0.232289, throughput 9.41755K wps
[Epoch 137 Batch 30/162] avg loss 0.00107265, throughput 9.6354K wps
[Epoch 137 Batch 60/162] avg loss 0.0010436, throughput 9.3048K wps
[Epoch 137 Batch 90/162] avg loss 0.00106676, throughput 9.47605K wps
[Epoch 137 Batch 120/162] avg loss 0.00101127, throughput 9.33882K wps
[Epoch 137 Batch 150/162] avg loss 0.0010889, throughput 9.41961K wps
Begin Testing...
[Epoch 137] train avg loss 0.00104105, dev acc 0.9089, dev avg loss 0.232055, throughput 9.42202K wps
[Epoch 138 Batch 30/162] avg loss 0.00101001, throughput 9.52996K wps
[Epoch 138 Batch 60/162] avg loss 0.00107274, throughput 9.40712K wps
[Epoch 138 Batch 90/162] avg loss 0.00112835, throughput 9.54277K wps
[Epoch 138 Batch 120/162] avg loss 0.00109074, throughput 9.36165K wps
[Epoch 138 Batch 150/162] avg loss 0.000926584, throughput 9.47375K wps
Begin Testing...
[Epoch 138] train avg loss 0.00102443, dev acc 0.9100, dev avg loss 0.234792, throughput 9.45976K wps
Observed Improvement.
Begin Testing...
[Epoch 139 Batch 30/162] avg loss 0.000917554, throughput 9.50602K wps
[Epoch 139 Batch 60/162] avg loss 0.00102483, throughput 9.56401K wps
[Epoch 139 Batch 90/162] avg loss 0.00104301, throughput 9.30412K wps
[Epoch 139 Batch 120/162] avg loss 0.000902007, throughput 9.42485K wps
[Epoch 139 Batch 150/162] avg loss 0.000978564, throughput 9.34043K wps
Begin Testing...
[Epoch 139] train avg loss 0.000973383, dev acc 0.9089, dev avg loss 0.23379, throughput 9.41532K wps
[Epoch 140 Batch 30/162] avg loss 0.00112363, throughput 9.58172K wps
[Epoch 140 Batch 60/162] avg loss 0.000927711, throughput 9.35552K wps
[Epoch 140 Batch 90/162] avg loss 0.00105667, throughput 9.27178K wps
[Epoch 140 Batch 120/162] avg loss 0.00109079, throughput 9.33046K wps
[Epoch 140 Batch 150/162] avg loss 0.000950986, throughput 9.45505K wps
Begin Testing...
[Epoch 140] train avg loss 0.00103573, dev acc 0.9089, dev avg loss 0.232376, throughput 9.38938K wps
[Epoch 141 Batch 30/162] avg loss 0.0010825, throughput 9.57874K wps
[Epoch 141 Batch 60/162] avg loss 0.000925724, throughput 9.40497K wps
[Epoch 141 Batch 90/162] avg loss 0.000907817, throughput 9.54674K wps
[Epoch 141 Batch 120/162] avg loss 0.00105981, throughput 9.22967K wps
[Epoch 141 Batch 150/162] avg loss 0.00104694, throughput 9.48152K wps
Begin Testing...
[Epoch 141] train avg loss 0.000989632, dev acc 0.9100, dev avg loss 0.232313, throughput 9.45458K wps
Observed Improvement.
Begin Testing...
[Epoch 142 Batch 30/162] avg loss 0.000791197, throughput 9.63629K wps
[Epoch 142 Batch 60/162] avg loss 0.00111298, throughput 9.30087K wps
[Epoch 142 Batch 90/162] avg loss 0.00109017, throughput 9.34436K wps
[Epoch 142 Batch 120/162] avg loss 0.000968016, throughput 9.51697K wps
[Epoch 142 Batch 150/162] avg loss 0.00100595, throughput 9.40806K wps
Begin Testing...
[Epoch 142] train avg loss 0.000993251, dev acc 0.9089, dev avg loss 0.232322, throughput 9.45164K wps
[Epoch 143 Batch 30/162] avg loss 0.000981228, throughput 9.60639K wps
[Epoch 143 Batch 60/162] avg loss 0.00095341, throughput 9.27932K wps
[Epoch 143 Batch 90/162] avg loss 0.000919176, throughput 9.42186K wps
[Epoch 143 Batch 120/162] avg loss 0.00104658, throughput 9.38615K wps
[Epoch 143 Batch 150/162] avg loss 0.000867069, throughput 9.33962K wps
Begin Testing...
[Epoch 143] train avg loss 0.000950099, dev acc 0.9078, dev avg loss 0.232937, throughput 9.42267K wps
[Epoch 144 Batch 30/162] avg loss 0.000974376, throughput 9.6411K wps
[Epoch 144 Batch 60/162] avg loss 0.00100757, throughput 9.38647K wps
[Epoch 144 Batch 90/162] avg loss 0.000908716, throughput 9.53237K wps
[Epoch 144 Batch 120/162] avg loss 0.000833444, throughput 9.28649K wps
[Epoch 144 Batch 150/162] avg loss 0.000887538, throughput 9.41759K wps
Begin Testing...
[Epoch 144] train avg loss 0.000939386, dev acc 0.9089, dev avg loss 0.23314, throughput 9.44574K wps
[Epoch 145 Batch 30/162] avg loss 0.00106849, throughput 9.56686K wps
[Epoch 145 Batch 60/162] avg loss 0.00101651, throughput 9.45993K wps
[Epoch 145 Batch 90/162] avg loss 0.000900312, throughput 9.31088K wps
[Epoch 145 Batch 120/162] avg loss 0.00108238, throughput 9.24731K wps
[Epoch 145 Batch 150/162] avg loss 0.000922512, throughput 9.37205K wps
Begin Testing...
[Epoch 145] train avg loss 0.000991222, dev acc 0.9089, dev avg loss 0.233307, throughput 9.37975K wps
[Epoch 146 Batch 30/162] avg loss 0.000977123, throughput 9.64177K wps
[Epoch 146 Batch 60/162] avg loss 0.000843502, throughput 9.33919K wps
[Epoch 146 Batch 90/162] avg loss 0.000988307, throughput 9.34273K wps
[Epoch 146 Batch 120/162] avg loss 0.000975136, throughput 9.5271K wps
[Epoch 146 Batch 150/162] avg loss 0.0010214, throughput 9.32937K wps
Begin Testing...
[Epoch 146] train avg loss 0.000946645, dev acc 0.9089, dev avg loss 0.23337, throughput 9.42273K wps
[Epoch 147 Batch 30/162] avg loss 0.000967704, throughput 9.7456K wps
[Epoch 147 Batch 60/162] avg loss 0.000919638, throughput 9.37475K wps
[Epoch 147 Batch 90/162] avg loss 0.00104986, throughput 9.57752K wps
[Epoch 147 Batch 120/162] avg loss 0.00083452, throughput 9.30869K wps
[Epoch 147 Batch 150/162] avg loss 0.000892878, throughput 9.39343K wps
Begin Testing...
[Epoch 147] train avg loss 0.000933025, dev acc 0.9089, dev avg loss 0.234006, throughput 9.45983K wps
[Epoch 148 Batch 30/162] avg loss 0.000899146, throughput 9.68824K wps
[Epoch 148 Batch 60/162] avg loss 0.000853725, throughput 9.4349K wps
[Epoch 148 Batch 90/162] avg loss 0.000912254, throughput 9.31751K wps
[Epoch 148 Batch 120/162] avg loss 0.000917275, throughput 9.34575K wps
[Epoch 148 Batch 150/162] avg loss 0.00105725, throughput 9.50296K wps
Begin Testing...
[Epoch 148] train avg loss 0.000918427, dev acc 0.9078, dev avg loss 0.233348, throughput 9.4517K wps
[Epoch 149 Batch 30/162] avg loss 0.000901917, throughput 9.57802K wps
[Epoch 149 Batch 60/162] avg loss 0.0010974, throughput 9.31524K wps
[Epoch 149 Batch 90/162] avg loss 0.000937287, throughput 9.2491K wps
[Epoch 149 Batch 120/162] avg loss 0.000894007, throughput 9.41471K wps
[Epoch 149 Batch 150/162] avg loss 0.000729819, throughput 9.5361K wps
Begin Testing...
[Epoch 149] train avg loss 0.000923802, dev acc 0.9111, dev avg loss 0.234489, throughput 9.43384K wps
Observed Improvement.
Begin Testing...
[Epoch 150 Batch 30/162] avg loss 0.000804541, throughput 9.64593K wps
[Epoch 150 Batch 60/162] avg loss 0.000940878, throughput 9.30358K wps
[Epoch 150 Batch 90/162] avg loss 0.00092122, throughput 9.45642K wps
[Epoch 150 Batch 120/162] avg loss 0.000889109, throughput 9.56398K wps
[Epoch 150 Batch 150/162] avg loss 0.000911743, throughput 9.31791K wps
Begin Testing...
[Epoch 150] train avg loss 0.000909455, dev acc 0.9133, dev avg loss 0.237016, throughput 9.45791K wps
Observed Improvement.
Begin Testing...
[Epoch 151 Batch 30/162] avg loss 0.00104687, throughput 9.62529K wps
[Epoch 151 Batch 60/162] avg loss 0.000783363, throughput 9.30955K wps
[Epoch 151 Batch 90/162] avg loss 0.000865708, throughput 9.47896K wps
[Epoch 151 Batch 120/162] avg loss 0.000898965, throughput 9.18108K wps
[Epoch 151 Batch 150/162] avg loss 0.000912511, throughput 9.32852K wps
Begin Testing...
[Epoch 151] train avg loss 0.000901924, dev acc 0.9089, dev avg loss 0.234193, throughput 9.39453K wps
[Epoch 152 Batch 30/162] avg loss 0.000942033, throughput 9.66154K wps
[Epoch 152 Batch 60/162] avg loss 0.000917, throughput 9.32729K wps
[Epoch 152 Batch 90/162] avg loss 0.000853672, throughput 9.46519K wps
[Epoch 152 Batch 120/162] avg loss 0.000865698, throughput 9.61104K wps
[Epoch 152 Batch 150/162] avg loss 0.000930457, throughput 9.46066K wps
Begin Testing...
[Epoch 152] train avg loss 0.000898754, dev acc 0.9089, dev avg loss 0.233933, throughput 9.50791K wps
[Epoch 153 Batch 30/162] avg loss 0.000848278, throughput 9.49512K wps
[Epoch 153 Batch 60/162] avg loss 0.000874891, throughput 9.34026K wps
[Epoch 153 Batch 90/162] avg loss 0.000924783, throughput 9.47398K wps
[Epoch 153 Batch 120/162] avg loss 0.0008464, throughput 9.41661K wps
[Epoch 153 Batch 150/162] avg loss 0.000908126, throughput 9.47812K wps
Begin Testing...
[Epoch 153] train avg loss 0.000874369, dev acc 0.9100, dev avg loss 0.234371, throughput 9.44979K wps
[Epoch 154 Batch 30/162] avg loss 0.000920672, throughput 9.74402K wps
[Epoch 154 Batch 60/162] avg loss 0.000814592, throughput 9.55446K wps
[Epoch 154 Batch 90/162] avg loss 0.000811327, throughput 9.28855K wps
[Epoch 154 Batch 120/162] avg loss 0.000805069, throughput 9.51916K wps
[Epoch 154 Batch 150/162] avg loss 0.000926132, throughput 9.39829K wps
Begin Testing...
[Epoch 154] train avg loss 0.000854013, dev acc 0.9089, dev avg loss 0.234696, throughput 9.50781K wps
[Epoch 155 Batch 30/162] avg loss 0.000819967, throughput 9.41411K wps
[Epoch 155 Batch 60/162] avg loss 0.000851742, throughput 9.42135K wps
[Epoch 155 Batch 90/162] avg loss 0.00085012, throughput 9.38537K wps
[Epoch 155 Batch 120/162] avg loss 0.000828359, throughput 9.33566K wps
[Epoch 155 Batch 150/162] avg loss 0.000979425, throughput 9.41623K wps
Begin Testing...
[Epoch 155] train avg loss 0.000861041, dev acc 0.9100, dev avg loss 0.234617, throughput 9.39749K wps
[Epoch 156 Batch 30/162] avg loss 0.000908386, throughput 9.62097K wps
[Epoch 156 Batch 60/162] avg loss 0.000821374, throughput 9.6243K wps
[Epoch 156 Batch 90/162] avg loss 0.000867861, throughput 9.4248K wps
[Epoch 156 Batch 120/162] avg loss 0.00078668, throughput 9.45719K wps
[Epoch 156 Batch 150/162] avg loss 0.00089233, throughput 9.41718K wps
Begin Testing...
[Epoch 156] train avg loss 0.00085236, dev acc 0.9089, dev avg loss 0.234838, throughput 9.51593K wps
[Epoch 157 Batch 30/162] avg loss 0.000736066, throughput 9.51409K wps
[Epoch 157 Batch 60/162] avg loss 0.000908002, throughput 9.40162K wps
[Epoch 157 Batch 90/162] avg loss 0.00083561, throughput 9.23029K wps
[Epoch 157 Batch 120/162] avg loss 0.000859348, throughput 9.47912K wps
[Epoch 157 Batch 150/162] avg loss 0.000805165, throughput 9.65545K wps
Begin Testing...
[Epoch 157] train avg loss 0.000831235, dev acc 0.9100, dev avg loss 0.235804, throughput 9.45897K wps
[Epoch 158 Batch 30/162] avg loss 0.000951143, throughput 9.41079K wps
[Epoch 158 Batch 60/162] avg loss 0.000778569, throughput 9.39475K wps
[Epoch 158 Batch 90/162] avg loss 0.000875491, throughput 9.5724K wps
[Epoch 158 Batch 120/162] avg loss 0.000952165, throughput 9.30331K wps
[Epoch 158 Batch 150/162] avg loss 0.000788414, throughput 9.56461K wps
Begin Testing...
[Epoch 158] train avg loss 0.000845678, dev acc 0.9078, dev avg loss 0.235733, throughput 9.4523K wps
[Epoch 159 Batch 30/162] avg loss 0.00085577, throughput 9.47013K wps
[Epoch 159 Batch 60/162] avg loss 0.000891118, throughput 9.27658K wps
[Epoch 159 Batch 90/162] avg loss 0.000861628, throughput 9.23381K wps
[Epoch 159 Batch 120/162] avg loss 0.000698481, throughput 9.2417K wps
[Epoch 159 Batch 150/162] avg loss 0.000864441, throughput 9.4254K wps
Begin Testing...
[Epoch 159] train avg loss 0.000825486, dev acc 0.9100, dev avg loss 0.234453, throughput 9.31329K wps
[Epoch 160 Batch 30/162] avg loss 0.000769882, throughput 9.48976K wps
[Epoch 160 Batch 60/162] avg loss 0.00079528, throughput 9.56516K wps
[Epoch 160 Batch 90/162] avg loss 0.000809148, throughput 9.2531K wps
[Epoch 160 Batch 120/162] avg loss 0.000830905, throughput 9.29415K wps
[Epoch 160 Batch 150/162] avg loss 0.00083244, throughput 9.44979K wps
Begin Testing...
[Epoch 160] train avg loss 0.000820002, dev acc 0.9111, dev avg loss 0.235537, throughput 9.39205K wps
[Epoch 161 Batch 30/162] avg loss 0.000844816, throughput 9.63445K wps
[Epoch 161 Batch 60/162] avg loss 0.000784664, throughput 9.44139K wps
[Epoch 161 Batch 90/162] avg loss 0.000801659, throughput 9.46971K wps
[Epoch 161 Batch 120/162] avg loss 0.000830579, throughput 9.46503K wps
[Epoch 161 Batch 150/162] avg loss 0.000821105, throughput 9.23749K wps
Begin Testing...
[Epoch 161] train avg loss 0.000821155, dev acc 0.9089, dev avg loss 0.235196, throughput 9.43005K wps
[Epoch 162 Batch 30/162] avg loss 0.000723523, throughput 9.50028K wps
[Epoch 162 Batch 60/162] avg loss 0.000831475, throughput 9.27321K wps
[Epoch 162 Batch 90/162] avg loss 0.000753669, throughput 9.24562K wps
[Epoch 162 Batch 120/162] avg loss 0.00094508, throughput 8.29039K wps
[Epoch 162 Batch 150/162] avg loss 0.000757828, throughput 9.53707K wps
Begin Testing...
[Epoch 162] train avg loss 0.000813578, dev acc 0.9100, dev avg loss 0.235683, throughput 9.15311K wps
[Epoch 163 Batch 30/162] avg loss 0.000832987, throughput 9.63042K wps
[Epoch 163 Batch 60/162] avg loss 0.000769862, throughput 9.48367K wps
[Epoch 163 Batch 90/162] avg loss 0.000810397, throughput 9.39692K wps
[Epoch 163 Batch 120/162] avg loss 0.000790574, throughput 9.33818K wps
[Epoch 163 Batch 150/162] avg loss 0.000766768, throughput 9.42976K wps
Begin Testing...
[Epoch 163] train avg loss 0.000794569, dev acc 0.9111, dev avg loss 0.235793, throughput 9.45488K wps
[Epoch 164 Batch 30/162] avg loss 0.000749019, throughput 9.55306K wps
[Epoch 164 Batch 60/162] avg loss 0.000769312, throughput 9.35973K wps
[Epoch 164 Batch 90/162] avg loss 0.000880183, throughput 9.47812K wps
[Epoch 164 Batch 120/162] avg loss 0.000738201, throughput 9.50744K wps
[Epoch 164 Batch 150/162] avg loss 0.000762934, throughput 9.31955K wps
Begin Testing...
[Epoch 164] train avg loss 0.000772451, dev acc 0.9100, dev avg loss 0.23632, throughput 9.44936K wps
[Epoch 165 Batch 30/162] avg loss 0.000771005, throughput 9.69773K wps
[Epoch 165 Batch 60/162] avg loss 0.000749823, throughput 9.37011K wps
[Epoch 165 Batch 90/162] avg loss 0.000906661, throughput 9.61047K wps
[Epoch 165 Batch 120/162] avg loss 0.000740001, throughput 9.45996K wps
[Epoch 165 Batch 150/162] avg loss 0.000830273, throughput 9.39055K wps
Begin Testing...
[Epoch 165] train avg loss 0.000794587, dev acc 0.9100, dev avg loss 0.236553, throughput 9.48944K wps
[Epoch 166 Batch 30/162] avg loss 0.000768177, throughput 9.5646K wps
[Epoch 166 Batch 60/162] avg loss 0.000824933, throughput 9.41575K wps
[Epoch 166 Batch 90/162] avg loss 0.000821158, throughput 9.27244K wps
[Epoch 166 Batch 120/162] avg loss 0.000800608, throughput 9.36184K wps
[Epoch 166 Batch 150/162] avg loss 0.000828734, throughput 9.3954K wps
Begin Testing...
[Epoch 166] train avg loss 0.000802713, dev acc 0.9100, dev avg loss 0.235764, throughput 9.39036K wps
[Epoch 167 Batch 30/162] avg loss 0.000790269, throughput 9.63225K wps
[Epoch 167 Batch 60/162] avg loss 0.00070167, throughput 9.40472K wps
[Epoch 167 Batch 90/162] avg loss 0.000787989, throughput 9.3872K wps
[Epoch 167 Batch 120/162] avg loss 0.000739007, throughput 9.42673K wps
[Epoch 167 Batch 150/162] avg loss 0.000794844, throughput 9.60576K wps
Begin Testing...
[Epoch 167] train avg loss 0.000764162, dev acc 0.9100, dev avg loss 0.236202, throughput 9.49419K wps
[Epoch 168 Batch 30/162] avg loss 0.000801519, throughput 9.55981K wps
[Epoch 168 Batch 60/162] avg loss 0.000683453, throughput 9.40529K wps
[Epoch 168 Batch 90/162] avg loss 0.000784027, throughput 9.26729K wps
[Epoch 168 Batch 120/162] avg loss 0.000743444, throughput 9.22664K wps
[Epoch 168 Batch 150/162] avg loss 0.00074639, throughput 9.33994K wps
Begin Testing...
[Epoch 168] train avg loss 0.000755654, dev acc 0.9078, dev avg loss 0.237907, throughput 9.36235K wps
[Epoch 169 Batch 30/162] avg loss 0.000706642, throughput 9.48942K wps
[Epoch 169 Batch 60/162] avg loss 0.000801273, throughput 9.38395K wps
[Epoch 169 Batch 90/162] avg loss 0.000779325, throughput 9.32791K wps
[Epoch 169 Batch 120/162] avg loss 0.000761004, throughput 9.44284K wps
[Epoch 169 Batch 150/162] avg loss 0.000729448, throughput 9.20202K wps
Begin Testing...
[Epoch 169] train avg loss 0.000756275, dev acc 0.9122, dev avg loss 0.237194, throughput 9.36017K wps
[Epoch 170 Batch 30/162] avg loss 0.00066477, throughput 9.37133K wps
[Epoch 170 Batch 60/162] avg loss 0.000778539, throughput 9.27414K wps
[Epoch 170 Batch 90/162] avg loss 0.000825341, throughput 9.49727K wps
[Epoch 170 Batch 120/162] avg loss 0.000736744, throughput 9.31899K wps
[Epoch 170 Batch 150/162] avg loss 0.000736793, throughput 9.42257K wps
Begin Testing...
[Epoch 170] train avg loss 0.000740752, dev acc 0.9100, dev avg loss 0.237422, throughput 9.3939K wps
[Epoch 171 Batch 30/162] avg loss 0.000799584, throughput 9.54575K wps
[Epoch 171 Batch 60/162] avg loss 0.000789217, throughput 9.62761K wps
[Epoch 171 Batch 90/162] avg loss 0.000725629, throughput 9.48674K wps
[Epoch 171 Batch 120/162] avg loss 0.00070809, throughput 9.38962K wps
[Epoch 171 Batch 150/162] avg loss 0.000717389, throughput 9.47381K wps
Begin Testing...
[Epoch 171] train avg loss 0.000756386, dev acc 0.9100, dev avg loss 0.237227, throughput 9.48263K wps
[Epoch 172 Batch 30/162] avg loss 0.000752299, throughput 9.6368K wps
[Epoch 172 Batch 60/162] avg loss 0.000679785, throughput 9.32707K wps
[Epoch 172 Batch 90/162] avg loss 0.000817014, throughput 9.35232K wps
[Epoch 172 Batch 120/162] avg loss 0.000623984, throughput 9.42069K wps
[Epoch 172 Batch 150/162] avg loss 0.000732433, throughput 9.23809K wps
Begin Testing...
[Epoch 172] train avg loss 0.000726017, dev acc 0.9078, dev avg loss 0.236804, throughput 9.39972K wps
[Epoch 173 Batch 30/162] avg loss 0.000755986, throughput 9.46317K wps
[Epoch 173 Batch 60/162] avg loss 0.000702394, throughput 9.3325K wps
[Epoch 173 Batch 90/162] avg loss 0.000796277, throughput 9.46001K wps
[Epoch 173 Batch 120/162] avg loss 0.000652132, throughput 9.4137K wps
[Epoch 173 Batch 150/162] avg loss 0.000687601, throughput 9.40747K wps
Begin Testing...
[Epoch 173] train avg loss 0.000724834, dev acc 0.9111, dev avg loss 0.237712, throughput 9.4084K wps
[Epoch 174 Batch 30/162] avg loss 0.000735784, throughput 9.46282K wps
[Epoch 174 Batch 60/162] avg loss 0.000731652, throughput 9.31279K wps
[Epoch 174 Batch 90/162] avg loss 0.000656412, throughput 9.47202K wps
[Epoch 174 Batch 120/162] avg loss 0.000698891, throughput 9.44206K wps
[Epoch 174 Batch 150/162] avg loss 0.000741097, throughput 9.48087K wps
Begin Testing...
[Epoch 174] train avg loss 0.000713348, dev acc 0.9078, dev avg loss 0.237118, throughput 9.44297K wps
[Epoch 175 Batch 30/162] avg loss 0.000743771, throughput 9.39445K wps
[Epoch 175 Batch 60/162] avg loss 0.000682886, throughput 9.32839K wps
[Epoch 175 Batch 90/162] avg loss 0.000649598, throughput 9.52288K wps
[Epoch 175 Batch 120/162] avg loss 0.000667763, throughput 9.41873K wps
[Epoch 175 Batch 150/162] avg loss 0.000806995, throughput 9.35282K wps
Begin Testing...
[Epoch 175] train avg loss 0.000703635, dev acc 0.9100, dev avg loss 0.237228, throughput 9.3915K wps
[Epoch 176 Batch 30/162] avg loss 0.000832819, throughput 9.40682K wps
[Epoch 176 Batch 60/162] avg loss 0.000710265, throughput 9.35888K wps
[Epoch 176 Batch 90/162] avg loss 0.000667312, throughput 9.3304K wps
[Epoch 176 Batch 120/162] avg loss 0.000637296, throughput 9.33409K wps
[Epoch 176 Batch 150/162] avg loss 0.000669669, throughput 9.34223K wps
Begin Testing...
[Epoch 176] train avg loss 0.000699142, dev acc 0.9111, dev avg loss 0.237637, throughput 9.36195K wps
[Epoch 177 Batch 30/162] avg loss 0.000693528, throughput 9.58115K wps
[Epoch 177 Batch 60/162] avg loss 0.000650959, throughput 9.5932K wps
[Epoch 177 Batch 90/162] avg loss 0.000784742, throughput 9.39822K wps
[Epoch 177 Batch 120/162] avg loss 0.000633501, throughput 9.46298K wps
[Epoch 177 Batch 150/162] avg loss 0.000686833, throughput 9.48952K wps
Begin Testing...
[Epoch 177] train avg loss 0.000681109, dev acc 0.9089, dev avg loss 0.240385, throughput 9.48392K wps
[Epoch 178 Batch 30/162] avg loss 0.000640096, throughput 9.43652K wps
[Epoch 178 Batch 60/162] avg loss 0.000612319, throughput 9.249K wps
[Epoch 178 Batch 90/162] avg loss 0.000734316, throughput 9.46672K wps
[Epoch 178 Batch 120/162] avg loss 0.000597709, throughput 9.36726K wps
[Epoch 178 Batch 150/162] avg loss 0.000672073, throughput 9.3237K wps
Begin Testing...
[Epoch 178] train avg loss 0.000665362, dev acc 0.9100, dev avg loss 0.239272, throughput 9.3727K wps
[Epoch 179 Batch 30/162] avg loss 0.000633614, throughput 9.57979K wps
[Epoch 179 Batch 60/162] avg loss 0.000687075, throughput 9.30941K wps
[Epoch 179 Batch 90/162] avg loss 0.000608268, throughput 9.52255K wps
[Epoch 179 Batch 120/162] avg loss 0.00064015, throughput 9.51619K wps
[Epoch 179 Batch 150/162] avg loss 0.000611095, throughput 9.47367K wps
Begin Testing...
[Epoch 179] train avg loss 0.000634524, dev acc 0.9111, dev avg loss 0.239065, throughput 9.45879K wps
[Epoch 180 Batch 30/162] avg loss 0.000640002, throughput 9.44108K wps
[Epoch 180 Batch 60/162] avg loss 0.00066036, throughput 9.46485K wps
[Epoch 180 Batch 90/162] avg loss 0.000695246, throughput 9.2436K wps
[Epoch 180 Batch 120/162] avg loss 0.000639377, throughput 9.31758K wps
[Epoch 180 Batch 150/162] avg loss 0.000728602, throughput 9.51451K wps
Begin Testing...
[Epoch 180] train avg loss 0.000674831, dev acc 0.9100, dev avg loss 0.238442, throughput 9.40345K wps
[Epoch 181 Batch 30/162] avg loss 0.000615604, throughput 9.36744K wps
[Epoch 181 Batch 60/162] avg loss 0.000634235, throughput 9.42493K wps
[Epoch 181 Batch 90/162] avg loss 0.000672926, throughput 9.29464K wps
[Epoch 181 Batch 120/162] avg loss 0.000726044, throughput 9.24447K wps
[Epoch 181 Batch 150/162] avg loss 0.000637706, throughput 9.29073K wps
Begin Testing...
[Epoch 181] train avg loss 0.000656826, dev acc 0.9100, dev avg loss 0.238993, throughput 9.3223K wps
[Epoch 182 Batch 30/162] avg loss 0.000695117, throughput 9.49686K wps
[Epoch 182 Batch 60/162] avg loss 0.000682117, throughput 9.32149K wps
[Epoch 182 Batch 90/162] avg loss 0.000699859, throughput 9.47571K wps
[Epoch 182 Batch 120/162] avg loss 0.000691186, throughput 9.33104K wps
[Epoch 182 Batch 150/162] avg loss 0.000641777, throughput 9.35585K wps
Begin Testing...
[Epoch 182] train avg loss 0.000676474, dev acc 0.9100, dev avg loss 0.238849, throughput 9.36984K wps
[Epoch 183 Batch 30/162] avg loss 0.000643806, throughput 9.55162K wps
[Epoch 183 Batch 60/162] avg loss 0.000675814, throughput 9.37733K wps
[Epoch 183 Batch 90/162] avg loss 0.000622648, throughput 9.45701K wps
[Epoch 183 Batch 120/162] avg loss 0.000677972, throughput 9.2751K wps
[Epoch 183 Batch 150/162] avg loss 0.00067654, throughput 9.56866K wps
Begin Testing...
[Epoch 183] train avg loss 0.000662198, dev acc 0.9100, dev avg loss 0.239296, throughput 9.43471K wps
[Epoch 184 Batch 30/162] avg loss 0.000643572, throughput 9.61286K wps
[Epoch 184 Batch 60/162] avg loss 0.000679541, throughput 9.335K wps
[Epoch 184 Batch 90/162] avg loss 0.0006128, throughput 9.36744K wps
[Epoch 184 Batch 120/162] avg loss 0.000675135, throughput 9.30609K wps
[Epoch 184 Batch 150/162] avg loss 0.0006197, throughput 9.26366K wps
Begin Testing...
[Epoch 184] train avg loss 0.000652314, dev acc 0.9089, dev avg loss 0.240087, throughput 9.37998K wps
[Epoch 185 Batch 30/162] avg loss 0.000701048, throughput 9.41467K wps
[Epoch 185 Batch 60/162] avg loss 0.000639268, throughput 9.28428K wps
[Epoch 185 Batch 90/162] avg loss 0.000594154, throughput 9.30481K wps
[Epoch 185 Batch 120/162] avg loss 0.000658006, throughput 9.23882K wps
[Epoch 185 Batch 150/162] avg loss 0.000591176, throughput 9.28895K wps
Begin Testing...
[Epoch 185] train avg loss 0.000641609, dev acc 0.9089, dev avg loss 0.24066, throughput 9.30514K wps
[Epoch 186 Batch 30/162] avg loss 0.000669224, throughput 9.59649K wps
[Epoch 186 Batch 60/162] avg loss 0.000661967, throughput 9.34397K wps
[Epoch 186 Batch 90/162] avg loss 0.000709543, throughput 9.43368K wps
[Epoch 186 Batch 120/162] avg loss 0.00056775, throughput 9.41121K wps
[Epoch 186 Batch 150/162] avg loss 0.000713935, throughput 9.23548K wps
Begin Testing...
[Epoch 186] train avg loss 0.000658315, dev acc 0.9100, dev avg loss 0.239777, throughput 9.39768K wps
[Epoch 187 Batch 30/162] avg loss 0.000600564, throughput 9.70658K wps
[Epoch 187 Batch 60/162] avg loss 0.000598476, throughput 9.38072K wps
[Epoch 187 Batch 90/162] avg loss 0.000622142, throughput 9.38747K wps
[Epoch 187 Batch 120/162] avg loss 0.000639972, throughput 9.47634K wps
[Epoch 187 Batch 150/162] avg loss 0.000609056, throughput 9.28248K wps
Begin Testing...
[Epoch 187] train avg loss 0.000613317, dev acc 0.9100, dev avg loss 0.240447, throughput 9.42922K wps
[Epoch 188 Batch 30/162] avg loss 0.000773164, throughput 9.52025K wps
[Epoch 188 Batch 60/162] avg loss 0.000617776, throughput 9.22389K wps
[Epoch 188 Batch 90/162] avg loss 0.000667963, throughput 9.36549K wps
[Epoch 188 Batch 120/162] avg loss 0.000657483, throughput 9.47177K wps
[Epoch 188 Batch 150/162] avg loss 0.000693193, throughput 9.39425K wps
Begin Testing...
[Epoch 188] train avg loss 0.000672095, dev acc 0.9100, dev avg loss 0.23955, throughput 9.38678K wps
[Epoch 189 Batch 30/162] avg loss 0.000638646, throughput 9.53631K wps
[Epoch 189 Batch 60/162] avg loss 0.000609101, throughput 9.43035K wps
[Epoch 189 Batch 90/162] avg loss 0.000576269, throughput 9.30208K wps
[Epoch 189 Batch 120/162] avg loss 0.00057591, throughput 9.44965K wps
[Epoch 189 Batch 150/162] avg loss 0.000644149, throughput 9.20806K wps
Begin Testing...
[Epoch 189] train avg loss 0.000622232, dev acc 0.9111, dev avg loss 0.239722, throughput 9.37656K wps
[Epoch 190 Batch 30/162] avg loss 0.000607203, throughput 9.60465K wps
[Epoch 190 Batch 60/162] avg loss 0.000647973, throughput 9.3575K wps
[Epoch 190 Batch 90/162] avg loss 0.000602054, throughput 9.25115K wps
[Epoch 190 Batch 120/162] avg loss 0.000639988, throughput 9.40754K wps
[Epoch 190 Batch 150/162] avg loss 0.000545217, throughput 9.29879K wps
Begin Testing...
[Epoch 190] train avg loss 0.000615136, dev acc 0.9100, dev avg loss 0.239798, throughput 9.37632K wps
[Epoch 191 Batch 30/162] avg loss 0.00058518, throughput 9.59361K wps
[Epoch 191 Batch 60/162] avg loss 0.000575453, throughput 9.39359K wps
[Epoch 191 Batch 90/162] avg loss 0.00065633, throughput 9.43916K wps
[Epoch 191 Batch 120/162] avg loss 0.000648766, throughput 9.27674K wps
[Epoch 191 Batch 150/162] avg loss 0.000553561, throughput 9.4239K wps
Begin Testing...
[Epoch 191] train avg loss 0.000602676, dev acc 0.9078, dev avg loss 0.240976, throughput 9.3981K wps
[Epoch 192 Batch 30/162] avg loss 0.000592778, throughput 9.44805K wps
[Epoch 192 Batch 60/162] avg loss 0.000596172, throughput 9.42278K wps
[Epoch 192 Batch 90/162] avg loss 0.000573371, throughput 9.26366K wps
[Epoch 192 Batch 120/162] avg loss 0.000575005, throughput 9.25155K wps
[Epoch 192 Batch 150/162] avg loss 0.000636236, throughput 9.32154K wps
Begin Testing...
[Epoch 192] train avg loss 0.000593533, dev acc 0.9100, dev avg loss 0.240407, throughput 9.35108K wps
[Epoch 193 Batch 30/162] avg loss 0.000658213, throughput 9.72959K wps
[Epoch 193 Batch 60/162] avg loss 0.000577393, throughput 9.44005K wps
[Epoch 193 Batch 90/162] avg loss 0.000586158, throughput 9.30096K wps
[Epoch 193 Batch 120/162] avg loss 0.000573065, throughput 9.52292K wps
[Epoch 193 Batch 150/162] avg loss 0.000586835, throughput 9.41289K wps
Begin Testing...
[Epoch 193] train avg loss 0.000595349, dev acc 0.9100, dev avg loss 0.240989, throughput 9.46127K wps
[Epoch 194 Batch 30/162] avg loss 0.000662096, throughput 9.62171K wps
[Epoch 194 Batch 60/162] avg loss 0.000583193, throughput 9.47202K wps
[Epoch 194 Batch 90/162] avg loss 0.000577051, throughput 9.35759K wps
[Epoch 194 Batch 120/162] avg loss 0.000570031, throughput 9.41577K wps
[Epoch 194 Batch 150/162] avg loss 0.0006127, throughput 9.41688K wps
Begin Testing...
[Epoch 194] train avg loss 0.000599261, dev acc 0.9089, dev avg loss 0.241427, throughput 9.43003K wps
[Epoch 195 Batch 30/162] avg loss 0.000605583, throughput 9.50723K wps
[Epoch 195 Batch 60/162] avg loss 0.000595955, throughput 9.30366K wps
[Epoch 195 Batch 90/162] avg loss 0.000495778, throughput 9.33759K wps
[Epoch 195 Batch 120/162] avg loss 0.000552032, throughput 9.37243K wps
[Epoch 195 Batch 150/162] avg loss 0.000550038, throughput 9.5454K wps
Begin Testing...
[Epoch 195] train avg loss 0.00057086, dev acc 0.9089, dev avg loss 0.241859, throughput 9.40634K wps
[Epoch 196 Batch 30/162] avg loss 0.000520767, throughput 9.49918K wps
[Epoch 196 Batch 60/162] avg loss 0.000716071, throughput 9.22411K wps
[Epoch 196 Batch 90/162] avg loss 0.000616169, throughput 9.24138K wps
[Epoch 196 Batch 120/162] avg loss 0.000636962, throughput 9.54738K wps
[Epoch 196 Batch 150/162] avg loss 0.000596003, throughput 9.23651K wps
Begin Testing...
[Epoch 196] train avg loss 0.000611805, dev acc 0.9089, dev avg loss 0.241178, throughput 9.33831K wps
[Epoch 197 Batch 30/162] avg loss 0.000599157, throughput 9.55958K wps
[Epoch 197 Batch 60/162] avg loss 0.000653632, throughput 9.36023K wps
[Epoch 197 Batch 90/162] avg loss 0.000706283, throughput 9.3413K wps
[Epoch 197 Batch 120/162] avg loss 0.000546314, throughput 9.47379K wps
[Epoch 197 Batch 150/162] avg loss 0.00065786, throughput 9.35161K wps
Begin Testing...
[Epoch 197] train avg loss 0.000622945, dev acc 0.9100, dev avg loss 0.242742, throughput 9.42742K wps
[Epoch 198 Batch 30/162] avg loss 0.000501483, throughput 9.53983K wps
[Epoch 198 Batch 60/162] avg loss 0.000557501, throughput 9.52597K wps
[Epoch 198 Batch 90/162] avg loss 0.00052395, throughput 9.19008K wps
[Epoch 198 Batch 120/162] avg loss 0.000651491, throughput 9.311K wps
[Epoch 198 Batch 150/162] avg loss 0.000543924, throughput 9.3464K wps
Begin Testing...
[Epoch 198] train avg loss 0.000554811, dev acc 0.9089, dev avg loss 0.242213, throughput 9.38769K wps
[Epoch 199 Batch 30/162] avg loss 0.000553013, throughput 9.62338K wps
[Epoch 199 Batch 60/162] avg loss 0.000695971, throughput 9.38402K wps
[Epoch 199 Batch 90/162] avg loss 0.000585834, throughput 9.46732K wps
[Epoch 199 Batch 120/162] avg loss 0.000557989, throughput 9.41661K wps
[Epoch 199 Batch 150/162] avg loss 0.000537312, throughput 9.31667K wps
Begin Testing...
[Epoch 199] train avg loss 0.000579422, dev acc 0.9089, dev avg loss 0.242216, throughput 9.42672K wps
Test loss 0.240849, test acc 0.9000
Total time cost 449.56s
[Epoch 0 Batch 30/162] avg loss 0.0140726, throughput 7.39141K wps
[Epoch 0 Batch 60/162] avg loss 0.0140383, throughput 9.46102K wps
[Epoch 0 Batch 90/162] avg loss 0.0137832, throughput 9.39688K wps
[Epoch 0 Batch 120/162] avg loss 0.0136756, throughput 9.50621K wps
[Epoch 0 Batch 150/162] avg loss 0.0135761, throughput 9.26843K wps
Begin Testing...
[Epoch 0] train avg loss 0.0138065, dev acc 0.5733, dev avg loss 0.670652, throughput 8.96478K wps
Observed Improvement.
Begin Testing...
[Epoch 1 Batch 30/162] avg loss 0.0133764, throughput 9.56917K wps
[Epoch 1 Batch 60/162] avg loss 0.0131806, throughput 9.34241K wps
[Epoch 1 Batch 90/162] avg loss 0.0131952, throughput 9.43099K wps
[Epoch 1 Batch 120/162] avg loss 0.0130626, throughput 9.4115K wps
[Epoch 1 Batch 150/162] avg loss 0.0128902, throughput 9.33768K wps
Begin Testing...
[Epoch 1] train avg loss 0.0131225, dev acc 0.8422, dev avg loss 0.635717, throughput 9.40402K wps
Observed Improvement.
Begin Testing...
[Epoch 2 Batch 30/162] avg loss 0.0126453, throughput 9.61325K wps
[Epoch 2 Batch 60/162] avg loss 0.0126212, throughput 9.53795K wps
[Epoch 2 Batch 90/162] avg loss 0.0126731, throughput 9.37195K wps
[Epoch 2 Batch 120/162] avg loss 0.0122695, throughput 9.29239K wps
[Epoch 2 Batch 150/162] avg loss 0.0121763, throughput 9.36317K wps
Begin Testing...
[Epoch 2] train avg loss 0.0124558, dev acc 0.8644, dev avg loss 0.600753, throughput 9.43269K wps
Observed Improvement.
Begin Testing...
[Epoch 3 Batch 30/162] avg loss 0.0120106, throughput 9.59417K wps
[Epoch 3 Batch 60/162] avg loss 0.0119057, throughput 9.59146K wps
[Epoch 3 Batch 90/162] avg loss 0.0116017, throughput 9.41008K wps
[Epoch 3 Batch 120/162] avg loss 0.0116738, throughput 9.36862K wps
[Epoch 3 Batch 150/162] avg loss 0.0115794, throughput 9.28908K wps
Begin Testing...
[Epoch 3] train avg loss 0.0117007, dev acc 0.8378, dev avg loss 0.561953, throughput 9.4522K wps
[Epoch 4 Batch 30/162] avg loss 0.0112274, throughput 9.60132K wps
[Epoch 4 Batch 60/162] avg loss 0.0109353, throughput 9.43363K wps
[Epoch 4 Batch 90/162] avg loss 0.0109218, throughput 9.47301K wps
[Epoch 4 Batch 120/162] avg loss 0.0107493, throughput 9.49994K wps
[Epoch 4 Batch 150/162] avg loss 0.0106113, throughput 9.24157K wps
Begin Testing...
[Epoch 4] train avg loss 0.0108482, dev acc 0.8789, dev avg loss 0.518359, throughput 9.44586K wps
Observed Improvement.
Begin Testing...
[Epoch 5 Batch 30/162] avg loss 0.0104453, throughput 9.54439K wps
[Epoch 5 Batch 60/162] avg loss 0.0101523, throughput 9.42843K wps
[Epoch 5 Batch 90/162] avg loss 0.0101504, throughput 9.34022K wps
[Epoch 5 Batch 120/162] avg loss 0.0100014, throughput 9.35983K wps
[Epoch 5 Batch 150/162] avg loss 0.00994029, throughput 9.22321K wps
Begin Testing...
[Epoch 5] train avg loss 0.0101024, dev acc 0.8756, dev avg loss 0.478165, throughput 9.37346K wps
[Epoch 6 Batch 30/162] avg loss 0.00950364, throughput 9.57595K wps
[Epoch 6 Batch 60/162] avg loss 0.00935815, throughput 9.32741K wps
[Epoch 6 Batch 90/162] avg loss 0.00937507, throughput 9.42836K wps
[Epoch 6 Batch 120/162] avg loss 0.00913393, throughput 9.3183K wps
[Epoch 6 Batch 150/162] avg loss 0.00919421, throughput 9.48266K wps
Begin Testing...
[Epoch 6] train avg loss 0.00929135, dev acc 0.8822, dev avg loss 0.440005, throughput 9.41961K wps
Observed Improvement.
Begin Testing...
[Epoch 7 Batch 30/162] avg loss 0.00869316, throughput 9.4913K wps
[Epoch 7 Batch 60/162] avg loss 0.0090485, throughput 9.26146K wps
[Epoch 7 Batch 90/162] avg loss 0.0085844, throughput 9.32471K wps
[Epoch 7 Batch 120/162] avg loss 0.00854258, throughput 9.47627K wps
[Epoch 7 Batch 150/162] avg loss 0.00856236, throughput 9.35085K wps
Begin Testing...
[Epoch 7] train avg loss 0.00863502, dev acc 0.8800, dev avg loss 0.409862, throughput 9.40692K wps
[Epoch 8 Batch 30/162] avg loss 0.00828883, throughput 9.60198K wps
[Epoch 8 Batch 60/162] avg loss 0.00796956, throughput 9.41286K wps
[Epoch 8 Batch 90/162] avg loss 0.00830067, throughput 9.3611K wps
[Epoch 8 Batch 120/162] avg loss 0.00792313, throughput 9.27207K wps
[Epoch 8 Batch 150/162] avg loss 0.00797801, throughput 9.58641K wps
Begin Testing...
[Epoch 8] train avg loss 0.00807945, dev acc 0.8833, dev avg loss 0.384831, throughput 9.43896K wps
Observed Improvement.
Begin Testing...
[Epoch 9 Batch 30/162] avg loss 0.00766453, throughput 9.62713K wps
[Epoch 9 Batch 60/162] avg loss 0.00750802, throughput 9.4043K wps
[Epoch 9 Batch 90/162] avg loss 0.00775625, throughput 9.43724K wps
[Epoch 9 Batch 120/162] avg loss 0.00767138, throughput 9.52052K wps
[Epoch 9 Batch 150/162] avg loss 0.00756405, throughput 9.47361K wps
Begin Testing...
[Epoch 9] train avg loss 0.00761668, dev acc 0.8867, dev avg loss 0.365238, throughput 9.4848K wps
Observed Improvement.
Begin Testing...
[Epoch 10 Batch 30/162] avg loss 0.00731415, throughput 9.53568K wps
[Epoch 10 Batch 60/162] avg loss 0.00770711, throughput 9.33454K wps
[Epoch 10 Batch 90/162] avg loss 0.0070536, throughput 9.47978K wps
[Epoch 10 Batch 120/162] avg loss 0.00719121, throughput 9.49022K wps
[Epoch 10 Batch 150/162] avg loss 0.00721707, throughput 9.42229K wps
Begin Testing...
[Epoch 10] train avg loss 0.00729648, dev acc 0.8900, dev avg loss 0.349222, throughput 9.44327K wps
Observed Improvement.
Begin Testing...
[Epoch 11 Batch 30/162] avg loss 0.00690867, throughput 9.54089K wps
[Epoch 11 Batch 60/162] avg loss 0.00717689, throughput 9.48441K wps
[Epoch 11 Batch 90/162] avg loss 0.0068918, throughput 9.28624K wps
[Epoch 11 Batch 120/162] avg loss 0.0071228, throughput 9.44055K wps
[Epoch 11 Batch 150/162] avg loss 0.00669772, throughput 9.31503K wps
Begin Testing...
[Epoch 11] train avg loss 0.00697559, dev acc 0.8944, dev avg loss 0.336424, throughput 9.41227K wps
Observed Improvement.
Begin Testing...
[Epoch 12 Batch 30/162] avg loss 0.00693336, throughput 9.54545K wps
[Epoch 12 Batch 60/162] avg loss 0.00663406, throughput 9.35629K wps
[Epoch 12 Batch 90/162] avg loss 0.00648058, throughput 9.51424K wps
[Epoch 12 Batch 120/162] avg loss 0.00676808, throughput 9.42161K wps
[Epoch 12 Batch 150/162] avg loss 0.00687599, throughput 9.23493K wps
Begin Testing...
[Epoch 12] train avg loss 0.00674375, dev acc 0.8922, dev avg loss 0.326054, throughput 9.39659K wps
[Epoch 13 Batch 30/162] avg loss 0.00646833, throughput 9.53929K wps
[Epoch 13 Batch 60/162] avg loss 0.00647085, throughput 9.55425K wps
[Epoch 13 Batch 90/162] avg loss 0.00639204, throughput 9.33323K wps
[Epoch 13 Batch 120/162] avg loss 0.00652885, throughput 9.37294K wps
[Epoch 13 Batch 150/162] avg loss 0.00633697, throughput 9.29223K wps
Begin Testing...
[Epoch 13] train avg loss 0.006429, dev acc 0.8967, dev avg loss 0.315885, throughput 9.41412K wps
Observed Improvement.
Begin Testing...
[Epoch 14 Batch 30/162] avg loss 0.00604977, throughput 9.72143K wps
[Epoch 14 Batch 60/162] avg loss 0.00650789, throughput 9.20917K wps
[Epoch 14 Batch 90/162] avg loss 0.00629555, throughput 9.21399K wps
[Epoch 14 Batch 120/162] avg loss 0.00603097, throughput 9.34961K wps
[Epoch 14 Batch 150/162] avg loss 0.00663879, throughput 9.23972K wps
Begin Testing...
[Epoch 14] train avg loss 0.00628041, dev acc 0.9000, dev avg loss 0.308417, throughput 9.36123K wps
Observed Improvement.
Begin Testing...
[Epoch 15 Batch 30/162] avg loss 0.00609191, throughput 9.70632K wps
[Epoch 15 Batch 60/162] avg loss 0.00608192, throughput 9.40351K wps
[Epoch 15 Batch 90/162] avg loss 0.00635003, throughput 9.35678K wps
[Epoch 15 Batch 120/162] avg loss 0.00639329, throughput 9.49997K wps
[Epoch 15 Batch 150/162] avg loss 0.00603197, throughput 9.32145K wps
Begin Testing...
[Epoch 15] train avg loss 0.0061382, dev acc 0.8967, dev avg loss 0.303301, throughput 9.45129K wps
[Epoch 16 Batch 30/162] avg loss 0.00598804, throughput 9.71958K wps
[Epoch 16 Batch 60/162] avg loss 0.00622243, throughput 9.46557K wps
[Epoch 16 Batch 90/162] avg loss 0.00580353, throughput 9.31757K wps
[Epoch 16 Batch 120/162] avg loss 0.00564709, throughput 9.66407K wps
[Epoch 16 Batch 150/162] avg loss 0.00628037, throughput 9.39222K wps
Begin Testing...
[Epoch 16] train avg loss 0.00594148, dev acc 0.8989, dev avg loss 0.295985, throughput 9.50332K wps
[Epoch 17 Batch 30/162] avg loss 0.00580788, throughput 9.65908K wps
[Epoch 17 Batch 60/162] avg loss 0.00586131, throughput 9.35996K wps
[Epoch 17 Batch 90/162] avg loss 0.00601423, throughput 9.41336K wps
[Epoch 17 Batch 120/162] avg loss 0.005885, throughput 9.35865K wps
[Epoch 17 Batch 150/162] avg loss 0.00577548, throughput 9.40211K wps
Begin Testing...
[Epoch 17] train avg loss 0.00585919, dev acc 0.9000, dev avg loss 0.290959, throughput 9.44395K wps
Observed Improvement.
Begin Testing...
[Epoch 18 Batch 30/162] avg loss 0.00584757, throughput 9.55851K wps
[Epoch 18 Batch 60/162] avg loss 0.00571791, throughput 9.4662K wps
[Epoch 18 Batch 90/162] avg loss 0.0056391, throughput 9.49958K wps
[Epoch 18 Batch 120/162] avg loss 0.00571704, throughput 9.29116K wps
[Epoch 18 Batch 150/162] avg loss 0.00547214, throughput 9.55196K wps
Begin Testing...
[Epoch 18] train avg loss 0.00566763, dev acc 0.9078, dev avg loss 0.286611, throughput 9.45075K wps
Observed Improvement.
Begin Testing...
[Epoch 19 Batch 30/162] avg loss 0.00576338, throughput 9.73011K wps
[Epoch 19 Batch 60/162] avg loss 0.00537843, throughput 9.51917K wps
[Epoch 19 Batch 90/162] avg loss 0.00584516, throughput 9.29239K wps
[Epoch 19 Batch 120/162] avg loss 0.00548162, throughput 9.33315K wps
[Epoch 19 Batch 150/162] avg loss 0.00554966, throughput 9.51513K wps
Begin Testing...
[Epoch 19] train avg loss 0.00560618, dev acc 0.9011, dev avg loss 0.280949, throughput 9.46913K wps
[Epoch 20 Batch 30/162] avg loss 0.00561643, throughput 9.68142K wps
[Epoch 20 Batch 60/162] avg loss 0.00528191, throughput 9.5243K wps
[Epoch 20 Batch 90/162] avg loss 0.00544346, throughput 9.56445K wps
[Epoch 20 Batch 120/162] avg loss 0.00574723, throughput 9.53746K wps
[Epoch 20 Batch 150/162] avg loss 0.00529826, throughput 9.39591K wps
Begin Testing...
[Epoch 20] train avg loss 0.00544785, dev acc 0.9056, dev avg loss 0.276786, throughput 9.53617K wps
[Epoch 21 Batch 30/162] avg loss 0.00524969, throughput 9.57011K wps
[Epoch 21 Batch 60/162] avg loss 0.00548531, throughput 9.49003K wps
[Epoch 21 Batch 90/162] avg loss 0.00517617, throughput 9.33073K wps
[Epoch 21 Batch 120/162] avg loss 0.00527949, throughput 9.47007K wps
[Epoch 21 Batch 150/162] avg loss 0.00526136, throughput 9.34523K wps
Begin Testing...
[Epoch 21] train avg loss 0.00531254, dev acc 0.9056, dev avg loss 0.273125, throughput 9.45373K wps
[Epoch 22 Batch 30/162] avg loss 0.00510113, throughput 9.6955K wps
[Epoch 22 Batch 60/162] avg loss 0.00489239, throughput 9.56468K wps
[Epoch 22 Batch 90/162] avg loss 0.00510778, throughput 9.40444K wps
[Epoch 22 Batch 120/162] avg loss 0.00564604, throughput 9.3401K wps
[Epoch 22 Batch 150/162] avg loss 0.00520852, throughput 9.40036K wps
Begin Testing...
[Epoch 22] train avg loss 0.0052046, dev acc 0.9067, dev avg loss 0.270058, throughput 9.46844K wps
[Epoch 23 Batch 30/162] avg loss 0.00501132, throughput 9.45797K wps
[Epoch 23 Batch 60/162] avg loss 0.00527927, throughput 9.50068K wps
[Epoch 23 Batch 90/162] avg loss 0.00513286, throughput 9.45796K wps
[Epoch 23 Batch 120/162] avg loss 0.00516424, throughput 9.5026K wps
[Epoch 23 Batch 150/162] avg loss 0.00464099, throughput 9.45687K wps
Begin Testing...
[Epoch 23] train avg loss 0.00506607, dev acc 0.9056, dev avg loss 0.266341, throughput 9.47333K wps
[Epoch 24 Batch 30/162] avg loss 0.00508284, throughput 9.45475K wps
[Epoch 24 Batch 60/162] avg loss 0.00504651, throughput 9.29793K wps
[Epoch 24 Batch 90/162] avg loss 0.00519296, throughput 9.44859K wps
[Epoch 24 Batch 120/162] avg loss 0.00511292, throughput 9.3703K wps
[Epoch 24 Batch 150/162] avg loss 0.00481905, throughput 9.3752K wps
Begin Testing...
[Epoch 24] train avg loss 0.00500738, dev acc 0.9100, dev avg loss 0.262825, throughput 9.38826K wps
Observed Improvement.
Begin Testing...
[Epoch 25 Batch 30/162] avg loss 0.00489864, throughput 9.56647K wps
[Epoch 25 Batch 60/162] avg loss 0.00520048, throughput 9.55132K wps
[Epoch 25 Batch 90/162] avg loss 0.0046729, throughput 9.5449K wps
[Epoch 25 Batch 120/162] avg loss 0.00470683, throughput 9.48367K wps
[Epoch 25 Batch 150/162] avg loss 0.00493507, throughput 9.40231K wps
Begin Testing...
[Epoch 25] train avg loss 0.00494214, dev acc 0.9067, dev avg loss 0.260009, throughput 9.49344K wps
[Epoch 26 Batch 30/162] avg loss 0.00523675, throughput 9.73317K wps
[Epoch 26 Batch 60/162] avg loss 0.00442369, throughput 9.4471K wps
[Epoch 26 Batch 90/162] avg loss 0.00478524, throughput 9.50189K wps
[Epoch 26 Batch 120/162] avg loss 0.00457294, throughput 9.33065K wps
[Epoch 26 Batch 150/162] avg loss 0.00506119, throughput 9.34235K wps
Begin Testing...
[Epoch 26] train avg loss 0.00482379, dev acc 0.9111, dev avg loss 0.25761, throughput 9.47817K wps
Observed Improvement.
Begin Testing...
[Epoch 27 Batch 30/162] avg loss 0.00465707, throughput 9.55972K wps
[Epoch 27 Batch 60/162] avg loss 0.00476009, throughput 9.29239K wps
[Epoch 27 Batch 90/162] avg loss 0.00458761, throughput 9.3975K wps
[Epoch 27 Batch 120/162] avg loss 0.00510205, throughput 9.32087K wps
[Epoch 27 Batch 150/162] avg loss 0.00485559, throughput 9.37847K wps
Begin Testing...
[Epoch 27] train avg loss 0.00477718, dev acc 0.9111, dev avg loss 0.255202, throughput 9.40621K wps
Observed Improvement.
Begin Testing...
[Epoch 28 Batch 30/162] avg loss 0.00459179, throughput 9.4494K wps
[Epoch 28 Batch 60/162] avg loss 0.00477292, throughput 9.29028K wps
[Epoch 28 Batch 90/162] avg loss 0.00475007, throughput 9.35829K wps
[Epoch 28 Batch 120/162] avg loss 0.00456027, throughput 9.31185K wps
[Epoch 28 Batch 150/162] avg loss 0.00476986, throughput 9.46715K wps
Begin Testing...
[Epoch 28] train avg loss 0.00464604, dev acc 0.9111, dev avg loss 0.252468, throughput 9.36739K wps
Observed Improvement.
Begin Testing...
[Epoch 29 Batch 30/162] avg loss 0.00495152, throughput 9.55223K wps
[Epoch 29 Batch 60/162] avg loss 0.00472496, throughput 9.37179K wps
[Epoch 29 Batch 90/162] avg loss 0.00440736, throughput 9.48296K wps
[Epoch 29 Batch 120/162] avg loss 0.00450659, throughput 9.41146K wps
[Epoch 29 Batch 150/162] avg loss 0.00443813, throughput 9.42477K wps
Begin Testing...
[Epoch 29] train avg loss 0.00455219, dev acc 0.9122, dev avg loss 0.250449, throughput 9.45928K wps
Observed Improvement.
Begin Testing...
[Epoch 30 Batch 30/162] avg loss 0.00450807, throughput 9.44462K wps
[Epoch 30 Batch 60/162] avg loss 0.00411041, throughput 9.24222K wps
[Epoch 30 Batch 90/162] avg loss 0.00421674, throughput 9.40247K wps
[Epoch 30 Batch 120/162] avg loss 0.00494776, throughput 9.40438K wps
[Epoch 30 Batch 150/162] avg loss 0.00468724, throughput 9.25262K wps
Begin Testing...
[Epoch 30] train avg loss 0.00453674, dev acc 0.9133, dev avg loss 0.249946, throughput 9.34694K wps
Observed Improvement.
Begin Testing...
[Epoch 31 Batch 30/162] avg loss 0.00443646, throughput 9.5003K wps
[Epoch 31 Batch 60/162] avg loss 0.00432406, throughput 9.41434K wps
[Epoch 31 Batch 90/162] avg loss 0.00490161, throughput 9.54533K wps
[Epoch 31 Batch 120/162] avg loss 0.00435793, throughput 9.5111K wps
[Epoch 31 Batch 150/162] avg loss 0.00419358, throughput 9.38685K wps
Begin Testing...
[Epoch 31] train avg loss 0.00443906, dev acc 0.9122, dev avg loss 0.247184, throughput 9.47838K wps
[Epoch 32 Batch 30/162] avg loss 0.0043791, throughput 9.59499K wps
[Epoch 32 Batch 60/162] avg loss 0.00418867, throughput 9.55898K wps
[Epoch 32 Batch 90/162] avg loss 0.00421759, throughput 9.51839K wps
[Epoch 32 Batch 120/162] avg loss 0.0043961, throughput 9.38598K wps
[Epoch 32 Batch 150/162] avg loss 0.00467612, throughput 9.29118K wps
Begin Testing...
[Epoch 32] train avg loss 0.0043886, dev acc 0.9100, dev avg loss 0.245101, throughput 9.4542K wps
[Epoch 33 Batch 30/162] avg loss 0.0041956, throughput 9.51613K wps
[Epoch 33 Batch 60/162] avg loss 0.00444644, throughput 9.36423K wps
[Epoch 33 Batch 90/162] avg loss 0.00451396, throughput 9.23207K wps
[Epoch 33 Batch 120/162] avg loss 0.00449578, throughput 9.38102K wps
[Epoch 33 Batch 150/162] avg loss 0.00411528, throughput 9.23957K wps
Begin Testing...
[Epoch 33] train avg loss 0.00432072, dev acc 0.9111, dev avg loss 0.242991, throughput 9.32943K wps
[Epoch 34 Batch 30/162] avg loss 0.00404961, throughput 9.41653K wps
[Epoch 34 Batch 60/162] avg loss 0.00407636, throughput 9.388K wps
[Epoch 34 Batch 90/162] avg loss 0.00410159, throughput 9.3557K wps
[Epoch 34 Batch 120/162] avg loss 0.00455285, throughput 9.37711K wps
[Epoch 34 Batch 150/162] avg loss 0.00447952, throughput 9.40347K wps
Begin Testing...
[Epoch 34] train avg loss 0.00422945, dev acc 0.9167, dev avg loss 0.24166, throughput 9.38692K wps
Observed Improvement.
Begin Testing...
[Epoch 35 Batch 30/162] avg loss 0.00405558, throughput 9.56081K wps
[Epoch 35 Batch 60/162] avg loss 0.00425252, throughput 9.53486K wps
[Epoch 35 Batch 90/162] avg loss 0.00398845, throughput 9.32025K wps
[Epoch 35 Batch 120/162] avg loss 0.00420538, throughput 9.40946K wps
[Epoch 35 Batch 150/162] avg loss 0.00426934, throughput 9.36935K wps
Begin Testing...
[Epoch 35] train avg loss 0.00418682, dev acc 0.9111, dev avg loss 0.242246, throughput 9.43124K wps
[Epoch 36 Batch 30/162] avg loss 0.00448431, throughput 9.55008K wps
[Epoch 36 Batch 60/162] avg loss 0.00398378, throughput 9.34084K wps
[Epoch 36 Batch 90/162] avg loss 0.00405471, throughput 9.26619K wps
[Epoch 36 Batch 120/162] avg loss 0.00436351, throughput 9.30262K wps
[Epoch 36 Batch 150/162] avg loss 0.00359431, throughput 9.45511K wps
Begin Testing...
[Epoch 36] train avg loss 0.00408005, dev acc 0.9156, dev avg loss 0.23805, throughput 9.37372K wps
[Epoch 37 Batch 30/162] avg loss 0.00443964, throughput 9.59793K wps
[Epoch 37 Batch 60/162] avg loss 0.00417021, throughput 9.25536K wps
[Epoch 37 Batch 90/162] avg loss 0.00369751, throughput 9.48843K wps
[Epoch 37 Batch 120/162] avg loss 0.00391587, throughput 9.25898K wps
[Epoch 37 Batch 150/162] avg loss 0.00380923, throughput 9.34119K wps
Begin Testing...
[Epoch 37] train avg loss 0.00403013, dev acc 0.9122, dev avg loss 0.237798, throughput 9.3771K wps
[Epoch 38 Batch 30/162] avg loss 0.00368733, throughput 9.55393K wps
[Epoch 38 Batch 60/162] avg loss 0.00391182, throughput 9.37262K wps
[Epoch 38 Batch 90/162] avg loss 0.00396453, throughput 9.43634K wps
[Epoch 38 Batch 120/162] avg loss 0.0041397, throughput 9.44248K wps
[Epoch 38 Batch 150/162] avg loss 0.00424483, throughput 9.51457K wps
Begin Testing...
[Epoch 38] train avg loss 0.00396421, dev acc 0.9156, dev avg loss 0.235457, throughput 9.46659K wps
[Epoch 39 Batch 30/162] avg loss 0.00384706, throughput 9.46862K wps
[Epoch 39 Batch 60/162] avg loss 0.00376509, throughput 9.39534K wps
[Epoch 39 Batch 90/162] avg loss 0.00371358, throughput 9.39119K wps
[Epoch 39 Batch 120/162] avg loss 0.0036837, throughput 9.38283K wps
[Epoch 39 Batch 150/162] avg loss 0.00421515, throughput 9.36233K wps
Begin Testing...
[Epoch 39] train avg loss 0.00384098, dev acc 0.9156, dev avg loss 0.233676, throughput 9.39655K wps
[Epoch 40 Batch 30/162] avg loss 0.0038304, throughput 9.50969K wps
[Epoch 40 Batch 60/162] avg loss 0.00371572, throughput 9.41708K wps
[Epoch 40 Batch 90/162] avg loss 0.00422123, throughput 9.42236K wps
[Epoch 40 Batch 120/162] avg loss 0.00393215, throughput 9.52516K wps
[Epoch 40 Batch 150/162] avg loss 0.00343454, throughput 9.39763K wps
Begin Testing...
[Epoch 40] train avg loss 0.00386679, dev acc 0.9156, dev avg loss 0.233001, throughput 9.45388K wps
[Epoch 41 Batch 30/162] avg loss 0.00365855, throughput 9.71311K wps
[Epoch 41 Batch 60/162] avg loss 0.00360422, throughput 9.37947K wps
[Epoch 41 Batch 90/162] avg loss 0.00393725, throughput 9.3635K wps
[Epoch 41 Batch 120/162] avg loss 0.00386439, throughput 9.22235K wps
[Epoch 41 Batch 150/162] avg loss 0.00364322, throughput 9.45094K wps
Begin Testing...
[Epoch 41] train avg loss 0.00376707, dev acc 0.9111, dev avg loss 0.232585, throughput 9.43671K wps
[Epoch 42 Batch 30/162] avg loss 0.00356105, throughput 9.66854K wps
[Epoch 42 Batch 60/162] avg loss 0.00405164, throughput 9.46931K wps
[Epoch 42 Batch 90/162] avg loss 0.00365844, throughput 9.45482K wps
[Epoch 42 Batch 120/162] avg loss 0.00366082, throughput 9.3757K wps
[Epoch 42 Batch 150/162] avg loss 0.00370244, throughput 9.62455K wps
Begin Testing...
[Epoch 42] train avg loss 0.00371542, dev acc 0.9156, dev avg loss 0.230118, throughput 9.50509K wps
[Epoch 43 Batch 30/162] avg loss 0.0037323, throughput 9.6174K wps
[Epoch 43 Batch 60/162] avg loss 0.00351933, throughput 9.3256K wps
[Epoch 43 Batch 90/162] avg loss 0.00373353, throughput 9.42253K wps
[Epoch 43 Batch 120/162] avg loss 0.00364719, throughput 9.56141K wps
[Epoch 43 Batch 150/162] avg loss 0.00365445, throughput 9.23229K wps
Begin Testing...
[Epoch 43] train avg loss 0.00364265, dev acc 0.9178, dev avg loss 0.228829, throughput 9.43623K wps
Observed Improvement.
Begin Testing...
[Epoch 44 Batch 30/162] avg loss 0.00348581, throughput 9.70121K wps
[Epoch 44 Batch 60/162] avg loss 0.00367857, throughput 9.21387K wps
[Epoch 44 Batch 90/162] avg loss 0.00366131, throughput 9.38915K wps
[Epoch 44 Batch 120/162] avg loss 0.00388536, throughput 9.33646K wps
[Epoch 44 Batch 150/162] avg loss 0.00338824, throughput 9.41887K wps
Begin Testing...
[Epoch 44] train avg loss 0.00362926, dev acc 0.9167, dev avg loss 0.22834, throughput 9.37063K wps
[Epoch 45 Batch 30/162] avg loss 0.0035622, throughput 9.46774K wps
[Epoch 45 Batch 60/162] avg loss 0.00341074, throughput 9.43816K wps
[Epoch 45 Batch 90/162] avg loss 0.00357691, throughput 9.25185K wps
[Epoch 45 Batch 120/162] avg loss 0.00365014, throughput 9.48571K wps
[Epoch 45 Batch 150/162] avg loss 0.00372701, throughput 9.30793K wps
Begin Testing...
[Epoch 45] train avg loss 0.0035557, dev acc 0.9189, dev avg loss 0.227091, throughput 9.38509K wps
Observed Improvement.
Begin Testing...
[Epoch 46 Batch 30/162] avg loss 0.00373619, throughput 9.54032K wps
[Epoch 46 Batch 60/162] avg loss 0.00390801, throughput 9.21746K wps
[Epoch 46 Batch 90/162] avg loss 0.00321295, throughput 9.44002K wps
[Epoch 46 Batch 120/162] avg loss 0.00347194, throughput 9.64346K wps
[Epoch 46 Batch 150/162] avg loss 0.0032535, throughput 9.49093K wps
Begin Testing...
[Epoch 46] train avg loss 0.00349264, dev acc 0.9178, dev avg loss 0.225902, throughput 9.44335K wps
[Epoch 47 Batch 30/162] avg loss 0.00321667, throughput 9.52042K wps
[Epoch 47 Batch 60/162] avg loss 0.0036551, throughput 9.5516K wps
[Epoch 47 Batch 90/162] avg loss 0.00330308, throughput 9.26896K wps
[Epoch 47 Batch 120/162] avg loss 0.00351787, throughput 9.53198K wps
[Epoch 47 Batch 150/162] avg loss 0.00354322, throughput 9.50896K wps
Begin Testing...
[Epoch 47] train avg loss 0.0034673, dev acc 0.9178, dev avg loss 0.22496, throughput 9.48018K wps
[Epoch 48 Batch 30/162] avg loss 0.00339867, throughput 9.58202K wps
[Epoch 48 Batch 60/162] avg loss 0.00356258, throughput 9.48541K wps
[Epoch 48 Batch 90/162] avg loss 0.00366516, throughput 9.54232K wps
[Epoch 48 Batch 120/162] avg loss 0.00348893, throughput 9.47152K wps
[Epoch 48 Batch 150/162] avg loss 0.00322505, throughput 9.32159K wps
Begin Testing...
[Epoch 48] train avg loss 0.00342565, dev acc 0.9178, dev avg loss 0.224293, throughput 9.44939K wps
[Epoch 49 Batch 30/162] avg loss 0.00342041, throughput 9.46309K wps
[Epoch 49 Batch 60/162] avg loss 0.00356455, throughput 9.30137K wps
[Epoch 49 Batch 90/162] avg loss 0.00336422, throughput 9.20449K wps
[Epoch 49 Batch 120/162] avg loss 0.0031011, throughput 9.21031K wps
[Epoch 49 Batch 150/162] avg loss 0.00367254, throughput 9.30586K wps
Begin Testing...
[Epoch 49] train avg loss 0.00340466, dev acc 0.9178, dev avg loss 0.223225, throughput 9.30418K wps
[Epoch 50 Batch 30/162] avg loss 0.00347847, throughput 9.46388K wps
[Epoch 50 Batch 60/162] avg loss 0.00312123, throughput 9.36587K wps
[Epoch 50 Batch 90/162] avg loss 0.00341098, throughput 9.37377K wps
[Epoch 50 Batch 120/162] avg loss 0.00322874, throughput 9.34715K wps
[Epoch 50 Batch 150/162] avg loss 0.00349886, throughput 9.31039K wps
Begin Testing...
[Epoch 50] train avg loss 0.00334845, dev acc 0.9189, dev avg loss 0.222546, throughput 9.38239K wps
Observed Improvement.
Begin Testing...
[Epoch 51 Batch 30/162] avg loss 0.00304245, throughput 9.57039K wps
[Epoch 51 Batch 60/162] avg loss 0.00320162, throughput 9.40134K wps
[Epoch 51 Batch 90/162] avg loss 0.0035254, throughput 9.22304K wps
[Epoch 51 Batch 120/162] avg loss 0.00332416, throughput 9.21957K wps
[Epoch 51 Batch 150/162] avg loss 0.00290001, throughput 9.46228K wps
Begin Testing...
[Epoch 51] train avg loss 0.00320044, dev acc 0.9178, dev avg loss 0.221703, throughput 9.37985K wps
[Epoch 52 Batch 30/162] avg loss 0.00330129, throughput 9.53952K wps
[Epoch 52 Batch 60/162] avg loss 0.00319879, throughput 9.42616K wps
[Epoch 52 Batch 90/162] avg loss 0.00331638, throughput 9.37914K wps
[Epoch 52 Batch 120/162] avg loss 0.0031235, throughput 9.45869K wps
[Epoch 52 Batch 150/162] avg loss 0.00322151, throughput 9.35381K wps
Begin Testing...
[Epoch 52] train avg loss 0.00324313, dev acc 0.9178, dev avg loss 0.221029, throughput 9.41941K wps
[Epoch 53 Batch 30/162] avg loss 0.00312318, throughput 9.48197K wps
[Epoch 53 Batch 60/162] avg loss 0.00293857, throughput 9.30991K wps
[Epoch 53 Batch 90/162] avg loss 0.00348293, throughput 9.29317K wps
[Epoch 53 Batch 120/162] avg loss 0.00308333, throughput 9.43754K wps
[Epoch 53 Batch 150/162] avg loss 0.00317115, throughput 9.38241K wps
Begin Testing...
[Epoch 53] train avg loss 0.00313697, dev acc 0.9189, dev avg loss 0.219898, throughput 9.37372K wps
Observed Improvement.
Begin Testing...
[Epoch 54 Batch 30/162] avg loss 0.00306758, throughput 9.53444K wps
[Epoch 54 Batch 60/162] avg loss 0.0030859, throughput 9.1984K wps
[Epoch 54 Batch 90/162] avg loss 0.00317954, throughput 9.1559K wps
[Epoch 54 Batch 120/162] avg loss 0.00295329, throughput 9.38802K wps
[Epoch 54 Batch 150/162] avg loss 0.00323905, throughput 9.28328K wps
Begin Testing...
[Epoch 54] train avg loss 0.00309711, dev acc 0.9156, dev avg loss 0.219606, throughput 9.30942K wps
[Epoch 55 Batch 30/162] avg loss 0.00309912, throughput 9.55629K wps
[Epoch 55 Batch 60/162] avg loss 0.00328996, throughput 9.26022K wps
[Epoch 55 Batch 90/162] avg loss 0.00283497, throughput 9.3437K wps
[Epoch 55 Batch 120/162] avg loss 0.00309373, throughput 9.33788K wps
[Epoch 55 Batch 150/162] avg loss 0.00322807, throughput 9.32481K wps
Begin Testing...
[Epoch 55] train avg loss 0.00314959, dev acc 0.9189, dev avg loss 0.218877, throughput 9.35597K wps
Observed Improvement.
Begin Testing...
[Epoch 56 Batch 30/162] avg loss 0.00299748, throughput 9.57061K wps
[Epoch 56 Batch 60/162] avg loss 0.00324153, throughput 9.3992K wps
[Epoch 56 Batch 90/162] avg loss 0.0029744, throughput 9.34628K wps
[Epoch 56 Batch 120/162] avg loss 0.00312703, throughput 9.26112K wps
[Epoch 56 Batch 150/162] avg loss 0.0027195, throughput 9.38876K wps
Begin Testing...
[Epoch 56] train avg loss 0.0030296, dev acc 0.9189, dev avg loss 0.218157, throughput 9.39806K wps
Observed Improvement.
Begin Testing...
[Epoch 57 Batch 30/162] avg loss 0.00297457, throughput 9.53425K wps
[Epoch 57 Batch 60/162] avg loss 0.00302075, throughput 9.39484K wps
[Epoch 57 Batch 90/162] avg loss 0.00308144, throughput 9.471K wps
[Epoch 57 Batch 120/162] avg loss 0.00301216, throughput 9.33223K wps
[Epoch 57 Batch 150/162] avg loss 0.00285948, throughput 9.5359K wps
Begin Testing...
[Epoch 57] train avg loss 0.00298452, dev acc 0.9211, dev avg loss 0.216941, throughput 9.44929K wps
Observed Improvement.
Begin Testing...
[Epoch 58 Batch 30/162] avg loss 0.0030519, throughput 9.42538K wps
[Epoch 58 Batch 60/162] avg loss 0.00304264, throughput 9.36764K wps
[Epoch 58 Batch 90/162] avg loss 0.00318671, throughput 9.45294K wps
[Epoch 58 Batch 120/162] avg loss 0.00294361, throughput 9.41266K wps
[Epoch 58 Batch 150/162] avg loss 0.00281069, throughput 9.29921K wps
Begin Testing...
[Epoch 58] train avg loss 0.00299149, dev acc 0.9200, dev avg loss 0.21641, throughput 9.37769K wps
[Epoch 59 Batch 30/162] avg loss 0.00309003, throughput 9.76164K wps
[Epoch 59 Batch 60/162] avg loss 0.00288478, throughput 9.43423K wps
[Epoch 59 Batch 90/162] avg loss 0.0030056, throughput 9.36587K wps
[Epoch 59 Batch 120/162] avg loss 0.00284529, throughput 9.49486K wps
[Epoch 59 Batch 150/162] avg loss 0.00296807, throughput 9.34959K wps
Begin Testing...
[Epoch 59] train avg loss 0.00294704, dev acc 0.9189, dev avg loss 0.21637, throughput 9.48271K wps
[Epoch 60 Batch 30/162] avg loss 0.00309789, throughput 9.58397K wps
[Epoch 60 Batch 60/162] avg loss 0.0028392, throughput 9.416K wps
[Epoch 60 Batch 90/162] avg loss 0.00275479, throughput 9.4663K wps
[Epoch 60 Batch 120/162] avg loss 0.00266363, throughput 9.38289K wps
[Epoch 60 Batch 150/162] avg loss 0.00285094, throughput 9.39935K wps
Begin Testing...
[Epoch 60] train avg loss 0.00286126, dev acc 0.9189, dev avg loss 0.215594, throughput 9.43455K wps
[Epoch 61 Batch 30/162] avg loss 0.00253019, throughput 9.76977K wps
[Epoch 61 Batch 60/162] avg loss 0.00296642, throughput 9.38459K wps
[Epoch 61 Batch 90/162] avg loss 0.00277417, throughput 9.35544K wps
[Epoch 61 Batch 120/162] avg loss 0.00305958, throughput 9.33514K wps
[Epoch 61 Batch 150/162] avg loss 0.00266039, throughput 9.2967K wps
Begin Testing...
[Epoch 61] train avg loss 0.00282061, dev acc 0.9211, dev avg loss 0.214901, throughput 9.42117K wps
Observed Improvement.
Begin Testing...
[Epoch 62 Batch 30/162] avg loss 0.00269149, throughput 9.6304K wps
[Epoch 62 Batch 60/162] avg loss 0.00275886, throughput 9.28424K wps
[Epoch 62 Batch 90/162] avg loss 0.00278965, throughput 9.46314K wps
[Epoch 62 Batch 120/162] avg loss 0.00278281, throughput 9.41509K wps
[Epoch 62 Batch 150/162] avg loss 0.00296652, throughput 9.46217K wps
Begin Testing...
[Epoch 62] train avg loss 0.00280655, dev acc 0.9200, dev avg loss 0.214234, throughput 9.45876K wps
[Epoch 63 Batch 30/162] avg loss 0.0026918, throughput 9.60831K wps
[Epoch 63 Batch 60/162] avg loss 0.00272578, throughput 9.43545K wps
[Epoch 63 Batch 90/162] avg loss 0.00265863, throughput 9.32025K wps
[Epoch 63 Batch 120/162] avg loss 0.00290196, throughput 9.13743K wps
[Epoch 63 Batch 150/162] avg loss 0.00266342, throughput 9.28168K wps
Begin Testing...
[Epoch 63] train avg loss 0.0027211, dev acc 0.9211, dev avg loss 0.214618, throughput 9.34955K wps
Observed Improvement.
Begin Testing...
[Epoch 64 Batch 30/162] avg loss 0.0026126, throughput 9.56088K wps
[Epoch 64 Batch 60/162] avg loss 0.00261168, throughput 9.39434K wps
[Epoch 64 Batch 90/162] avg loss 0.00284288, throughput 9.35492K wps
[Epoch 64 Batch 120/162] avg loss 0.00285913, throughput 9.30435K wps
[Epoch 64 Batch 150/162] avg loss 0.00259517, throughput 9.2835K wps
Begin Testing...
[Epoch 64] train avg loss 0.00270896, dev acc 0.9222, dev avg loss 0.212864, throughput 9.38041K wps
Observed Improvement.
Begin Testing...
[Epoch 65 Batch 30/162] avg loss 0.00266184, throughput 9.457K wps
[Epoch 65 Batch 60/162] avg loss 0.00256257, throughput 9.45132K wps
[Epoch 65 Batch 90/162] avg loss 0.00249786, throughput 9.32666K wps
[Epoch 65 Batch 120/162] avg loss 0.00299726, throughput 9.40122K wps
[Epoch 65 Batch 150/162] avg loss 0.00266928, throughput 9.27458K wps
Begin Testing...
[Epoch 65] train avg loss 0.00266813, dev acc 0.9222, dev avg loss 0.212283, throughput 9.37011K wps
Observed Improvement.
Begin Testing...
[Epoch 66 Batch 30/162] avg loss 0.00250264, throughput 9.52543K wps
[Epoch 66 Batch 60/162] avg loss 0.00241669, throughput 9.27055K wps
[Epoch 66 Batch 90/162] avg loss 0.00282027, throughput 9.42991K wps
[Epoch 66 Batch 120/162] avg loss 0.00269262, throughput 9.23692K wps
[Epoch 66 Batch 150/162] avg loss 0.00266666, throughput 9.27168K wps
Begin Testing...
[Epoch 66] train avg loss 0.00259419, dev acc 0.9222, dev avg loss 0.211483, throughput 9.33889K wps
Observed Improvement.
Begin Testing...
[Epoch 67 Batch 30/162] avg loss 0.00249504, throughput 9.47676K wps
[Epoch 67 Batch 60/162] avg loss 0.00250064, throughput 9.29991K wps
[Epoch 67 Batch 90/162] avg loss 0.00257394, throughput 9.36837K wps
[Epoch 67 Batch 120/162] avg loss 0.00274929, throughput 9.35763K wps
[Epoch 67 Batch 150/162] avg loss 0.00266347, throughput 9.44608K wps
Begin Testing...
[Epoch 67] train avg loss 0.00259588, dev acc 0.9200, dev avg loss 0.210808, throughput 9.37524K wps
[Epoch 68 Batch 30/162] avg loss 0.0026491, throughput 9.55714K wps
[Epoch 68 Batch 60/162] avg loss 0.00293124, throughput 9.24097K wps
[Epoch 68 Batch 90/162] avg loss 0.0023547, throughput 9.39898K wps
[Epoch 68 Batch 120/162] avg loss 0.00261437, throughput 9.43415K wps
[Epoch 68 Batch 150/162] avg loss 0.0023147, throughput 9.37737K wps
Begin Testing...
[Epoch 68] train avg loss 0.00255004, dev acc 0.9200, dev avg loss 0.210856, throughput 9.4112K wps
[Epoch 69 Batch 30/162] avg loss 0.00236264, throughput 9.76615K wps
[Epoch 69 Batch 60/162] avg loss 0.00254204, throughput 9.53889K wps
[Epoch 69 Batch 90/162] avg loss 0.00256414, throughput 9.38684K wps
[Epoch 69 Batch 120/162] avg loss 0.00262695, throughput 9.44273K wps
[Epoch 69 Batch 150/162] avg loss 0.00251046, throughput 9.40966K wps
Begin Testing...
[Epoch 69] train avg loss 0.00253583, dev acc 0.9200, dev avg loss 0.210559, throughput 9.50258K wps
[Epoch 70 Batch 30/162] avg loss 0.0024204, throughput 9.53716K wps
[Epoch 70 Batch 60/162] avg loss 0.00244967, throughput 9.4404K wps
[Epoch 70 Batch 90/162] avg loss 0.00240488, throughput 9.36018K wps
[Epoch 70 Batch 120/162] avg loss 0.00263349, throughput 9.45604K wps
[Epoch 70 Batch 150/162] avg loss 0.00234914, throughput 9.35992K wps
Begin Testing...
[Epoch 70] train avg loss 0.00245254, dev acc 0.9200, dev avg loss 0.209931, throughput 9.43631K wps
[Epoch 71 Batch 30/162] avg loss 0.00240789, throughput 9.68664K wps
[Epoch 71 Batch 60/162] avg loss 0.00248945, throughput 9.3224K wps
[Epoch 71 Batch 90/162] avg loss 0.00256101, throughput 9.40859K wps
[Epoch 71 Batch 120/162] avg loss 0.00226569, throughput 9.23253K wps
[Epoch 71 Batch 150/162] avg loss 0.00262153, throughput 9.38478K wps
Begin Testing...
[Epoch 71] train avg loss 0.00248414, dev acc 0.9211, dev avg loss 0.209473, throughput 9.38691K wps
[Epoch 72 Batch 30/162] avg loss 0.00231295, throughput 9.40852K wps
[Epoch 72 Batch 60/162] avg loss 0.00227905, throughput 9.18122K wps
[Epoch 72 Batch 90/162] avg loss 0.00249192, throughput 9.33384K wps
[Epoch 72 Batch 120/162] avg loss 0.00220115, throughput 9.49662K wps
[Epoch 72 Batch 150/162] avg loss 0.00257822, throughput 9.53931K wps
Begin Testing...
[Epoch 72] train avg loss 0.00237862, dev acc 0.9222, dev avg loss 0.208919, throughput 9.37349K wps
Observed Improvement.
Begin Testing...
[Epoch 73 Batch 30/162] avg loss 0.00248132, throughput 9.66719K wps
[Epoch 73 Batch 60/162] avg loss 0.00223944, throughput 9.3788K wps
[Epoch 73 Batch 90/162] avg loss 0.00237729, throughput 9.33332K wps
[Epoch 73 Batch 120/162] avg loss 0.00263944, throughput 9.51485K wps
[Epoch 73 Batch 150/162] avg loss 0.00230547, throughput 9.52591K wps
Begin Testing...
[Epoch 73] train avg loss 0.00239745, dev acc 0.9222, dev avg loss 0.208237, throughput 9.47115K wps
Observed Improvement.
Begin Testing...
[Epoch 74 Batch 30/162] avg loss 0.00233452, throughput 9.72133K wps
[Epoch 74 Batch 60/162] avg loss 0.00235288, throughput 9.42341K wps
[Epoch 74 Batch 90/162] avg loss 0.00223811, throughput 9.50422K wps
[Epoch 74 Batch 120/162] avg loss 0.00246239, throughput 9.34197K wps
[Epoch 74 Batch 150/162] avg loss 0.00239308, throughput 9.39556K wps
Begin Testing...
[Epoch 74] train avg loss 0.00235093, dev acc 0.9211, dev avg loss 0.207968, throughput 9.46835K wps
[Epoch 75 Batch 30/162] avg loss 0.00257944, throughput 9.6299K wps
[Epoch 75 Batch 60/162] avg loss 0.00218365, throughput 9.38988K wps
[Epoch 75 Batch 90/162] avg loss 0.0021832, throughput 9.56108K wps
[Epoch 75 Batch 120/162] avg loss 0.00231687, throughput 9.35633K wps
[Epoch 75 Batch 150/162] avg loss 0.00227732, throughput 9.26077K wps
Begin Testing...
[Epoch 75] train avg loss 0.00234944, dev acc 0.9211, dev avg loss 0.207756, throughput 9.44125K wps
[Epoch 76 Batch 30/162] avg loss 0.00246977, throughput 9.57019K wps
[Epoch 76 Batch 60/162] avg loss 0.00233069, throughput 9.47189K wps
[Epoch 76 Batch 90/162] avg loss 0.00205494, throughput 9.47045K wps
[Epoch 76 Batch 120/162] avg loss 0.00233756, throughput 9.3395K wps
[Epoch 76 Batch 150/162] avg loss 0.00245006, throughput 9.38939K wps
Begin Testing...
[Epoch 76] train avg loss 0.00231847, dev acc 0.9233, dev avg loss 0.208611, throughput 9.43029K wps
Observed Improvement.
Begin Testing...
[Epoch 77 Batch 30/162] avg loss 0.0024733, throughput 9.6933K wps
[Epoch 77 Batch 60/162] avg loss 0.00224544, throughput 9.36658K wps
[Epoch 77 Batch 90/162] avg loss 0.00225306, throughput 9.3526K wps
[Epoch 77 Batch 120/162] avg loss 0.00239805, throughput 9.40154K wps
[Epoch 77 Batch 150/162] avg loss 0.00195458, throughput 9.40228K wps
Begin Testing...
[Epoch 77] train avg loss 0.00228019, dev acc 0.9233, dev avg loss 0.206629, throughput 9.44618K wps
Observed Improvement.
Begin Testing...
[Epoch 78 Batch 30/162] avg loss 0.00207113, throughput 9.69232K wps
[Epoch 78 Batch 60/162] avg loss 0.00228975, throughput 9.23539K wps
[Epoch 78 Batch 90/162] avg loss 0.00230338, throughput 9.42575K wps
[Epoch 78 Batch 120/162] avg loss 0.00218398, throughput 9.37048K wps
[Epoch 78 Batch 150/162] avg loss 0.00197783, throughput 9.32321K wps
Begin Testing...
[Epoch 78] train avg loss 0.00217135, dev acc 0.9211, dev avg loss 0.206674, throughput 9.40951K wps
[Epoch 79 Batch 30/162] avg loss 0.00235005, throughput 9.47869K wps
[Epoch 79 Batch 60/162] avg loss 0.00216836, throughput 9.3978K wps
[Epoch 79 Batch 90/162] avg loss 0.00214022, throughput 9.37198K wps
[Epoch 79 Batch 120/162] avg loss 0.00223165, throughput 9.33747K wps
[Epoch 79 Batch 150/162] avg loss 0.00198613, throughput 9.41899K wps
Begin Testing...
[Epoch 79] train avg loss 0.00215863, dev acc 0.9222, dev avg loss 0.206344, throughput 9.39235K wps
[Epoch 80 Batch 30/162] avg loss 0.00222975, throughput 9.6298K wps
[Epoch 80 Batch 60/162] avg loss 0.00184238, throughput 9.28435K wps
[Epoch 80 Batch 90/162] avg loss 0.00252918, throughput 9.25054K wps
[Epoch 80 Batch 120/162] avg loss 0.00220173, throughput 9.49821K wps
[Epoch 80 Batch 150/162] avg loss 0.00183768, throughput 9.28979K wps
Begin Testing...
[Epoch 80] train avg loss 0.00212165, dev acc 0.9244, dev avg loss 0.206298, throughput 9.40949K wps
Observed Improvement.
Begin Testing...
[Epoch 81 Batch 30/162] avg loss 0.00230204, throughput 9.49307K wps
[Epoch 81 Batch 60/162] avg loss 0.00216045, throughput 9.31312K wps
[Epoch 81 Batch 90/162] avg loss 0.00231145, throughput 9.29803K wps
[Epoch 81 Batch 120/162] avg loss 0.00209262, throughput 9.33757K wps
[Epoch 81 Batch 150/162] avg loss 0.0019635, throughput 9.31314K wps
Begin Testing...
[Epoch 81] train avg loss 0.0021713, dev acc 0.9233, dev avg loss 0.205901, throughput 9.34852K wps
[Epoch 82 Batch 30/162] avg loss 0.00190619, throughput 9.55957K wps
[Epoch 82 Batch 60/162] avg loss 0.00234425, throughput 9.35162K wps
[Epoch 82 Batch 90/162] avg loss 0.00232376, throughput 9.22046K wps
[Epoch 82 Batch 120/162] avg loss 0.00202978, throughput 9.54604K wps
[Epoch 82 Batch 150/162] avg loss 0.00209262, throughput 9.52053K wps
Begin Testing...
[Epoch 82] train avg loss 0.00211971, dev acc 0.9222, dev avg loss 0.205093, throughput 9.42898K wps
[Epoch 83 Batch 30/162] avg loss 0.00217346, throughput 9.32124K wps
[Epoch 83 Batch 60/162] avg loss 0.00185793, throughput 9.43477K wps
[Epoch 83 Batch 90/162] avg loss 0.00228981, throughput 9.39923K wps
[Epoch 83 Batch 120/162] avg loss 0.00215277, throughput 9.38176K wps
[Epoch 83 Batch 150/162] avg loss 0.0022617, throughput 9.43604K wps
Begin Testing...
[Epoch 83] train avg loss 0.00215133, dev acc 0.9244, dev avg loss 0.205591, throughput 9.37267K wps
Observed Improvement.
Begin Testing...
[Epoch 84 Batch 30/162] avg loss 0.00187361, throughput 9.58512K wps
[Epoch 84 Batch 60/162] avg loss 0.00205785, throughput 9.26876K wps
[Epoch 84 Batch 90/162] avg loss 0.00206367, throughput 9.509K wps
[Epoch 84 Batch 120/162] avg loss 0.00219402, throughput 9.33053K wps
[Epoch 84 Batch 150/162] avg loss 0.00198974, throughput 9.37345K wps
Begin Testing...
[Epoch 84] train avg loss 0.00203205, dev acc 0.9222, dev avg loss 0.204304, throughput 9.39575K wps
[Epoch 85 Batch 30/162] avg loss 0.00208433, throughput 9.5062K wps
[Epoch 85 Batch 60/162] avg loss 0.00177657, throughput 9.31559K wps
[Epoch 85 Batch 90/162] avg loss 0.00198931, throughput 9.29835K wps
[Epoch 85 Batch 120/162] avg loss 0.00200297, throughput 9.39833K wps
[Epoch 85 Batch 150/162] avg loss 0.00227748, throughput 9.29818K wps
Begin Testing...
[Epoch 85] train avg loss 0.00202292, dev acc 0.9256, dev avg loss 0.203931, throughput 9.3396K wps
Observed Improvement.
Begin Testing...
[Epoch 86 Batch 30/162] avg loss 0.00196756, throughput 9.52375K wps
[Epoch 86 Batch 60/162] avg loss 0.0019974, throughput 9.23258K wps
[Epoch 86 Batch 90/162] avg loss 0.00204913, throughput 9.34431K wps
[Epoch 86 Batch 120/162] avg loss 0.00179861, throughput 9.4073K wps
[Epoch 86 Batch 150/162] avg loss 0.0018706, throughput 9.37705K wps
Begin Testing...
[Epoch 86] train avg loss 0.0019604, dev acc 0.9244, dev avg loss 0.204496, throughput 9.37745K wps
[Epoch 87 Batch 30/162] avg loss 0.00190248, throughput 9.5779K wps
[Epoch 87 Batch 60/162] avg loss 0.00211798, throughput 9.29668K wps
[Epoch 87 Batch 90/162] avg loss 0.00191901, throughput 9.24771K wps
[Epoch 87 Batch 120/162] avg loss 0.00195456, throughput 9.44121K wps
[Epoch 87 Batch 150/162] avg loss 0.00204954, throughput 9.27785K wps
Begin Testing...
[Epoch 87] train avg loss 0.00199495, dev acc 0.9256, dev avg loss 0.203738, throughput 9.37875K wps
Observed Improvement.
Begin Testing...
[Epoch 88 Batch 30/162] avg loss 0.0019946, throughput 9.41906K wps
[Epoch 88 Batch 60/162] avg loss 0.00193633, throughput 9.30126K wps
[Epoch 88 Batch 90/162] avg loss 0.00185998, throughput 9.5232K wps
[Epoch 88 Batch 120/162] avg loss 0.00195051, throughput 9.43933K wps
[Epoch 88 Batch 150/162] avg loss 0.00214931, throughput 9.31672K wps
Begin Testing...
[Epoch 88] train avg loss 0.00195381, dev acc 0.9256, dev avg loss 0.203697, throughput 9.39728K wps
Observed Improvement.
Begin Testing...
[Epoch 89 Batch 30/162] avg loss 0.00192866, throughput 9.52825K wps
[Epoch 89 Batch 60/162] avg loss 0.00197791, throughput 9.26104K wps
[Epoch 89 Batch 90/162] avg loss 0.00193645, throughput 9.40345K wps
[Epoch 89 Batch 120/162] avg loss 0.00217938, throughput 9.37801K wps
[Epoch 89 Batch 150/162] avg loss 0.00199219, throughput 9.28022K wps
Begin Testing...
[Epoch 89] train avg loss 0.00200452, dev acc 0.9244, dev avg loss 0.204087, throughput 9.38397K wps
[Epoch 90 Batch 30/162] avg loss 0.00185857, throughput 9.51826K wps
[Epoch 90 Batch 60/162] avg loss 0.00202954, throughput 9.3183K wps
[Epoch 90 Batch 90/162] avg loss 0.00203731, throughput 9.44054K wps
[Epoch 90 Batch 120/162] avg loss 0.00187401, throughput 9.31878K wps
[Epoch 90 Batch 150/162] avg loss 0.00185091, throughput 9.23475K wps
Begin Testing...
[Epoch 90] train avg loss 0.00190673, dev acc 0.9256, dev avg loss 0.202735, throughput 9.38357K wps
Observed Improvement.
Begin Testing...
[Epoch 91 Batch 30/162] avg loss 0.00167195, throughput 9.48947K wps
[Epoch 91 Batch 60/162] avg loss 0.00180569, throughput 9.41164K wps
[Epoch 91 Batch 90/162] avg loss 0.00181314, throughput 9.38445K wps
[Epoch 91 Batch 120/162] avg loss 0.00202195, throughput 9.43237K wps
[Epoch 91 Batch 150/162] avg loss 0.00214542, throughput 9.30149K wps
Begin Testing...
[Epoch 91] train avg loss 0.00186402, dev acc 0.9233, dev avg loss 0.202484, throughput 9.41859K wps
[Epoch 92 Batch 30/162] avg loss 0.00202393, throughput 9.57583K wps
[Epoch 92 Batch 60/162] avg loss 0.00166262, throughput 9.4278K wps
[Epoch 92 Batch 90/162] avg loss 0.0018026, throughput 9.32236K wps
[Epoch 92 Batch 120/162] avg loss 0.0016359, throughput 9.46342K wps
[Epoch 92 Batch 150/162] avg loss 0.00196304, throughput 9.4185K wps
Begin Testing...
[Epoch 92] train avg loss 0.00184312, dev acc 0.9256, dev avg loss 0.202851, throughput 9.4406K wps
Observed Improvement.
Begin Testing...
[Epoch 93 Batch 30/162] avg loss 0.00192257, throughput 9.51555K wps
[Epoch 93 Batch 60/162] avg loss 0.00183915, throughput 9.23505K wps
[Epoch 93 Batch 90/162] avg loss 0.00188178, throughput 9.42059K wps
[Epoch 93 Batch 120/162] avg loss 0.00172586, throughput 9.35112K wps
[Epoch 93 Batch 150/162] avg loss 0.00162763, throughput 9.27694K wps
Begin Testing...
[Epoch 93] train avg loss 0.0018171, dev acc 0.9244, dev avg loss 0.202735, throughput 9.35486K wps
[Epoch 94 Batch 30/162] avg loss 0.00163137, throughput 9.48787K wps
[Epoch 94 Batch 60/162] avg loss 0.00172054, throughput 9.52759K wps
[Epoch 94 Batch 90/162] avg loss 0.00191297, throughput 9.29313K wps
[Epoch 94 Batch 120/162] avg loss 0.0017799, throughput 9.39914K wps
[Epoch 94 Batch 150/162] avg loss 0.00186745, throughput 9.41909K wps
Begin Testing...
[Epoch 94] train avg loss 0.00177816, dev acc 0.9267, dev avg loss 0.201452, throughput 9.4095K wps
Observed Improvement.
Begin Testing...
[Epoch 95 Batch 30/162] avg loss 0.00165395, throughput 9.54676K wps
[Epoch 95 Batch 60/162] avg loss 0.00188043, throughput 9.23283K wps
[Epoch 95 Batch 90/162] avg loss 0.00188678, throughput 9.33186K wps
[Epoch 95 Batch 120/162] avg loss 0.00183553, throughput 9.30612K wps
[Epoch 95 Batch 150/162] avg loss 0.00167964, throughput 9.36738K wps
Begin Testing...
[Epoch 95] train avg loss 0.00177887, dev acc 0.9278, dev avg loss 0.201506, throughput 9.35049K wps
Observed Improvement.
Begin Testing...
[Epoch 96 Batch 30/162] avg loss 0.0018067, throughput 9.36084K wps
[Epoch 96 Batch 60/162] avg loss 0.00150185, throughput 9.45998K wps
[Epoch 96 Batch 90/162] avg loss 0.00174129, throughput 9.40955K wps
[Epoch 96 Batch 120/162] avg loss 0.00168815, throughput 9.43815K wps
[Epoch 96 Batch 150/162] avg loss 0.00183747, throughput 9.38871K wps
Begin Testing...
[Epoch 96] train avg loss 0.00172719, dev acc 0.9256, dev avg loss 0.202846, throughput 9.4169K wps
[Epoch 97 Batch 30/162] avg loss 0.0016267, throughput 9.4709K wps
[Epoch 97 Batch 60/162] avg loss 0.00164389, throughput 9.51736K wps
[Epoch 97 Batch 90/162] avg loss 0.00153075, throughput 9.30254K wps
[Epoch 97 Batch 120/162] avg loss 0.00184117, throughput 9.53896K wps
[Epoch 97 Batch 150/162] avg loss 0.00179041, throughput 9.29699K wps
Begin Testing...
[Epoch 97] train avg loss 0.00169207, dev acc 0.9256, dev avg loss 0.201731, throughput 9.40968K wps
[Epoch 98 Batch 30/162] avg loss 0.00161255, throughput 9.55141K wps
[Epoch 98 Batch 60/162] avg loss 0.0017042, throughput 9.25578K wps
[Epoch 98 Batch 90/162] avg loss 0.00162267, throughput 9.31859K wps
[Epoch 98 Batch 120/162] avg loss 0.00183334, throughput 9.42684K wps
[Epoch 98 Batch 150/162] avg loss 0.0018736, throughput 9.26711K wps
Begin Testing...
[Epoch 98] train avg loss 0.00172534, dev acc 0.9278, dev avg loss 0.20133, throughput 9.35885K wps
Observed Improvement.
Begin Testing...
[Epoch 99 Batch 30/162] avg loss 0.00178203, throughput 9.69662K wps
[Epoch 99 Batch 60/162] avg loss 0.00168355, throughput 9.25415K wps
[Epoch 99 Batch 90/162] avg loss 0.00180107, throughput 9.45727K wps
[Epoch 99 Batch 120/162] avg loss 0.00156636, throughput 9.33142K wps
[Epoch 99 Batch 150/162] avg loss 0.00161922, throughput 9.42126K wps
Begin Testing...
[Epoch 99] train avg loss 0.00169234, dev acc 0.9278, dev avg loss 0.201103, throughput 9.43427K wps
Observed Improvement.
Begin Testing...
[Epoch 100 Batch 30/162] avg loss 0.0016259, throughput 9.72245K wps
[Epoch 100 Batch 60/162] avg loss 0.00157466, throughput 9.47123K wps
[Epoch 100 Batch 90/162] avg loss 0.00155395, throughput 9.48682K wps
[Epoch 100 Batch 120/162] avg loss 0.00169453, throughput 9.49779K wps
[Epoch 100 Batch 150/162] avg loss 0.00187769, throughput 9.32434K wps
Begin Testing...
[Epoch 100] train avg loss 0.00167188, dev acc 0.9278, dev avg loss 0.201584, throughput 9.50583K wps
Observed Improvement.
Begin Testing...
[Epoch 101 Batch 30/162] avg loss 0.00150735, throughput 9.55371K wps
[Epoch 101 Batch 60/162] avg loss 0.00189267, throughput 9.30413K wps
[Epoch 101 Batch 90/162] avg loss 0.00151687, throughput 9.31869K wps
[Epoch 101 Batch 120/162] avg loss 0.00165239, throughput 9.52223K wps
[Epoch 101 Batch 150/162] avg loss 0.0016823, throughput 9.32521K wps
Begin Testing...
[Epoch 101] train avg loss 0.00164527, dev acc 0.9244, dev avg loss 0.201661, throughput 9.39745K wps
[Epoch 102 Batch 30/162] avg loss 0.00150766, throughput 9.43727K wps
[Epoch 102 Batch 60/162] avg loss 0.00164845, throughput 9.2507K wps
[Epoch 102 Batch 90/162] avg loss 0.00175769, throughput 9.35673K wps
[Epoch 102 Batch 120/162] avg loss 0.00178624, throughput 9.4864K wps
[Epoch 102 Batch 150/162] avg loss 0.00154615, throughput 9.39803K wps
Begin Testing...
[Epoch 102] train avg loss 0.00164297, dev acc 0.9256, dev avg loss 0.201279, throughput 9.4006K wps
[Epoch 103 Batch 30/162] avg loss 0.00161365, throughput 9.58287K wps
[Epoch 103 Batch 60/162] avg loss 0.00161417, throughput 9.36368K wps
[Epoch 103 Batch 90/162] avg loss 0.00167703, throughput 9.27214K wps
[Epoch 103 Batch 120/162] avg loss 0.00152906, throughput 9.445K wps
[Epoch 103 Batch 150/162] avg loss 0.00155203, throughput 9.4719K wps
Begin Testing...
[Epoch 103] train avg loss 0.00158686, dev acc 0.9267, dev avg loss 0.20078, throughput 9.41522K wps
[Epoch 104 Batch 30/162] avg loss 0.0015189, throughput 9.55253K wps
[Epoch 104 Batch 60/162] avg loss 0.00157563, throughput 9.37204K wps
[Epoch 104 Batch 90/162] avg loss 0.00151875, throughput 9.47342K wps
[Epoch 104 Batch 120/162] avg loss 0.00146111, throughput 9.28916K wps
[Epoch 104 Batch 150/162] avg loss 0.00167138, throughput 9.40638K wps
Begin Testing...
[Epoch 104] train avg loss 0.00155683, dev acc 0.9267, dev avg loss 0.200666, throughput 9.42508K wps
[Epoch 105 Batch 30/162] avg loss 0.00149955, throughput 9.63626K wps
[Epoch 105 Batch 60/162] avg loss 0.00157579, throughput 9.2207K wps
[Epoch 105 Batch 90/162] avg loss 0.00165946, throughput 9.4933K wps
[Epoch 105 Batch 120/162] avg loss 0.00146336, throughput 9.41137K wps
[Epoch 105 Batch 150/162] avg loss 0.00153732, throughput 9.30036K wps
Begin Testing...
[Epoch 105] train avg loss 0.00155901, dev acc 0.9289, dev avg loss 0.199678, throughput 9.40358K wps
Observed Improvement.
Begin Testing...
[Epoch 106 Batch 30/162] avg loss 0.00142332, throughput 9.66661K wps
[Epoch 106 Batch 60/162] avg loss 0.00158049, throughput 9.40541K wps
[Epoch 106 Batch 90/162] avg loss 0.00157714, throughput 9.41834K wps
[Epoch 106 Batch 120/162] avg loss 0.00141014, throughput 9.35045K wps
[Epoch 106 Batch 150/162] avg loss 0.00164792, throughput 9.43854K wps
Begin Testing...
[Epoch 106] train avg loss 0.00152312, dev acc 0.9278, dev avg loss 0.200344, throughput 9.4436K wps
[Epoch 107 Batch 30/162] avg loss 0.00140457, throughput 9.57794K wps
[Epoch 107 Batch 60/162] avg loss 0.00154623, throughput 9.51554K wps
[Epoch 107 Batch 90/162] avg loss 0.00144479, throughput 9.43262K wps
[Epoch 107 Batch 120/162] avg loss 0.00160065, throughput 9.45721K wps
[Epoch 107 Batch 150/162] avg loss 0.0015154, throughput 9.43334K wps
Begin Testing...
[Epoch 107] train avg loss 0.00151217, dev acc 0.9244, dev avg loss 0.201207, throughput 9.46985K wps
[Epoch 108 Batch 30/162] avg loss 0.00157157, throughput 9.5936K wps
[Epoch 108 Batch 60/162] avg loss 0.00155498, throughput 9.36322K wps
[Epoch 108 Batch 90/162] avg loss 0.00149941, throughput 9.37888K wps
[Epoch 108 Batch 120/162] avg loss 0.00141142, throughput 9.22913K wps
[Epoch 108 Batch 150/162] avg loss 0.00173523, throughput 9.35308K wps
Begin Testing...
[Epoch 108] train avg loss 0.00154691, dev acc 0.9278, dev avg loss 0.200023, throughput 9.38223K wps
[Epoch 109 Batch 30/162] avg loss 0.0013554, throughput 9.6K wps
[Epoch 109 Batch 60/162] avg loss 0.00149964, throughput 9.3052K wps
[Epoch 109 Batch 90/162] avg loss 0.00153402, throughput 9.29517K wps
[Epoch 109 Batch 120/162] avg loss 0.001613, throughput 9.28471K wps
[Epoch 109 Batch 150/162] avg loss 0.00141026, throughput 9.37518K wps
Begin Testing...
[Epoch 109] train avg loss 0.00150107, dev acc 0.9278, dev avg loss 0.199981, throughput 9.3516K wps
[Epoch 110 Batch 30/162] avg loss 0.00147237, throughput 9.49169K wps
[Epoch 110 Batch 60/162] avg loss 0.00133405, throughput 9.4005K wps
[Epoch 110 Batch 90/162] avg loss 0.00153076, throughput 9.60073K wps
[Epoch 110 Batch 120/162] avg loss 0.00142159, throughput 9.35705K wps
[Epoch 110 Batch 150/162] avg loss 0.00150565, throughput 9.50525K wps
Begin Testing...
[Epoch 110] train avg loss 0.00147086, dev acc 0.9278, dev avg loss 0.199383, throughput 9.4574K wps
[Epoch 111 Batch 30/162] avg loss 0.00139733, throughput 9.35636K wps
[Epoch 111 Batch 60/162] avg loss 0.00153088, throughput 9.3379K wps
[Epoch 111 Batch 90/162] avg loss 0.00133508, throughput 9.30945K wps
[Epoch 111 Batch 120/162] avg loss 0.00143346, throughput 9.27795K wps
[Epoch 111 Batch 150/162] avg loss 0.00156747, throughput 9.28131K wps
Begin Testing...
[Epoch 111] train avg loss 0.00145299, dev acc 0.9278, dev avg loss 0.199196, throughput 9.32067K wps
[Epoch 112 Batch 30/162] avg loss 0.00141012, throughput 9.49453K wps
[Epoch 112 Batch 60/162] avg loss 0.00164314, throughput 9.31206K wps
[Epoch 112 Batch 90/162] avg loss 0.00135187, throughput 9.31677K wps
[Epoch 112 Batch 120/162] avg loss 0.00145206, throughput 9.49807K wps
[Epoch 112 Batch 150/162] avg loss 0.00138705, throughput 9.59828K wps
Begin Testing...
[Epoch 112] train avg loss 0.0014432, dev acc 0.9311, dev avg loss 0.199142, throughput 9.44057K wps
Observed Improvement.
Begin Testing...
[Epoch 113 Batch 30/162] avg loss 0.0013675, throughput 9.80452K wps
[Epoch 113 Batch 60/162] avg loss 0.00140924, throughput 9.24686K wps
[Epoch 113 Batch 90/162] avg loss 0.00136147, throughput 9.41749K wps
[Epoch 113 Batch 120/162] avg loss 0.00138529, throughput 9.19369K wps
[Epoch 113 Batch 150/162] avg loss 0.00147477, throughput 9.44531K wps
Begin Testing...
[Epoch 113] train avg loss 0.00138702, dev acc 0.9289, dev avg loss 0.199933, throughput 9.4024K wps
[Epoch 114 Batch 30/162] avg loss 0.00127599, throughput 9.65191K wps
[Epoch 114 Batch 60/162] avg loss 0.00142577, throughput 9.34895K wps
[Epoch 114 Batch 90/162] avg loss 0.00155588, throughput 9.31974K wps
[Epoch 114 Batch 120/162] avg loss 0.00124705, throughput 9.319K wps
[Epoch 114 Batch 150/162] avg loss 0.00156108, throughput 9.2978K wps
Begin Testing...
[Epoch 114] train avg loss 0.00140714, dev acc 0.9267, dev avg loss 0.199581, throughput 9.38194K wps
[Epoch 115 Batch 30/162] avg loss 0.0012881, throughput 9.43695K wps
[Epoch 115 Batch 60/162] avg loss 0.00130016, throughput 9.57039K wps
[Epoch 115 Batch 90/162] avg loss 0.00149962, throughput 9.3521K wps
[Epoch 115 Batch 120/162] avg loss 0.00131644, throughput 9.29963K wps
[Epoch 115 Batch 150/162] avg loss 0.00136541, throughput 9.43992K wps
Begin Testing...
[Epoch 115] train avg loss 0.00136657, dev acc 0.9267, dev avg loss 0.199445, throughput 9.40994K wps
[Epoch 116 Batch 30/162] avg loss 0.00144005, throughput 9.5065K wps
[Epoch 116 Batch 60/162] avg loss 0.00121961, throughput 9.38819K wps
[Epoch 116 Batch 90/162] avg loss 0.00146966, throughput 9.44021K wps
[Epoch 116 Batch 120/162] avg loss 0.00128273, throughput 9.38758K wps
[Epoch 116 Batch 150/162] avg loss 0.00126835, throughput 9.37983K wps
Begin Testing...
[Epoch 116] train avg loss 0.00134921, dev acc 0.9278, dev avg loss 0.19927, throughput 9.41364K wps
[Epoch 117 Batch 30/162] avg loss 0.00128174, throughput 9.70739K wps
[Epoch 117 Batch 60/162] avg loss 0.00136264, throughput 9.47089K wps
[Epoch 117 Batch 90/162] avg loss 0.00132684, throughput 9.28686K wps
[Epoch 117 Batch 120/162] avg loss 0.00131676, throughput 9.39006K wps
[Epoch 117 Batch 150/162] avg loss 0.00164203, throughput 9.41568K wps
Begin Testing...
[Epoch 117] train avg loss 0.00136938, dev acc 0.9256, dev avg loss 0.200608, throughput 9.46134K wps
[Epoch 118 Batch 30/162] avg loss 0.00146328, throughput 9.43066K wps
[Epoch 118 Batch 60/162] avg loss 0.00131696, throughput 9.39652K wps
[Epoch 118 Batch 90/162] avg loss 0.00116184, throughput 9.40948K wps
[Epoch 118 Batch 120/162] avg loss 0.00123258, throughput 9.44742K wps
[Epoch 118 Batch 150/162] avg loss 0.00131881, throughput 9.45972K wps
Begin Testing...
[Epoch 118] train avg loss 0.00130584, dev acc 0.9311, dev avg loss 0.199374, throughput 9.42266K wps
Observed Improvement.
Begin Testing...
[Epoch 119 Batch 30/162] avg loss 0.00122319, throughput 9.47212K wps
[Epoch 119 Batch 60/162] avg loss 0.00136209, throughput 9.36105K wps
[Epoch 119 Batch 90/162] avg loss 0.00125783, throughput 9.22653K wps
[Epoch 119 Batch 120/162] avg loss 0.00128243, throughput 9.20817K wps
[Epoch 119 Batch 150/162] avg loss 0.00131484, throughput 9.42213K wps
Begin Testing...
[Epoch 119] train avg loss 0.00129441, dev acc 0.9300, dev avg loss 0.199031, throughput 9.32579K wps
[Epoch 120 Batch 30/162] avg loss 0.00126199, throughput 9.60037K wps
[Epoch 120 Batch 60/162] avg loss 0.00133035, throughput 9.45657K wps
[Epoch 120 Batch 90/162] avg loss 0.00128758, throughput 9.28242K wps
[Epoch 120 Batch 120/162] avg loss 0.00128132, throughput 9.41893K wps
[Epoch 120 Batch 150/162] avg loss 0.00130215, throughput 9.36201K wps
Begin Testing...
[Epoch 120] train avg loss 0.00129626, dev acc 0.9267, dev avg loss 0.199177, throughput 9.41975K wps
[Epoch 121 Batch 30/162] avg loss 0.00114233, throughput 9.58115K wps
[Epoch 121 Batch 60/162] avg loss 0.00125421, throughput 9.46377K wps
[Epoch 121 Batch 90/162] avg loss 0.00131915, throughput 9.26709K wps
[Epoch 121 Batch 120/162] avg loss 0.0011579, throughput 9.42059K wps
[Epoch 121 Batch 150/162] avg loss 0.00128974, throughput 9.44377K wps
Begin Testing...
[Epoch 121] train avg loss 0.00125055, dev acc 0.9256, dev avg loss 0.199479, throughput 9.42811K wps
[Epoch 122 Batch 30/162] avg loss 0.0012395, throughput 9.58242K wps
[Epoch 122 Batch 60/162] avg loss 0.00127934, throughput 9.32832K wps
[Epoch 122 Batch 90/162] avg loss 0.00135989, throughput 9.45706K wps
[Epoch 122 Batch 120/162] avg loss 0.00122102, throughput 9.56583K wps
[Epoch 122 Batch 150/162] avg loss 0.00137403, throughput 9.3757K wps
Begin Testing...
[Epoch 122] train avg loss 0.00127423, dev acc 0.9256, dev avg loss 0.200026, throughput 9.45075K wps
[Epoch 123 Batch 30/162] avg loss 0.00123097, throughput 9.64759K wps
[Epoch 123 Batch 60/162] avg loss 0.00124824, throughput 9.24797K wps
[Epoch 123 Batch 90/162] avg loss 0.00129555, throughput 9.52339K wps
[Epoch 123 Batch 120/162] avg loss 0.00129079, throughput 9.43908K wps
[Epoch 123 Batch 150/162] avg loss 0.00116047, throughput 9.38876K wps
Begin Testing...
[Epoch 123] train avg loss 0.00125174, dev acc 0.9322, dev avg loss 0.198537, throughput 9.44336K wps
Observed Improvement.
Begin Testing...
[Epoch 124 Batch 30/162] avg loss 0.00120542, throughput 9.45576K wps
[Epoch 124 Batch 60/162] avg loss 0.00110786, throughput 9.42433K wps
[Epoch 124 Batch 90/162] avg loss 0.00113806, throughput 9.27189K wps
[Epoch 124 Batch 120/162] avg loss 0.00112598, throughput 9.42717K wps
[Epoch 124 Batch 150/162] avg loss 0.00146502, throughput 9.23453K wps
Begin Testing...
[Epoch 124] train avg loss 0.00120405, dev acc 0.9311, dev avg loss 0.198407, throughput 9.36526K wps
[Epoch 125 Batch 30/162] avg loss 0.00114488, throughput 9.41631K wps
[Epoch 125 Batch 60/162] avg loss 0.00112474, throughput 9.16766K wps
[Epoch 125 Batch 90/162] avg loss 0.00120823, throughput 9.31222K wps
[Epoch 125 Batch 120/162] avg loss 0.00131459, throughput 9.54004K wps
[Epoch 125 Batch 150/162] avg loss 0.00123069, throughput 9.33013K wps
Begin Testing...
[Epoch 125] train avg loss 0.00119058, dev acc 0.9289, dev avg loss 0.19879, throughput 9.34311K wps
[Epoch 126 Batch 30/162] avg loss 0.00115111, throughput 9.4237K wps
[Epoch 126 Batch 60/162] avg loss 0.00115781, throughput 9.33512K wps
[Epoch 126 Batch 90/162] avg loss 0.00119558, throughput 9.18513K wps
[Epoch 126 Batch 120/162] avg loss 0.00126233, throughput 9.22263K wps
[Epoch 126 Batch 150/162] avg loss 0.00114824, throughput 9.25024K wps
Begin Testing...
[Epoch 126] train avg loss 0.00117731, dev acc 0.9267, dev avg loss 0.199156, throughput 9.28786K wps
[Epoch 127 Batch 30/162] avg loss 0.00108638, throughput 9.64139K wps
[Epoch 127 Batch 60/162] avg loss 0.00122516, throughput 9.32021K wps
[Epoch 127 Batch 90/162] avg loss 0.00125105, throughput 9.32356K wps
[Epoch 127 Batch 120/162] avg loss 0.00113019, throughput 9.35535K wps
[Epoch 127 Batch 150/162] avg loss 0.00131444, throughput 9.29766K wps
Begin Testing...
[Epoch 127] train avg loss 0.00118893, dev acc 0.9322, dev avg loss 0.19873, throughput 9.36964K wps
Observed Improvement.
Begin Testing...
[Epoch 128 Batch 30/162] avg loss 0.00130396, throughput 9.71072K wps
[Epoch 128 Batch 60/162] avg loss 0.00114678, throughput 9.37894K wps
[Epoch 128 Batch 90/162] avg loss 0.00102016, throughput 9.46432K wps
[Epoch 128 Batch 120/162] avg loss 0.00126073, throughput 9.26486K wps
[Epoch 128 Batch 150/162] avg loss 0.00125993, throughput 9.30034K wps
Begin Testing...
[Epoch 128] train avg loss 0.00119159, dev acc 0.9311, dev avg loss 0.199058, throughput 9.42134K wps
[Epoch 129 Batch 30/162] avg loss 0.00113545, throughput 9.54132K wps
[Epoch 129 Batch 60/162] avg loss 0.00119513, throughput 9.4028K wps
[Epoch 129 Batch 90/162] avg loss 0.00116344, throughput 9.2748K wps
[Epoch 129 Batch 120/162] avg loss 0.0011865, throughput 9.5926K wps
[Epoch 129 Batch 150/162] avg loss 0.00122651, throughput 9.59699K wps
Begin Testing...
[Epoch 129] train avg loss 0.00117882, dev acc 0.9311, dev avg loss 0.199058, throughput 9.48258K wps
[Epoch 130 Batch 30/162] avg loss 0.00106041, throughput 9.54652K wps
[Epoch 130 Batch 60/162] avg loss 0.00111793, throughput 9.4356K wps
[Epoch 130 Batch 90/162] avg loss 0.00113075, throughput 9.45868K wps
[Epoch 130 Batch 120/162] avg loss 0.00121747, throughput 9.33525K wps
[Epoch 130 Batch 150/162] avg loss 0.00117444, throughput 9.33262K wps
Begin Testing...
[Epoch 130] train avg loss 0.00114068, dev acc 0.9311, dev avg loss 0.199081, throughput 9.42246K wps
[Epoch 131 Batch 30/162] avg loss 0.00114748, throughput 9.52469K wps
[Epoch 131 Batch 60/162] avg loss 0.00116264, throughput 9.48833K wps
[Epoch 131 Batch 90/162] avg loss 0.00111973, throughput 9.24726K wps
[Epoch 131 Batch 120/162] avg loss 0.0010949, throughput 9.43146K wps
[Epoch 131 Batch 150/162] avg loss 0.00104831, throughput 9.30233K wps
Begin Testing...
[Epoch 131] train avg loss 0.00111375, dev acc 0.9311, dev avg loss 0.199268, throughput 9.40373K wps
[Epoch 132 Batch 30/162] avg loss 0.00117632, throughput 9.65601K wps
[Epoch 132 Batch 60/162] avg loss 0.00114367, throughput 9.38117K wps
[Epoch 132 Batch 90/162] avg loss 0.00119868, throughput 9.33662K wps
[Epoch 132 Batch 120/162] avg loss 0.00114817, throughput 9.44067K wps
[Epoch 132 Batch 150/162] avg loss 0.00101591, throughput 9.5173K wps
Begin Testing...
[Epoch 132] train avg loss 0.00113164, dev acc 0.9311, dev avg loss 0.199261, throughput 9.4615K wps
[Epoch 133 Batch 30/162] avg loss 0.0010862, throughput 9.70623K wps
[Epoch 133 Batch 60/162] avg loss 0.00109126, throughput 9.31304K wps
[Epoch 133 Batch 90/162] avg loss 0.00102857, throughput 9.47679K wps
[Epoch 133 Batch 120/162] avg loss 0.0010486, throughput 9.32354K wps
[Epoch 133 Batch 150/162] avg loss 0.00107382, throughput 9.23628K wps
Begin Testing...
[Epoch 133] train avg loss 0.00108574, dev acc 0.9322, dev avg loss 0.199188, throughput 9.41648K wps
Observed Improvement.
Begin Testing...
[Epoch 134 Batch 30/162] avg loss 0.00115467, throughput 9.47643K wps
[Epoch 134 Batch 60/162] avg loss 0.00110443, throughput 9.4626K wps
[Epoch 134 Batch 90/162] avg loss 0.00110047, throughput 9.21819K wps
[Epoch 134 Batch 120/162] avg loss 0.00104304, throughput 9.3083K wps
[Epoch 134 Batch 150/162] avg loss 0.00117293, throughput 9.41173K wps
Begin Testing...
[Epoch 134] train avg loss 0.00110549, dev acc 0.9322, dev avg loss 0.198988, throughput 9.36432K wps
Observed Improvement.
Begin Testing...
[Epoch 135 Batch 30/162] avg loss 0.000991745, throughput 9.65294K wps
[Epoch 135 Batch 60/162] avg loss 0.0010149, throughput 9.22678K wps
[Epoch 135 Batch 90/162] avg loss 0.00119886, throughput 9.41089K wps
[Epoch 135 Batch 120/162] avg loss 0.00118241, throughput 9.22216K wps
[Epoch 135 Batch 150/162] avg loss 0.00113847, throughput 9.36905K wps
Begin Testing...
[Epoch 135] train avg loss 0.0011135, dev acc 0.9300, dev avg loss 0.198867, throughput 9.36878K wps
[Epoch 136 Batch 30/162] avg loss 0.00108496, throughput 9.4102K wps
[Epoch 136 Batch 60/162] avg loss 0.00113656, throughput 9.34808K wps
[Epoch 136 Batch 90/162] avg loss 0.00104562, throughput 9.44032K wps
[Epoch 136 Batch 120/162] avg loss 0.000964258, throughput 9.44368K wps
[Epoch 136 Batch 150/162] avg loss 0.00103422, throughput 9.45424K wps
Begin Testing...
[Epoch 136] train avg loss 0.00105575, dev acc 0.9311, dev avg loss 0.198962, throughput 9.41019K wps
[Epoch 137 Batch 30/162] avg loss 0.000985504, throughput 9.58277K wps
[Epoch 137 Batch 60/162] avg loss 0.00108232, throughput 9.46336K wps
[Epoch 137 Batch 90/162] avg loss 0.00111065, throughput 9.5478K wps
[Epoch 137 Batch 120/162] avg loss 0.000938915, throughput 9.51131K wps
[Epoch 137 Batch 150/162] avg loss 0.000987049, throughput 9.42027K wps
Begin Testing...
[Epoch 137] train avg loss 0.00104075, dev acc 0.9300, dev avg loss 0.198662, throughput 9.49439K wps
[Epoch 138 Batch 30/162] avg loss 0.000983391, throughput 9.45691K wps
[Epoch 138 Batch 60/162] avg loss 0.00104918, throughput 9.3775K wps
[Epoch 138 Batch 90/162] avg loss 0.00114148, throughput 9.29709K wps
[Epoch 138 Batch 120/162] avg loss 0.00105716, throughput 9.47397K wps
[Epoch 138 Batch 150/162] avg loss 0.000948318, throughput 9.33785K wps
Begin Testing...
[Epoch 138] train avg loss 0.00103061, dev acc 0.9322, dev avg loss 0.198622, throughput 9.37361K wps
Observed Improvement.
Begin Testing...
[Epoch 139 Batch 30/162] avg loss 0.00101436, throughput 9.4718K wps
[Epoch 139 Batch 60/162] avg loss 0.00108273, throughput 9.42683K wps
[Epoch 139 Batch 90/162] avg loss 0.00106279, throughput 9.25619K wps
[Epoch 139 Batch 120/162] avg loss 0.0011005, throughput 9.53493K wps
[Epoch 139 Batch 150/162] avg loss 0.00114215, throughput 9.4667K wps
Begin Testing...
[Epoch 139] train avg loss 0.00107766, dev acc 0.9267, dev avg loss 0.199262, throughput 9.43912K wps
[Epoch 140 Batch 30/162] avg loss 0.00115548, throughput 9.67493K wps
[Epoch 140 Batch 60/162] avg loss 0.000960541, throughput 9.40587K wps
[Epoch 140 Batch 90/162] avg loss 0.0010266, throughput 9.3468K wps
[Epoch 140 Batch 120/162] avg loss 0.00101439, throughput 9.3208K wps
[Epoch 140 Batch 150/162] avg loss 0.000913721, throughput 9.59213K wps
Begin Testing...
[Epoch 140] train avg loss 0.00102539, dev acc 0.9278, dev avg loss 0.199041, throughput 9.46759K wps
[Epoch 141 Batch 30/162] avg loss 0.00104014, throughput 9.61873K wps
[Epoch 141 Batch 60/162] avg loss 0.00105698, throughput 9.27773K wps
[Epoch 141 Batch 90/162] avg loss 0.00100694, throughput 9.45673K wps
[Epoch 141 Batch 120/162] avg loss 0.00105628, throughput 9.39983K wps
[Epoch 141 Batch 150/162] avg loss 0.00104997, throughput 9.34983K wps
Begin Testing...
[Epoch 141] train avg loss 0.00103387, dev acc 0.9322, dev avg loss 0.199055, throughput 9.40671K wps
Observed Improvement.
Begin Testing...
[Epoch 142 Batch 30/162] avg loss 0.000887658, throughput 9.55299K wps
[Epoch 142 Batch 60/162] avg loss 0.00104784, throughput 9.33178K wps
[Epoch 142 Batch 90/162] avg loss 0.000964878, throughput 9.32028K wps
[Epoch 142 Batch 120/162] avg loss 0.00104766, throughput 9.34455K wps
[Epoch 142 Batch 150/162] avg loss 0.00104082, throughput 9.29959K wps
Begin Testing...
[Epoch 142] train avg loss 0.000993459, dev acc 0.9344, dev avg loss 0.198146, throughput 9.36844K wps
Observed Improvement.
Begin Testing...
[Epoch 143 Batch 30/162] avg loss 0.000976041, throughput 9.57822K wps
[Epoch 143 Batch 60/162] avg loss 0.000971394, throughput 9.26075K wps
[Epoch 143 Batch 90/162] avg loss 0.000987658, throughput 9.34284K wps
[Epoch 143 Batch 120/162] avg loss 0.00105291, throughput 9.36807K wps
[Epoch 143 Batch 150/162] avg loss 0.000950739, throughput 9.50905K wps
Begin Testing...
[Epoch 143] train avg loss 0.000973638, dev acc 0.9267, dev avg loss 0.199737, throughput 9.41619K wps
[Epoch 144 Batch 30/162] avg loss 0.000952713, throughput 9.41633K wps
[Epoch 144 Batch 60/162] avg loss 0.00102474, throughput 9.38271K wps
[Epoch 144 Batch 90/162] avg loss 0.000948891, throughput 9.46628K wps
[Epoch 144 Batch 120/162] avg loss 0.000970719, throughput 9.53042K wps
[Epoch 144 Batch 150/162] avg loss 0.0010117, throughput 9.55971K wps
Begin Testing...
[Epoch 144] train avg loss 0.000975485, dev acc 0.9333, dev avg loss 0.198338, throughput 9.47488K wps
[Epoch 145 Batch 30/162] avg loss 0.000999671, throughput 9.57715K wps
[Epoch 145 Batch 60/162] avg loss 0.0010243, throughput 9.52931K wps
[Epoch 145 Batch 90/162] avg loss 0.000994705, throughput 9.38481K wps
[Epoch 145 Batch 120/162] avg loss 0.000852265, throughput 9.36572K wps
[Epoch 145 Batch 150/162] avg loss 0.000894153, throughput 9.31787K wps
Begin Testing...
[Epoch 145] train avg loss 0.000959826, dev acc 0.9322, dev avg loss 0.197951, throughput 9.42148K wps
[Epoch 146 Batch 30/162] avg loss 0.0009213, throughput 9.51997K wps
[Epoch 146 Batch 60/162] avg loss 0.000996457, throughput 9.52512K wps
[Epoch 146 Batch 90/162] avg loss 0.000974005, throughput 9.33838K wps
[Epoch 146 Batch 120/162] avg loss 0.000958056, throughput 9.38998K wps
[Epoch 146 Batch 150/162] avg loss 0.000898112, throughput 9.30961K wps
Begin Testing...
[Epoch 146] train avg loss 0.000944362, dev acc 0.9311, dev avg loss 0.198404, throughput 9.40715K wps
[Epoch 147 Batch 30/162] avg loss 0.000833541, throughput 9.51444K wps
[Epoch 147 Batch 60/162] avg loss 0.000823794, throughput 9.5389K wps
[Epoch 147 Batch 90/162] avg loss 0.00104564, throughput 9.4814K wps
[Epoch 147 Batch 120/162] avg loss 0.00101268, throughput 9.45189K wps
[Epoch 147 Batch 150/162] avg loss 0.00108087, throughput 9.47941K wps
Begin Testing...
[Epoch 147] train avg loss 0.000962311, dev acc 0.9322, dev avg loss 0.198059, throughput 9.48631K wps
[Epoch 148 Batch 30/162] avg loss 0.000961202, throughput 9.59616K wps
[Epoch 148 Batch 60/162] avg loss 0.000909715, throughput 9.2111K wps
[Epoch 148 Batch 90/162] avg loss 0.00097298, throughput 9.42474K wps
[Epoch 148 Batch 120/162] avg loss 0.000981802, throughput 9.46109K wps
[Epoch 148 Batch 150/162] avg loss 0.000926924, throughput 9.42846K wps
Begin Testing...
[Epoch 148] train avg loss 0.000952834, dev acc 0.9300, dev avg loss 0.198461, throughput 9.42231K wps
[Epoch 149 Batch 30/162] avg loss 0.00094544, throughput 9.59141K wps
[Epoch 149 Batch 60/162] avg loss 0.000798458, throughput 9.39647K wps
[Epoch 149 Batch 90/162] avg loss 0.000941157, throughput 9.48778K wps
[Epoch 149 Batch 120/162] avg loss 0.00103026, throughput 9.37045K wps
[Epoch 149 Batch 150/162] avg loss 0.00100158, throughput 9.4224K wps
Begin Testing...
[Epoch 149] train avg loss 0.00095037, dev acc 0.9322, dev avg loss 0.199338, throughput 9.45777K wps
[Epoch 150 Batch 30/162] avg loss 0.000941625, throughput 9.73412K wps
[Epoch 150 Batch 60/162] avg loss 0.000918315, throughput 9.47646K wps
[Epoch 150 Batch 90/162] avg loss 0.000879122, throughput 9.38709K wps
[Epoch 150 Batch 120/162] avg loss 0.000972516, throughput 9.3787K wps
[Epoch 150 Batch 150/162] avg loss 0.00093079, throughput 9.26623K wps
Begin Testing...
[Epoch 150] train avg loss 0.000932415, dev acc 0.9322, dev avg loss 0.198455, throughput 9.44923K wps
[Epoch 151 Batch 30/162] avg loss 0.00101544, throughput 9.52437K wps
[Epoch 151 Batch 60/162] avg loss 0.000847919, throughput 9.28802K wps
[Epoch 151 Batch 90/162] avg loss 0.000894011, throughput 9.41631K wps
[Epoch 151 Batch 120/162] avg loss 0.000859629, throughput 9.24288K wps
[Epoch 151 Batch 150/162] avg loss 0.000998474, throughput 9.22177K wps
Begin Testing...
[Epoch 151] train avg loss 0.000921964, dev acc 0.9289, dev avg loss 0.198911, throughput 9.33929K wps
[Epoch 152 Batch 30/162] avg loss 0.000965267, throughput 9.56187K wps
[Epoch 152 Batch 60/162] avg loss 0.000948554, throughput 9.28427K wps
[Epoch 152 Batch 90/162] avg loss 0.000796564, throughput 9.40172K wps
[Epoch 152 Batch 120/162] avg loss 0.000977773, throughput 9.28411K wps
[Epoch 152 Batch 150/162] avg loss 0.00079192, throughput 9.35199K wps
Begin Testing...
[Epoch 152] train avg loss 0.000899693, dev acc 0.9344, dev avg loss 0.198833, throughput 9.37892K wps
Observed Improvement.
Begin Testing...
[Epoch 153 Batch 30/162] avg loss 0.000805971, throughput 9.44939K wps
[Epoch 153 Batch 60/162] avg loss 0.000873838, throughput 9.18985K wps
[Epoch 153 Batch 90/162] avg loss 0.000994105, throughput 9.57794K wps
[Epoch 153 Batch 120/162] avg loss 0.000861975, throughput 9.38475K wps
[Epoch 153 Batch 150/162] avg loss 0.000864721, throughput 9.4132K wps
Begin Testing...
[Epoch 153] train avg loss 0.000902006, dev acc 0.9289, dev avg loss 0.199371, throughput 9.4132K wps
[Epoch 154 Batch 30/162] avg loss 0.000854582, throughput 9.56797K wps
[Epoch 154 Batch 60/162] avg loss 0.000876126, throughput 9.38431K wps
[Epoch 154 Batch 90/162] avg loss 0.000881669, throughput 9.35013K wps
[Epoch 154 Batch 120/162] avg loss 0.000845758, throughput 9.40708K wps
[Epoch 154 Batch 150/162] avg loss 0.000901686, throughput 9.38583K wps
Begin Testing...
[Epoch 154] train avg loss 0.000883258, dev acc 0.9311, dev avg loss 0.198855, throughput 9.4104K wps
[Epoch 155 Batch 30/162] avg loss 0.000877631, throughput 9.62206K wps
[Epoch 155 Batch 60/162] avg loss 0.00094612, throughput 9.55996K wps
[Epoch 155 Batch 90/162] avg loss 0.000838889, throughput 9.31259K wps
[Epoch 155 Batch 120/162] avg loss 0.000806639, throughput 9.41421K wps
[Epoch 155 Batch 150/162] avg loss 0.000920395, throughput 9.22436K wps
Begin Testing...
[Epoch 155] train avg loss 0.000877413, dev acc 0.9344, dev avg loss 0.199072, throughput 9.4179K wps
Observed Improvement.
Begin Testing...
[Epoch 156 Batch 30/162] avg loss 0.000917375, throughput 9.66404K wps
[Epoch 156 Batch 60/162] avg loss 0.000891523, throughput 9.58484K wps
[Epoch 156 Batch 90/162] avg loss 0.000843721, throughput 9.5537K wps
[Epoch 156 Batch 120/162] avg loss 0.000926243, throughput 9.5482K wps
[Epoch 156 Batch 150/162] avg loss 0.000885359, throughput 9.49341K wps
Begin Testing...
[Epoch 156] train avg loss 0.000888057, dev acc 0.9300, dev avg loss 0.199528, throughput 9.54813K wps
[Epoch 157 Batch 30/162] avg loss 0.000883461, throughput 9.73445K wps
[Epoch 157 Batch 60/162] avg loss 0.00081473, throughput 9.37778K wps
[Epoch 157 Batch 90/162] avg loss 0.000874448, throughput 9.42499K wps
[Epoch 157 Batch 120/162] avg loss 0.000875227, throughput 9.26044K wps
[Epoch 157 Batch 150/162] avg loss 0.000830335, throughput 9.40601K wps
Begin Testing...
[Epoch 157] train avg loss 0.000852687, dev acc 0.9311, dev avg loss 0.199195, throughput 9.43721K wps
[Epoch 158 Batch 30/162] avg loss 0.00079989, throughput 9.70167K wps
[Epoch 158 Batch 60/162] avg loss 0.000988619, throughput 9.3357K wps
[Epoch 158 Batch 90/162] avg loss 0.000847354, throughput 9.32659K wps
[Epoch 158 Batch 120/162] avg loss 0.000791548, throughput 9.37937K wps
[Epoch 158 Batch 150/162] avg loss 0.000873491, throughput 9.41736K wps
Begin Testing...
[Epoch 158] train avg loss 0.000855033, dev acc 0.9311, dev avg loss 0.19911, throughput 9.42603K wps
[Epoch 159 Batch 30/162] avg loss 0.00077084, throughput 9.67629K wps
[Epoch 159 Batch 60/162] avg loss 0.000712891, throughput 9.58881K wps
[Epoch 159 Batch 90/162] avg loss 0.000753351, throughput 9.41029K wps
[Epoch 159 Batch 120/162] avg loss 0.000934713, throughput 9.5363K wps
[Epoch 159 Batch 150/162] avg loss 0.000829498, throughput 9.24886K wps
Begin Testing...
[Epoch 159] train avg loss 0.000808835, dev acc 0.9311, dev avg loss 0.19963, throughput 9.47987K wps
[Epoch 160 Batch 30/162] avg loss 0.000808971, throughput 9.51016K wps
[Epoch 160 Batch 60/162] avg loss 0.000879163, throughput 9.48295K wps
[Epoch 160 Batch 90/162] avg loss 0.000803701, throughput 9.26398K wps
[Epoch 160 Batch 120/162] avg loss 0.000860717, throughput 9.23514K wps
[Epoch 160 Batch 150/162] avg loss 0.000750216, throughput 9.42975K wps
Begin Testing...
[Epoch 160] train avg loss 0.000829286, dev acc 0.9311, dev avg loss 0.199735, throughput 9.3919K wps
[Epoch 161 Batch 30/162] avg loss 0.000974678, throughput 9.54246K wps
[Epoch 161 Batch 60/162] avg loss 0.000845298, throughput 9.41976K wps
[Epoch 161 Batch 90/162] avg loss 0.000808872, throughput 9.37979K wps
[Epoch 161 Batch 120/162] avg loss 0.000905347, throughput 9.24486K wps
[Epoch 161 Batch 150/162] avg loss 0.000705771, throughput 9.40624K wps
Begin Testing...
[Epoch 161] train avg loss 0.000844436, dev acc 0.9289, dev avg loss 0.199593, throughput 9.41192K wps
[Epoch 162 Batch 30/162] avg loss 0.000703221, throughput 9.69891K wps
[Epoch 162 Batch 60/162] avg loss 0.000955533, throughput 9.26135K wps
[Epoch 162 Batch 90/162] avg loss 0.000947767, throughput 9.27048K wps
[Epoch 162 Batch 120/162] avg loss 0.000730136, throughput 9.36203K wps
[Epoch 162 Batch 150/162] avg loss 0.000868703, throughput 9.51287K wps
Begin Testing...
[Epoch 162] train avg loss 0.000837811, dev acc 0.9289, dev avg loss 0.201457, throughput 9.43234K wps
[Epoch 163 Batch 30/162] avg loss 0.000796378, throughput 9.63017K wps
[Epoch 163 Batch 60/162] avg loss 0.000855796, throughput 9.3025K wps
[Epoch 163 Batch 90/162] avg loss 0.000751803, throughput 9.40544K wps
[Epoch 163 Batch 120/162] avg loss 0.000802561, throughput 9.47464K wps
[Epoch 163 Batch 150/162] avg loss 0.000814194, throughput 9.3446K wps
Begin Testing...
[Epoch 163] train avg loss 0.000806808, dev acc 0.9311, dev avg loss 0.199386, throughput 9.4354K wps
[Epoch 164 Batch 30/162] avg loss 0.000694118, throughput 9.72815K wps
[Epoch 164 Batch 60/162] avg loss 0.000772537, throughput 9.35489K wps
[Epoch 164 Batch 90/162] avg loss 0.000798749, throughput 9.45838K wps
[Epoch 164 Batch 120/162] avg loss 0.000732335, throughput 9.42304K wps
[Epoch 164 Batch 150/162] avg loss 0.000873458, throughput 9.4681K wps
Begin Testing...
[Epoch 164] train avg loss 0.000774546, dev acc 0.9322, dev avg loss 0.199672, throughput 9.47183K wps
[Epoch 165 Batch 30/162] avg loss 0.000755503, throughput 9.50419K wps
[Epoch 165 Batch 60/162] avg loss 0.000855095, throughput 9.48564K wps
[Epoch 165 Batch 90/162] avg loss 0.000718117, throughput 9.33272K wps
[Epoch 165 Batch 120/162] avg loss 0.000749274, throughput 9.50652K wps
[Epoch 165 Batch 150/162] avg loss 0.000889802, throughput 9.43312K wps
Begin Testing...
[Epoch 165] train avg loss 0.000797565, dev acc 0.9322, dev avg loss 0.199214, throughput 9.43677K wps
[Epoch 166 Batch 30/162] avg loss 0.000786124, throughput 9.50313K wps
[Epoch 166 Batch 60/162] avg loss 0.000757047, throughput 9.40146K wps
[Epoch 166 Batch 90/162] avg loss 0.000846746, throughput 9.37472K wps
[Epoch 166 Batch 120/162] avg loss 0.000822213, throughput 9.53916K wps
[Epoch 166 Batch 150/162] avg loss 0.000710454, throughput 9.288K wps
Begin Testing...
[Epoch 166] train avg loss 0.000792623, dev acc 0.9311, dev avg loss 0.19945, throughput 9.41592K wps
[Epoch 167 Batch 30/162] avg loss 0.00078465, throughput 9.7727K wps
[Epoch 167 Batch 60/162] avg loss 0.000749043, throughput 9.39833K wps
[Epoch 167 Batch 90/162] avg loss 0.000751108, throughput 9.28201K wps
[Epoch 167 Batch 120/162] avg loss 0.000842619, throughput 9.51506K wps
[Epoch 167 Batch 150/162] avg loss 0.000828416, throughput 9.48429K wps
Begin Testing...
[Epoch 167] train avg loss 0.000775596, dev acc 0.9344, dev avg loss 0.199765, throughput 9.4774K wps
Observed Improvement.
Begin Testing...
[Epoch 168 Batch 30/162] avg loss 0.000750158, throughput 9.48357K wps
[Epoch 168 Batch 60/162] avg loss 0.000751583, throughput 9.45087K wps
[Epoch 168 Batch 90/162] avg loss 0.0008342, throughput 9.40624K wps
[Epoch 168 Batch 120/162] avg loss 0.000724395, throughput 9.41174K wps
[Epoch 168 Batch 150/162] avg loss 0.000741043, throughput 9.29438K wps
Begin Testing...
[Epoch 168] train avg loss 0.000766779, dev acc 0.9311, dev avg loss 0.199704, throughput 9.41703K wps
[Epoch 169 Batch 30/162] avg loss 0.00074829, throughput 9.50446K wps
[Epoch 169 Batch 60/162] avg loss 0.000697833, throughput 9.35994K wps
[Epoch 169 Batch 90/162] avg loss 0.000785212, throughput 9.47869K wps
[Epoch 169 Batch 120/162] avg loss 0.000862278, throughput 9.40258K wps
[Epoch 169 Batch 150/162] avg loss 0.000765823, throughput 9.44211K wps
Begin Testing...
[Epoch 169] train avg loss 0.000781461, dev acc 0.9289, dev avg loss 0.200178, throughput 9.42419K wps
[Epoch 170 Batch 30/162] avg loss 0.000665183, throughput 9.59721K wps
[Epoch 170 Batch 60/162] avg loss 0.000721135, throughput 9.40824K wps
[Epoch 170 Batch 90/162] avg loss 0.0007045, throughput 9.32782K wps
[Epoch 170 Batch 120/162] avg loss 0.000762113, throughput 9.30466K wps
[Epoch 170 Batch 150/162] avg loss 0.000850951, throughput 9.24936K wps
Begin Testing...
[Epoch 170] train avg loss 0.000734841, dev acc 0.9333, dev avg loss 0.200847, throughput 9.36183K wps
[Epoch 171 Batch 30/162] avg loss 0.000838534, throughput 9.62985K wps
[Epoch 171 Batch 60/162] avg loss 0.000800776, throughput 9.42853K wps
[Epoch 171 Batch 90/162] avg loss 0.000743843, throughput 9.41032K wps
[Epoch 171 Batch 120/162] avg loss 0.000806597, throughput 9.29138K wps
[Epoch 171 Batch 150/162] avg loss 0.00069783, throughput 9.50147K wps
Begin Testing...
[Epoch 171] train avg loss 0.000776812, dev acc 0.9300, dev avg loss 0.200217, throughput 9.43441K wps
[Epoch 172 Batch 30/162] avg loss 0.000723996, throughput 9.50593K wps
[Epoch 172 Batch 60/162] avg loss 0.000680573, throughput 9.47756K wps
[Epoch 172 Batch 90/162] avg loss 0.000735702, throughput 9.35396K wps
[Epoch 172 Batch 120/162] avg loss 0.000859069, throughput 9.27223K wps
[Epoch 172 Batch 150/162] avg loss 0.000791226, throughput 9.30891K wps
Begin Testing...
[Epoch 172] train avg loss 0.000755193, dev acc 0.9311, dev avg loss 0.199716, throughput 9.39513K wps
[Epoch 173 Batch 30/162] avg loss 0.000786021, throughput 9.63078K wps
[Epoch 173 Batch 60/162] avg loss 0.000690998, throughput 9.32488K wps
[Epoch 173 Batch 90/162] avg loss 0.000712721, throughput 9.39249K wps
[Epoch 173 Batch 120/162] avg loss 0.000668333, throughput 9.41696K wps
[Epoch 173 Batch 150/162] avg loss 0.000770785, throughput 9.43979K wps
Begin Testing...
[Epoch 173] train avg loss 0.000714309, dev acc 0.9311, dev avg loss 0.200101, throughput 9.41934K wps
[Epoch 174 Batch 30/162] avg loss 0.000684969, throughput 9.76961K wps
[Epoch 174 Batch 60/162] avg loss 0.000735002, throughput 9.3513K wps
[Epoch 174 Batch 90/162] avg loss 0.000772401, throughput 9.34882K wps
[Epoch 174 Batch 120/162] avg loss 0.000708095, throughput 9.59746K wps
[Epoch 174 Batch 150/162] avg loss 0.000749415, throughput 9.32475K wps
Begin Testing...
[Epoch 174] train avg loss 0.000722941, dev acc 0.9333, dev avg loss 0.200322, throughput 9.4791K wps
[Epoch 175 Batch 30/162] avg loss 0.000776178, throughput 9.55876K wps
[Epoch 175 Batch 60/162] avg loss 0.000646456, throughput 9.36128K wps
[Epoch 175 Batch 90/162] avg loss 0.00077226, throughput 9.23862K wps
[Epoch 175 Batch 120/162] avg loss 0.000751662, throughput 9.40855K wps
[Epoch 175 Batch 150/162] avg loss 0.000762339, throughput 9.50439K wps
Begin Testing...
[Epoch 175] train avg loss 0.000741196, dev acc 0.9311, dev avg loss 0.200233, throughput 9.40618K wps
[Epoch 176 Batch 30/162] avg loss 0.000718871, throughput 9.54854K wps
[Epoch 176 Batch 60/162] avg loss 0.000761122, throughput 9.29165K wps
[Epoch 176 Batch 90/162] avg loss 0.000761968, throughput 9.51773K wps
[Epoch 176 Batch 120/162] avg loss 0.000767324, throughput 9.45121K wps
[Epoch 176 Batch 150/162] avg loss 0.000621879, throughput 9.37256K wps
Begin Testing...
[Epoch 176] train avg loss 0.000716771, dev acc 0.9322, dev avg loss 0.199961, throughput 9.42582K wps
[Epoch 177 Batch 30/162] avg loss 0.000749861, throughput 9.62679K wps
[Epoch 177 Batch 60/162] avg loss 0.00070125, throughput 9.32116K wps
[Epoch 177 Batch 90/162] avg loss 0.000648393, throughput 9.36197K wps
[Epoch 177 Batch 120/162] avg loss 0.000711747, throughput 9.3052K wps
[Epoch 177 Batch 150/162] avg loss 0.000782345, throughput 9.31305K wps
Begin Testing...
[Epoch 177] train avg loss 0.000716524, dev acc 0.9300, dev avg loss 0.2003, throughput 9.36782K wps
[Epoch 178 Batch 30/162] avg loss 0.000694793, throughput 9.58033K wps
[Epoch 178 Batch 60/162] avg loss 0.000639935, throughput 9.37018K wps
[Epoch 178 Batch 90/162] avg loss 0.000586784, throughput 9.32866K wps
[Epoch 178 Batch 120/162] avg loss 0.000634432, throughput 9.41552K wps
[Epoch 178 Batch 150/162] avg loss 0.000779367, throughput 9.34433K wps
Begin Testing...
[Epoch 178] train avg loss 0.000680366, dev acc 0.9322, dev avg loss 0.200603, throughput 9.39735K wps
[Epoch 179 Batch 30/162] avg loss 0.000694844, throughput 9.45319K wps
[Epoch 179 Batch 60/162] avg loss 0.000702892, throughput 9.41916K wps
[Epoch 179 Batch 90/162] avg loss 0.000746897, throughput 9.54811K wps
[Epoch 179 Batch 120/162] avg loss 0.000635815, throughput 9.36965K wps
[Epoch 179 Batch 150/162] avg loss 0.000724774, throughput 9.27574K wps
Begin Testing...
[Epoch 179] train avg loss 0.000697136, dev acc 0.9333, dev avg loss 0.200535, throughput 9.42392K wps
[Epoch 180 Batch 30/162] avg loss 0.000765003, throughput 9.43349K wps
[Epoch 180 Batch 60/162] avg loss 0.00069129, throughput 9.36319K wps
[Epoch 180 Batch 90/162] avg loss 0.000677021, throughput 9.31895K wps
[Epoch 180 Batch 120/162] avg loss 0.000646422, throughput 9.31929K wps
[Epoch 180 Batch 150/162] avg loss 0.000598789, throughput 9.3553K wps
Begin Testing...
[Epoch 180] train avg loss 0.000685517, dev acc 0.9333, dev avg loss 0.201007, throughput 9.38013K wps
[Epoch 181 Batch 30/162] avg loss 0.000629044, throughput 9.54956K wps
[Epoch 181 Batch 60/162] avg loss 0.000800723, throughput 9.44998K wps
[Epoch 181 Batch 90/162] avg loss 0.000714992, throughput 9.32744K wps
[Epoch 181 Batch 120/162] avg loss 0.000725197, throughput 9.30969K wps
[Epoch 181 Batch 150/162] avg loss 0.000709321, throughput 9.35682K wps
Begin Testing...
[Epoch 181] train avg loss 0.000712253, dev acc 0.9333, dev avg loss 0.200674, throughput 9.41087K wps
[Epoch 182 Batch 30/162] avg loss 0.000742768, throughput 9.64101K wps
[Epoch 182 Batch 60/162] avg loss 0.000596728, throughput 9.33111K wps
[Epoch 182 Batch 90/162] avg loss 0.000640098, throughput 9.42153K wps
[Epoch 182 Batch 120/162] avg loss 0.000700641, throughput 9.40956K wps
[Epoch 182 Batch 150/162] avg loss 0.000668972, throughput 9.48947K wps
Begin Testing...
[Epoch 182] train avg loss 0.000669023, dev acc 0.9333, dev avg loss 0.200928, throughput 9.46244K wps
[Epoch 183 Batch 30/162] avg loss 0.000716692, throughput 9.61282K wps
[Epoch 183 Batch 60/162] avg loss 0.000643486, throughput 9.30336K wps
[Epoch 183 Batch 90/162] avg loss 0.000788289, throughput 9.30866K wps
[Epoch 183 Batch 120/162] avg loss 0.000729306, throughput 9.5337K wps
[Epoch 183 Batch 150/162] avg loss 0.000670616, throughput 9.38849K wps
Begin Testing...
[Epoch 183] train avg loss 0.000703653, dev acc 0.9333, dev avg loss 0.201005, throughput 9.41901K wps
[Epoch 184 Batch 30/162] avg loss 0.000672865, throughput 9.57261K wps
[Epoch 184 Batch 60/162] avg loss 0.000736718, throughput 9.30365K wps
[Epoch 184 Batch 90/162] avg loss 0.000653935, throughput 9.46661K wps
[Epoch 184 Batch 120/162] avg loss 0.000689856, throughput 9.49916K wps
[Epoch 184 Batch 150/162] avg loss 0.000644164, throughput 9.50608K wps
Begin Testing...
[Epoch 184] train avg loss 0.000681758, dev acc 0.9333, dev avg loss 0.20135, throughput 9.46569K wps
[Epoch 185 Batch 30/162] avg loss 0.000623803, throughput 9.54737K wps
[Epoch 185 Batch 60/162] avg loss 0.000700642, throughput 9.57664K wps
[Epoch 185 Batch 90/162] avg loss 0.000708221, throughput 9.4525K wps
[Epoch 185 Batch 120/162] avg loss 0.000657087, throughput 9.38965K wps
[Epoch 185 Batch 150/162] avg loss 0.000690085, throughput 9.19199K wps
Begin Testing...
[Epoch 185] train avg loss 0.000683682, dev acc 0.9322, dev avg loss 0.201315, throughput 9.43165K wps
[Epoch 186 Batch 30/162] avg loss 0.000609656, throughput 9.55662K wps
[Epoch 186 Batch 60/162] avg loss 0.000669506, throughput 9.30114K wps
[Epoch 186 Batch 90/162] avg loss 0.000723473, throughput 9.32232K wps
[Epoch 186 Batch 120/162] avg loss 0.000599758, throughput 9.54169K wps
[Epoch 186 Batch 150/162] avg loss 0.000777129, throughput 9.25565K wps
Begin Testing...
[Epoch 186] train avg loss 0.00066921, dev acc 0.9333, dev avg loss 0.20093, throughput 9.39597K wps
[Epoch 187 Batch 30/162] avg loss 0.000697954, throughput 9.56033K wps
[Epoch 187 Batch 60/162] avg loss 0.000675127, throughput 9.35616K wps
[Epoch 187 Batch 90/162] avg loss 0.000613366, throughput 9.27318K wps
[Epoch 187 Batch 120/162] avg loss 0.000621279, throughput 9.39594K wps
[Epoch 187 Batch 150/162] avg loss 0.000611037, throughput 9.32422K wps
Begin Testing...
[Epoch 187] train avg loss 0.000645539, dev acc 0.9344, dev avg loss 0.201274, throughput 9.38123K wps
Observed Improvement.
Begin Testing...
[Epoch 188 Batch 30/162] avg loss 0.000683857, throughput 9.56185K wps
[Epoch 188 Batch 60/162] avg loss 0.000638752, throughput 9.33461K wps
[Epoch 188 Batch 90/162] avg loss 0.000626189, throughput 9.38546K wps
[Epoch 188 Batch 120/162] avg loss 0.000655623, throughput 9.36592K wps
[Epoch 188 Batch 150/162] avg loss 0.000704469, throughput 9.31537K wps
Begin Testing...
[Epoch 188] train avg loss 0.000660655, dev acc 0.9344, dev avg loss 0.202151, throughput 9.37124K wps
Observed Improvement.
Begin Testing...
[Epoch 189 Batch 30/162] avg loss 0.000737478, throughput 9.59331K wps
[Epoch 189 Batch 60/162] avg loss 0.000649084, throughput 9.3144K wps
[Epoch 189 Batch 90/162] avg loss 0.000635279, throughput 9.44601K wps
[Epoch 189 Batch 120/162] avg loss 0.000621244, throughput 9.50741K wps
[Epoch 189 Batch 150/162] avg loss 0.000643737, throughput 9.30601K wps
Begin Testing...
[Epoch 189] train avg loss 0.000656525, dev acc 0.9333, dev avg loss 0.20168, throughput 9.42289K wps
[Epoch 190 Batch 30/162] avg loss 0.00060197, throughput 9.54444K wps
[Epoch 190 Batch 60/162] avg loss 0.000594402, throughput 9.55803K wps
[Epoch 190 Batch 90/162] avg loss 0.000541556, throughput 9.3759K wps
[Epoch 190 Batch 120/162] avg loss 0.000611504, throughput 9.29774K wps
[Epoch 190 Batch 150/162] avg loss 0.0006616, throughput 9.23232K wps
Begin Testing...
[Epoch 190] train avg loss 0.000611076, dev acc 0.9322, dev avg loss 0.202138, throughput 9.3941K wps
[Epoch 191 Batch 30/162] avg loss 0.000583513, throughput 9.54981K wps
[Epoch 191 Batch 60/162] avg loss 0.000607934, throughput 9.38146K wps
[Epoch 191 Batch 90/162] avg loss 0.000771854, throughput 9.33102K wps
[Epoch 191 Batch 120/162] avg loss 0.000666878, throughput 9.27158K wps
[Epoch 191 Batch 150/162] avg loss 0.000567977, throughput 9.40895K wps
Begin Testing...
[Epoch 191] train avg loss 0.000644209, dev acc 0.9333, dev avg loss 0.20221, throughput 9.39202K wps
[Epoch 192 Batch 30/162] avg loss 0.000730085, throughput 9.49941K wps
[Epoch 192 Batch 60/162] avg loss 0.000583115, throughput 9.39319K wps
[Epoch 192 Batch 90/162] avg loss 0.000663236, throughput 9.45732K wps
[Epoch 192 Batch 120/162] avg loss 0.000608084, throughput 9.51189K wps
[Epoch 192 Batch 150/162] avg loss 0.000570181, throughput 9.56225K wps
Begin Testing...
[Epoch 192] train avg loss 0.000618323, dev acc 0.9322, dev avg loss 0.202221, throughput 9.45856K wps
[Epoch 193 Batch 30/162] avg loss 0.000572176, throughput 9.60178K wps
[Epoch 193 Batch 60/162] avg loss 0.000565171, throughput 9.22527K wps
[Epoch 193 Batch 90/162] avg loss 0.000631348, throughput 9.47041K wps
[Epoch 193 Batch 120/162] avg loss 0.000595776, throughput 9.33119K wps
[Epoch 193 Batch 150/162] avg loss 0.000636903, throughput 9.37086K wps
Begin Testing...
[Epoch 193] train avg loss 0.000610488, dev acc 0.9344, dev avg loss 0.202612, throughput 9.40767K wps
Observed Improvement.
Begin Testing...
[Epoch 194 Batch 30/162] avg loss 0.000685029, throughput 9.54745K wps
[Epoch 194 Batch 60/162] avg loss 0.00062756, throughput 9.2748K wps
[Epoch 194 Batch 90/162] avg loss 0.000635914, throughput 9.31694K wps
[Epoch 194 Batch 120/162] avg loss 0.000535939, throughput 9.36308K wps
[Epoch 194 Batch 150/162] avg loss 0.000592485, throughput 9.24411K wps
Begin Testing...
[Epoch 194] train avg loss 0.000609878, dev acc 0.9300, dev avg loss 0.202956, throughput 9.34897K wps
[Epoch 195 Batch 30/162] avg loss 0.000644012, throughput 9.55858K wps
[Epoch 195 Batch 60/162] avg loss 0.000567187, throughput 9.33549K wps
[Epoch 195 Batch 90/162] avg loss 0.00067197, throughput 9.34275K wps
[Epoch 195 Batch 120/162] avg loss 0.000632038, throughput 9.24271K wps
[Epoch 195 Batch 150/162] avg loss 0.000684886, throughput 9.475K wps
Begin Testing...
[Epoch 195] train avg loss 0.000636783, dev acc 0.9322, dev avg loss 0.202272, throughput 9.38602K wps
[Epoch 196 Batch 30/162] avg loss 0.000696987, throughput 9.51412K wps
[Epoch 196 Batch 60/162] avg loss 0.000623682, throughput 9.47651K wps
[Epoch 196 Batch 90/162] avg loss 0.000584644, throughput 9.39549K wps
[Epoch 196 Batch 120/162] avg loss 0.000528887, throughput 9.49222K wps
[Epoch 196 Batch 150/162] avg loss 0.0005824, throughput 9.44647K wps
Begin Testing...
[Epoch 196] train avg loss 0.000590905, dev acc 0.9333, dev avg loss 0.202797, throughput 9.45447K wps
[Epoch 197 Batch 30/162] avg loss 0.000651228, throughput 9.63591K wps
[Epoch 197 Batch 60/162] avg loss 0.000498823, throughput 9.3449K wps
[Epoch 197 Batch 90/162] avg loss 0.000651713, throughput 9.45516K wps
[Epoch 197 Batch 120/162] avg loss 0.000642572, throughput 9.35166K wps
[Epoch 197 Batch 150/162] avg loss 0.000611978, throughput 9.3093K wps
Begin Testing...
[Epoch 197] train avg loss 0.000606933, dev acc 0.9322, dev avg loss 0.202357, throughput 9.40384K wps
[Epoch 198 Batch 30/162] avg loss 0.000587599, throughput 9.59051K wps
[Epoch 198 Batch 60/162] avg loss 0.000625381, throughput 9.40191K wps
[Epoch 198 Batch 90/162] avg loss 0.000543357, throughput 9.46783K wps
[Epoch 198 Batch 120/162] avg loss 0.000748192, throughput 9.31532K wps
[Epoch 198 Batch 150/162] avg loss 0.000580954, throughput 9.30742K wps
Begin Testing...
[Epoch 198] train avg loss 0.000609185, dev acc 0.9344, dev avg loss 0.202302, throughput 9.4086K wps
Observed Improvement.
Begin Testing...
[Epoch 199 Batch 30/162] avg loss 0.00056139, throughput 9.52406K wps
[Epoch 199 Batch 60/162] avg loss 0.000584519, throughput 9.24864K wps
[Epoch 199 Batch 90/162] avg loss 0.000607698, throughput 9.32767K wps
[Epoch 199 Batch 120/162] avg loss 0.000543109, throughput 9.31358K wps
[Epoch 199 Batch 150/162] avg loss 0.00053073, throughput 9.47223K wps
Begin Testing...
[Epoch 199] train avg loss 0.000561443, dev acc 0.9311, dev avg loss 0.202647, throughput 9.37277K wps
Test loss 0.234349, test acc 0.9150
Total time cost 445.65s
[Epoch 0 Batch 30/162] avg loss 0.0140324, throughput 7.31594K wps
[Epoch 0 Batch 60/162] avg loss 0.0137697, throughput 9.53558K wps
[Epoch 0 Batch 90/162] avg loss 0.013667, throughput 9.42413K wps
[Epoch 0 Batch 120/162] avg loss 0.0135796, throughput 9.41052K wps
[Epoch 0 Batch 150/162] avg loss 0.0135132, throughput 9.30372K wps
Begin Testing...
[Epoch 0] train avg loss 0.0136906, dev acc 0.6933, dev avg loss 0.663671, throughput 8.95059K wps
Observed Improvement.
Begin Testing...
[Epoch 1 Batch 30/162] avg loss 0.0132853, throughput 9.56229K wps
[Epoch 1 Batch 60/162] avg loss 0.0132598, throughput 9.34035K wps
[Epoch 1 Batch 90/162] avg loss 0.0131167, throughput 9.30347K wps
[Epoch 1 Batch 120/162] avg loss 0.0128403, throughput 9.25713K wps
[Epoch 1 Batch 150/162] avg loss 0.0128463, throughput 9.50765K wps
Begin Testing...
[Epoch 1] train avg loss 0.0130497, dev acc 0.8122, dev avg loss 0.635224, throughput 9.38824K wps
Observed Improvement.
Begin Testing...
[Epoch 2 Batch 30/162] avg loss 0.0126278, throughput 9.46677K wps
[Epoch 2 Batch 60/162] avg loss 0.0126171, throughput 9.33006K wps
[Epoch 2 Batch 90/162] avg loss 0.0123981, throughput 9.42421K wps
[Epoch 2 Batch 120/162] avg loss 0.0123127, throughput 9.41583K wps
[Epoch 2 Batch 150/162] avg loss 0.0121766, throughput 9.29956K wps
Begin Testing...
[Epoch 2] train avg loss 0.0124066, dev acc 0.8222, dev avg loss 0.602879, throughput 9.39711K wps
Observed Improvement.
Begin Testing...
[Epoch 3 Batch 30/162] avg loss 0.012074, throughput 9.77186K wps
[Epoch 3 Batch 60/162] avg loss 0.0119799, throughput 9.41283K wps
[Epoch 3 Batch 90/162] avg loss 0.0116279, throughput 9.50668K wps
[Epoch 3 Batch 120/162] avg loss 0.0116906, throughput 9.35647K wps
[Epoch 3 Batch 150/162] avg loss 0.0114773, throughput 9.56254K wps
Begin Testing...
[Epoch 3] train avg loss 0.0117511, dev acc 0.8456, dev avg loss 0.566611, throughput 9.49692K wps
Observed Improvement.
Begin Testing...
[Epoch 4 Batch 30/162] avg loss 0.0113544, throughput 9.40342K wps
[Epoch 4 Batch 60/162] avg loss 0.0110726, throughput 9.44167K wps
[Epoch 4 Batch 90/162] avg loss 0.0109634, throughput 9.49997K wps
[Epoch 4 Batch 120/162] avg loss 0.0108205, throughput 9.45509K wps
[Epoch 4 Batch 150/162] avg loss 0.0105801, throughput 9.43651K wps
Begin Testing...
[Epoch 4] train avg loss 0.0109219, dev acc 0.8589, dev avg loss 0.526782, throughput 9.43199K wps
Observed Improvement.
Begin Testing...
[Epoch 5 Batch 30/162] avg loss 0.0104652, throughput 9.6518K wps
[Epoch 5 Batch 60/162] avg loss 0.0104402, throughput 9.27921K wps
[Epoch 5 Batch 90/162] avg loss 0.0102351, throughput 9.39766K wps
[Epoch 5 Batch 120/162] avg loss 0.00998377, throughput 9.39601K wps
[Epoch 5 Batch 150/162] avg loss 0.00978893, throughput 9.24047K wps
Begin Testing...
[Epoch 5] train avg loss 0.0101777, dev acc 0.8667, dev avg loss 0.488422, throughput 9.40287K wps
Observed Improvement.
Begin Testing...
[Epoch 6 Batch 30/162] avg loss 0.00962471, throughput 9.50385K wps
[Epoch 6 Batch 60/162] avg loss 0.00960506, throughput 9.22368K wps
[Epoch 6 Batch 90/162] avg loss 0.00940988, throughput 9.52823K wps
[Epoch 6 Batch 120/162] avg loss 0.00923151, throughput 9.32226K wps
[Epoch 6 Batch 150/162] avg loss 0.00923229, throughput 9.39732K wps
Begin Testing...
[Epoch 6] train avg loss 0.00941625, dev acc 0.8600, dev avg loss 0.453629, throughput 9.38266K wps
[Epoch 7 Batch 30/162] avg loss 0.00911647, throughput 9.38688K wps
[Epoch 7 Batch 60/162] avg loss 0.00882835, throughput 9.35864K wps
[Epoch 7 Batch 90/162] avg loss 0.00889105, throughput 9.25503K wps
[Epoch 7 Batch 120/162] avg loss 0.00855239, throughput 9.38802K wps
[Epoch 7 Batch 150/162] avg loss 0.00855025, throughput 9.38374K wps
Begin Testing...
[Epoch 7] train avg loss 0.00878858, dev acc 0.8667, dev avg loss 0.423234, throughput 9.33794K wps
Observed Improvement.
Begin Testing...
[Epoch 8 Batch 30/162] avg loss 0.00859895, throughput 9.50349K wps
[Epoch 8 Batch 60/162] avg loss 0.00849058, throughput 9.48742K wps
[Epoch 8 Batch 90/162] avg loss 0.00805606, throughput 9.36248K wps
[Epoch 8 Batch 120/162] avg loss 0.00792628, throughput 9.46156K wps
[Epoch 8 Batch 150/162] avg loss 0.00799315, throughput 9.52324K wps
Begin Testing...
[Epoch 8] train avg loss 0.00823134, dev acc 0.8722, dev avg loss 0.398932, throughput 9.46574K wps
Observed Improvement.
Begin Testing...
[Epoch 9 Batch 30/162] avg loss 0.00803663, throughput 9.69821K wps
[Epoch 9 Batch 60/162] avg loss 0.00798854, throughput 9.54757K wps
[Epoch 9 Batch 90/162] avg loss 0.00790172, throughput 9.50315K wps
[Epoch 9 Batch 120/162] avg loss 0.00761154, throughput 9.29564K wps
[Epoch 9 Batch 150/162] avg loss 0.0076133, throughput 9.59119K wps
Begin Testing...
[Epoch 9] train avg loss 0.00779802, dev acc 0.8767, dev avg loss 0.377216, throughput 9.52225K wps
Observed Improvement.
Begin Testing...
[Epoch 10 Batch 30/162] avg loss 0.00754552, throughput 9.46628K wps
[Epoch 10 Batch 60/162] avg loss 0.00787831, throughput 9.51402K wps
[Epoch 10 Batch 90/162] avg loss 0.00731, throughput 9.3669K wps
[Epoch 10 Batch 120/162] avg loss 0.00692783, throughput 9.52606K wps
[Epoch 10 Batch 150/162] avg loss 0.00738703, throughput 9.33564K wps
Begin Testing...
[Epoch 10] train avg loss 0.00739839, dev acc 0.8811, dev avg loss 0.360572, throughput 9.41177K wps
Observed Improvement.
Begin Testing...
[Epoch 11 Batch 30/162] avg loss 0.00698418, throughput 9.47101K wps
[Epoch 11 Batch 60/162] avg loss 0.00701242, throughput 9.54302K wps
[Epoch 11 Batch 90/162] avg loss 0.00728473, throughput 9.3188K wps
[Epoch 11 Batch 120/162] avg loss 0.00717908, throughput 9.35873K wps
[Epoch 11 Batch 150/162] avg loss 0.00703294, throughput 9.44976K wps
Begin Testing...
[Epoch 11] train avg loss 0.00709986, dev acc 0.8844, dev avg loss 0.346173, throughput 9.42604K wps
Observed Improvement.
Begin Testing...
[Epoch 12 Batch 30/162] avg loss 0.00696089, throughput 9.63775K wps
[Epoch 12 Batch 60/162] avg loss 0.00690618, throughput 9.33674K wps
[Epoch 12 Batch 90/162] avg loss 0.00688934, throughput 9.55114K wps
[Epoch 12 Batch 120/162] avg loss 0.006897, throughput 9.55849K wps
[Epoch 12 Batch 150/162] avg loss 0.00638437, throughput 9.29649K wps
Begin Testing...
[Epoch 12] train avg loss 0.00680112, dev acc 0.8856, dev avg loss 0.334045, throughput 9.4697K wps
Observed Improvement.
Begin Testing...
[Epoch 13 Batch 30/162] avg loss 0.00664919, throughput 9.41764K wps
[Epoch 13 Batch 60/162] avg loss 0.00679228, throughput 9.43987K wps
[Epoch 13 Batch 90/162] avg loss 0.00633877, throughput 9.26264K wps
[Epoch 13 Batch 120/162] avg loss 0.00683719, throughput 9.44104K wps
[Epoch 13 Batch 150/162] avg loss 0.00643042, throughput 9.16669K wps
Begin Testing...
[Epoch 13] train avg loss 0.00661544, dev acc 0.8867, dev avg loss 0.324257, throughput 9.34391K wps
Observed Improvement.
Begin Testing...
[Epoch 14 Batch 30/162] avg loss 0.00618894, throughput 9.48752K wps
[Epoch 14 Batch 60/162] avg loss 0.0063971, throughput 9.43725K wps
[Epoch 14 Batch 90/162] avg loss 0.00644065, throughput 9.27515K wps
[Epoch 14 Batch 120/162] avg loss 0.00612615, throughput 9.38826K wps
[Epoch 14 Batch 150/162] avg loss 0.00650961, throughput 9.22306K wps
Begin Testing...
[Epoch 14] train avg loss 0.00634747, dev acc 0.8900, dev avg loss 0.315319, throughput 9.36034K wps
Observed Improvement.
Begin Testing...
[Epoch 15 Batch 30/162] avg loss 0.0059935, throughput 9.46308K wps
[Epoch 15 Batch 60/162] avg loss 0.00645696, throughput 9.29715K wps
[Epoch 15 Batch 90/162] avg loss 0.00609282, throughput 9.34504K wps
[Epoch 15 Batch 120/162] avg loss 0.00612239, throughput 9.34152K wps
[Epoch 15 Batch 150/162] avg loss 0.00620774, throughput 9.44577K wps
Begin Testing...
[Epoch 15] train avg loss 0.00616604, dev acc 0.8878, dev avg loss 0.310088, throughput 9.3773K wps
[Epoch 16 Batch 30/162] avg loss 0.00619185, throughput 9.66085K wps
[Epoch 16 Batch 60/162] avg loss 0.00597811, throughput 9.27337K wps
[Epoch 16 Batch 90/162] avg loss 0.00608168, throughput 9.34676K wps
[Epoch 16 Batch 120/162] avg loss 0.0060234, throughput 9.29497K wps
[Epoch 16 Batch 150/162] avg loss 0.00622411, throughput 9.30938K wps
Begin Testing...
[Epoch 16] train avg loss 0.00607492, dev acc 0.8911, dev avg loss 0.301734, throughput 9.36725K wps
Observed Improvement.
Begin Testing...
[Epoch 17 Batch 30/162] avg loss 0.00532383, throughput 9.71143K wps
[Epoch 17 Batch 60/162] avg loss 0.00631796, throughput 9.25044K wps
[Epoch 17 Batch 90/162] avg loss 0.00572296, throughput 9.53512K wps
[Epoch 17 Batch 120/162] avg loss 0.0059069, throughput 9.46149K wps
[Epoch 17 Batch 150/162] avg loss 0.00613275, throughput 9.26207K wps
Begin Testing...
[Epoch 17] train avg loss 0.00588974, dev acc 0.8978, dev avg loss 0.29529, throughput 9.43012K wps
Observed Improvement.
Begin Testing...
[Epoch 18 Batch 30/162] avg loss 0.0063085, throughput 9.71707K wps
[Epoch 18 Batch 60/162] avg loss 0.00548676, throughput 9.34587K wps
[Epoch 18 Batch 90/162] avg loss 0.00558735, throughput 9.44843K wps
[Epoch 18 Batch 120/162] avg loss 0.00593278, throughput 9.43432K wps
[Epoch 18 Batch 150/162] avg loss 0.00505024, throughput 9.37806K wps
Begin Testing...
[Epoch 18] train avg loss 0.00571916, dev acc 0.8911, dev avg loss 0.289942, throughput 9.46878K wps
[Epoch 19 Batch 30/162] avg loss 0.00602063, throughput 9.39825K wps
[Epoch 19 Batch 60/162] avg loss 0.00534838, throughput 9.27807K wps
[Epoch 19 Batch 90/162] avg loss 0.00539531, throughput 9.4522K wps
[Epoch 19 Batch 120/162] avg loss 0.00563241, throughput 9.57426K wps
[Epoch 19 Batch 150/162] avg loss 0.00549891, throughput 9.39374K wps
Begin Testing...
[Epoch 19] train avg loss 0.00558118, dev acc 0.8944, dev avg loss 0.287854, throughput 9.41553K wps
[Epoch 20 Batch 30/162] avg loss 0.00539524, throughput 9.51736K wps
[Epoch 20 Batch 60/162] avg loss 0.00532709, throughput 9.32092K wps
[Epoch 20 Batch 90/162] avg loss 0.00552385, throughput 9.39393K wps
[Epoch 20 Batch 120/162] avg loss 0.00528053, throughput 9.40094K wps
[Epoch 20 Batch 150/162] avg loss 0.00551267, throughput 9.52752K wps
Begin Testing...
[Epoch 20] train avg loss 0.00543856, dev acc 0.8978, dev avg loss 0.280875, throughput 9.43418K wps
Observed Improvement.
Begin Testing...
[Epoch 21 Batch 30/162] avg loss 0.00546986, throughput 9.67087K wps
[Epoch 21 Batch 60/162] avg loss 0.00544896, throughput 9.24313K wps
[Epoch 21 Batch 90/162] avg loss 0.00513897, throughput 9.25813K wps
[Epoch 21 Batch 120/162] avg loss 0.0058843, throughput 9.52954K wps
[Epoch 21 Batch 150/162] avg loss 0.00520905, throughput 9.31865K wps
Begin Testing...
[Epoch 21] train avg loss 0.00543021, dev acc 0.9022, dev avg loss 0.277326, throughput 9.38967K wps
Observed Improvement.
Begin Testing...
[Epoch 22 Batch 30/162] avg loss 0.0052814, throughput 9.66198K wps
[Epoch 22 Batch 60/162] avg loss 0.00526309, throughput 9.30709K wps
[Epoch 22 Batch 90/162] avg loss 0.00492756, throughput 9.33057K wps
[Epoch 22 Batch 120/162] avg loss 0.00540431, throughput 9.36646K wps
[Epoch 22 Batch 150/162] avg loss 0.0049414, throughput 9.31729K wps
Begin Testing...
[Epoch 22] train avg loss 0.00519418, dev acc 0.8989, dev avg loss 0.2743, throughput 9.38903K wps
[Epoch 23 Batch 30/162] avg loss 0.00495773, throughput 9.58527K wps
[Epoch 23 Batch 60/162] avg loss 0.00509811, throughput 9.28268K wps
[Epoch 23 Batch 90/162] avg loss 0.00522338, throughput 9.28067K wps
[Epoch 23 Batch 120/162] avg loss 0.00526962, throughput 9.36207K wps
[Epoch 23 Batch 150/162] avg loss 0.00513892, throughput 9.34404K wps
Begin Testing...
[Epoch 23] train avg loss 0.00515225, dev acc 0.9000, dev avg loss 0.271179, throughput 9.38602K wps
[Epoch 24 Batch 30/162] avg loss 0.00527121, throughput 9.49717K wps
[Epoch 24 Batch 60/162] avg loss 0.00527566, throughput 9.24494K wps
[Epoch 24 Batch 90/162] avg loss 0.00490714, throughput 9.44236K wps
[Epoch 24 Batch 120/162] avg loss 0.00485533, throughput 9.46745K wps
[Epoch 24 Batch 150/162] avg loss 0.0054804, throughput 9.37747K wps
Begin Testing...
[Epoch 24] train avg loss 0.00509567, dev acc 0.9022, dev avg loss 0.266359, throughput 9.39867K wps
Observed Improvement.
Begin Testing...
[Epoch 25 Batch 30/162] avg loss 0.00514784, throughput 9.56632K wps
[Epoch 25 Batch 60/162] avg loss 0.00523556, throughput 9.42523K wps
[Epoch 25 Batch 90/162] avg loss 0.00474505, throughput 9.22163K wps
[Epoch 25 Batch 120/162] avg loss 0.0051623, throughput 9.41804K wps
[Epoch 25 Batch 150/162] avg loss 0.0046691, throughput 9.29541K wps
Begin Testing...
[Epoch 25] train avg loss 0.00495063, dev acc 0.9011, dev avg loss 0.264198, throughput 9.37651K wps
[Epoch 26 Batch 30/162] avg loss 0.00481072, throughput 9.58693K wps
[Epoch 26 Batch 60/162] avg loss 0.00515807, throughput 9.33722K wps
[Epoch 26 Batch 90/162] avg loss 0.00489274, throughput 9.53344K wps
[Epoch 26 Batch 120/162] avg loss 0.00501662, throughput 9.52266K wps
[Epoch 26 Batch 150/162] avg loss 0.00482542, throughput 9.29128K wps
Begin Testing...
[Epoch 26] train avg loss 0.00494098, dev acc 0.9033, dev avg loss 0.262455, throughput 9.43532K wps
Observed Improvement.
Begin Testing...
[Epoch 27 Batch 30/162] avg loss 0.00505959, throughput 9.53086K wps
[Epoch 27 Batch 60/162] avg loss 0.00461797, throughput 9.24031K wps
[Epoch 27 Batch 90/162] avg loss 0.00468589, throughput 9.47949K wps
[Epoch 27 Batch 120/162] avg loss 0.00475254, throughput 9.49519K wps
[Epoch 27 Batch 150/162] avg loss 0.0049145, throughput 9.49342K wps
Begin Testing...
[Epoch 27] train avg loss 0.00478827, dev acc 0.9044, dev avg loss 0.257602, throughput 9.42375K wps
Observed Improvement.
Begin Testing...
[Epoch 28 Batch 30/162] avg loss 0.00498717, throughput 9.4896K wps
[Epoch 28 Batch 60/162] avg loss 0.0044837, throughput 9.29405K wps
[Epoch 28 Batch 90/162] avg loss 0.0047602, throughput 9.22584K wps
[Epoch 28 Batch 120/162] avg loss 0.0044738, throughput 9.46349K wps
[Epoch 28 Batch 150/162] avg loss 0.00476912, throughput 9.28657K wps
Begin Testing...
[Epoch 28] train avg loss 0.00471349, dev acc 0.9044, dev avg loss 0.25515, throughput 9.33295K wps
Observed Improvement.
Begin Testing...
[Epoch 29 Batch 30/162] avg loss 0.00475557, throughput 9.5363K wps
[Epoch 29 Batch 60/162] avg loss 0.00439865, throughput 9.46315K wps
[Epoch 29 Batch 90/162] avg loss 0.00452953, throughput 9.27957K wps
[Epoch 29 Batch 120/162] avg loss 0.00509885, throughput 9.47118K wps
[Epoch 29 Batch 150/162] avg loss 0.00460024, throughput 9.34474K wps
Begin Testing...
[Epoch 29] train avg loss 0.00467039, dev acc 0.9078, dev avg loss 0.25361, throughput 9.41418K wps
Observed Improvement.
Begin Testing...
[Epoch 30 Batch 30/162] avg loss 0.00423084, throughput 9.57885K wps
[Epoch 30 Batch 60/162] avg loss 0.00505471, throughput 9.35229K wps
[Epoch 30 Batch 90/162] avg loss 0.00457207, throughput 9.30255K wps
[Epoch 30 Batch 120/162] avg loss 0.00457113, throughput 9.22787K wps
[Epoch 30 Batch 150/162] avg loss 0.00462759, throughput 9.35478K wps
Begin Testing...
[Epoch 30] train avg loss 0.00458903, dev acc 0.9089, dev avg loss 0.250859, throughput 9.34796K wps
Observed Improvement.
Begin Testing...
[Epoch 31 Batch 30/162] avg loss 0.00457845, throughput 9.7521K wps
[Epoch 31 Batch 60/162] avg loss 0.00443692, throughput 9.44723K wps
[Epoch 31 Batch 90/162] avg loss 0.0042473, throughput 9.49401K wps
[Epoch 31 Batch 120/162] avg loss 0.00449582, throughput 9.37604K wps
[Epoch 31 Batch 150/162] avg loss 0.00490598, throughput 9.27065K wps
Begin Testing...
[Epoch 31] train avg loss 0.00448793, dev acc 0.9044, dev avg loss 0.250111, throughput 9.47686K wps
[Epoch 32 Batch 30/162] avg loss 0.00438358, throughput 9.63072K wps
[Epoch 32 Batch 60/162] avg loss 0.00459197, throughput 9.30356K wps
[Epoch 32 Batch 90/162] avg loss 0.00415089, throughput 9.39326K wps
[Epoch 32 Batch 120/162] avg loss 0.00419427, throughput 9.38156K wps
[Epoch 32 Batch 150/162] avg loss 0.00442435, throughput 9.38964K wps
Begin Testing...
[Epoch 32] train avg loss 0.00434168, dev acc 0.9067, dev avg loss 0.247303, throughput 9.4318K wps
[Epoch 33 Batch 30/162] avg loss 0.00438759, throughput 9.72087K wps
[Epoch 33 Batch 60/162] avg loss 0.00426632, throughput 9.18907K wps
[Epoch 33 Batch 90/162] avg loss 0.00416034, throughput 9.44921K wps
[Epoch 33 Batch 120/162] avg loss 0.00445434, throughput 9.52463K wps
[Epoch 33 Batch 150/162] avg loss 0.00432861, throughput 9.50081K wps
Begin Testing...
[Epoch 33] train avg loss 0.00434186, dev acc 0.9089, dev avg loss 0.244891, throughput 9.46679K wps
Observed Improvement.
Begin Testing...
[Epoch 34 Batch 30/162] avg loss 0.00439183, throughput 9.5572K wps
[Epoch 34 Batch 60/162] avg loss 0.00444066, throughput 9.33924K wps
[Epoch 34 Batch 90/162] avg loss 0.00433305, throughput 9.49454K wps
[Epoch 34 Batch 120/162] avg loss 0.00399763, throughput 9.40104K wps
[Epoch 34 Batch 150/162] avg loss 0.0041683, throughput 9.36359K wps
Begin Testing...
[Epoch 34] train avg loss 0.0042618, dev acc 0.9122, dev avg loss 0.2424, throughput 9.41809K wps
Observed Improvement.
Begin Testing...
[Epoch 35 Batch 30/162] avg loss 0.00396095, throughput 9.53064K wps
[Epoch 35 Batch 60/162] avg loss 0.00422483, throughput 9.40817K wps
[Epoch 35 Batch 90/162] avg loss 0.0042559, throughput 9.3102K wps
[Epoch 35 Batch 120/162] avg loss 0.00440011, throughput 9.30157K wps
[Epoch 35 Batch 150/162] avg loss 0.00414947, throughput 9.4243K wps
Begin Testing...
[Epoch 35] train avg loss 0.00420051, dev acc 0.9078, dev avg loss 0.240934, throughput 9.3835K wps
[Epoch 36 Batch 30/162] avg loss 0.00419109, throughput 9.68718K wps
[Epoch 36 Batch 60/162] avg loss 0.00415082, throughput 9.38639K wps
[Epoch 36 Batch 90/162] avg loss 0.00407759, throughput 9.31439K wps
[Epoch 36 Batch 120/162] avg loss 0.00443489, throughput 9.37311K wps
[Epoch 36 Batch 150/162] avg loss 0.00407228, throughput 9.56743K wps
Begin Testing...
[Epoch 36] train avg loss 0.00418269, dev acc 0.9067, dev avg loss 0.240895, throughput 9.44762K wps
[Epoch 37 Batch 30/162] avg loss 0.00423626, throughput 9.45649K wps
[Epoch 37 Batch 60/162] avg loss 0.00430995, throughput 9.37398K wps
[Epoch 37 Batch 90/162] avg loss 0.00390362, throughput 9.29196K wps
[Epoch 37 Batch 120/162] avg loss 0.0039908, throughput 9.42386K wps
[Epoch 37 Batch 150/162] avg loss 0.00387184, throughput 9.41925K wps
Begin Testing...
[Epoch 37] train avg loss 0.00401795, dev acc 0.9067, dev avg loss 0.240547, throughput 9.38779K wps
[Epoch 38 Batch 30/162] avg loss 0.0036163, throughput 9.47012K wps
[Epoch 38 Batch 60/162] avg loss 0.00385092, throughput 9.31015K wps
[Epoch 38 Batch 90/162] avg loss 0.0038551, throughput 9.29771K wps
[Epoch 38 Batch 120/162] avg loss 0.00418608, throughput 9.29542K wps
[Epoch 38 Batch 150/162] avg loss 0.00415054, throughput 9.31799K wps
Begin Testing...
[Epoch 38] train avg loss 0.0039579, dev acc 0.9133, dev avg loss 0.235807, throughput 9.3362K wps
Observed Improvement.
Begin Testing...
[Epoch 39 Batch 30/162] avg loss 0.00393028, throughput 9.55193K wps
[Epoch 39 Batch 60/162] avg loss 0.00413509, throughput 9.34994K wps
[Epoch 39 Batch 90/162] avg loss 0.00382916, throughput 9.38689K wps
[Epoch 39 Batch 120/162] avg loss 0.00398161, throughput 9.28729K wps
[Epoch 39 Batch 150/162] avg loss 0.0038787, throughput 9.40004K wps
Begin Testing...
[Epoch 39] train avg loss 0.00393537, dev acc 0.9100, dev avg loss 0.238395, throughput 9.40121K wps
[Epoch 40 Batch 30/162] avg loss 0.00403528, throughput 9.53022K wps
[Epoch 40 Batch 60/162] avg loss 0.00352629, throughput 9.40001K wps
[Epoch 40 Batch 90/162] avg loss 0.00368269, throughput 9.27869K wps
[Epoch 40 Batch 120/162] avg loss 0.00396326, throughput 9.43897K wps
[Epoch 40 Batch 150/162] avg loss 0.00412601, throughput 9.28515K wps
Begin Testing...
[Epoch 40] train avg loss 0.00389714, dev acc 0.9044, dev avg loss 0.233922, throughput 9.39103K wps
[Epoch 41 Batch 30/162] avg loss 0.00374791, throughput 9.47088K wps
[Epoch 41 Batch 60/162] avg loss 0.00383387, throughput 9.35316K wps
[Epoch 41 Batch 90/162] avg loss 0.00386738, throughput 9.28109K wps
[Epoch 41 Batch 120/162] avg loss 0.00399312, throughput 9.53913K wps
[Epoch 41 Batch 150/162] avg loss 0.00365472, throughput 9.3243K wps
Begin Testing...
[Epoch 41] train avg loss 0.00382076, dev acc 0.9044, dev avg loss 0.233103, throughput 9.39679K wps
[Epoch 42 Batch 30/162] avg loss 0.00369949, throughput 9.40587K wps
[Epoch 42 Batch 60/162] avg loss 0.00348854, throughput 9.36362K wps
[Epoch 42 Batch 90/162] avg loss 0.00353717, throughput 9.33468K wps
[Epoch 42 Batch 120/162] avg loss 0.00380575, throughput 9.375K wps
[Epoch 42 Batch 150/162] avg loss 0.00376093, throughput 9.35851K wps
Begin Testing...
[Epoch 42] train avg loss 0.00367571, dev acc 0.9089, dev avg loss 0.236939, throughput 9.17275K wps
[Epoch 43 Batch 30/162] avg loss 0.00370348, throughput 9.65191K wps
[Epoch 43 Batch 60/162] avg loss 0.00404026, throughput 9.25523K wps
[Epoch 43 Batch 90/162] avg loss 0.00337673, throughput 9.46245K wps
[Epoch 43 Batch 120/162] avg loss 0.0038275, throughput 9.33324K wps
[Epoch 43 Batch 150/162] avg loss 0.00347124, throughput 9.53556K wps
Begin Testing...
[Epoch 43] train avg loss 0.00368728, dev acc 0.9100, dev avg loss 0.232106, throughput 9.4557K wps
[Epoch 44 Batch 30/162] avg loss 0.0032868, throughput 9.41737K wps
[Epoch 44 Batch 60/162] avg loss 0.00362133, throughput 9.4609K wps
[Epoch 44 Batch 90/162] avg loss 0.00386706, throughput 9.47567K wps
[Epoch 44 Batch 120/162] avg loss 0.00421878, throughput 9.26623K wps
[Epoch 44 Batch 150/162] avg loss 0.00317487, throughput 9.39186K wps
Begin Testing...
[Epoch 44] train avg loss 0.00365101, dev acc 0.9078, dev avg loss 0.228685, throughput 9.3968K wps
[Epoch 45 Batch 30/162] avg loss 0.00391822, throughput 9.68156K wps
[Epoch 45 Batch 60/162] avg loss 0.00368644, throughput 9.36446K wps
[Epoch 45 Batch 90/162] avg loss 0.00347633, throughput 9.37991K wps
[Epoch 45 Batch 120/162] avg loss 0.00338192, throughput 9.28825K wps
[Epoch 45 Batch 150/162] avg loss 0.00379859, throughput 9.52351K wps
Begin Testing...
[Epoch 45] train avg loss 0.00360312, dev acc 0.9122, dev avg loss 0.226747, throughput 9.43741K wps
[Epoch 46 Batch 30/162] avg loss 0.00337851, throughput 9.52281K wps
[Epoch 46 Batch 60/162] avg loss 0.00353297, throughput 9.42704K wps
[Epoch 46 Batch 90/162] avg loss 0.00361803, throughput 9.23996K wps
[Epoch 46 Batch 120/162] avg loss 0.00373974, throughput 9.41785K wps
[Epoch 46 Batch 150/162] avg loss 0.00347415, throughput 9.30365K wps
Begin Testing...
[Epoch 46] train avg loss 0.003556, dev acc 0.9089, dev avg loss 0.226497, throughput 9.3744K wps
[Epoch 47 Batch 30/162] avg loss 0.00317813, throughput 9.68463K wps
[Epoch 47 Batch 60/162] avg loss 0.00360974, throughput 9.35363K wps
[Epoch 47 Batch 90/162] avg loss 0.00345975, throughput 9.39537K wps
[Epoch 47 Batch 120/162] avg loss 0.00384116, throughput 9.39674K wps
[Epoch 47 Batch 150/162] avg loss 0.00356022, throughput 9.32371K wps
Begin Testing...
[Epoch 47] train avg loss 0.0035078, dev acc 0.9111, dev avg loss 0.22506, throughput 9.42128K wps
[Epoch 48 Batch 30/162] avg loss 0.00318078, throughput 9.50623K wps
[Epoch 48 Batch 60/162] avg loss 0.00334933, throughput 9.43885K wps
[Epoch 48 Batch 90/162] avg loss 0.00356736, throughput 9.2695K wps
[Epoch 48 Batch 120/162] avg loss 0.00351245, throughput 9.29588K wps
[Epoch 48 Batch 150/162] avg loss 0.00366584, throughput 9.46076K wps
Begin Testing...
[Epoch 48] train avg loss 0.00347328, dev acc 0.9100, dev avg loss 0.224737, throughput 9.39616K wps
[Epoch 49 Batch 30/162] avg loss 0.00315785, throughput 9.3593K wps
[Epoch 49 Batch 60/162] avg loss 0.00340777, throughput 9.42631K wps
[Epoch 49 Batch 90/162] avg loss 0.0033833, throughput 9.44367K wps
[Epoch 49 Batch 120/162] avg loss 0.00370163, throughput 9.3962K wps
[Epoch 49 Batch 150/162] avg loss 0.00358835, throughput 9.29303K wps
Begin Testing...
[Epoch 49] train avg loss 0.00343552, dev acc 0.9100, dev avg loss 0.22304, throughput 9.38923K wps
[Epoch 50 Batch 30/162] avg loss 0.00358856, throughput 9.53773K wps
[Epoch 50 Batch 60/162] avg loss 0.00340707, throughput 9.54984K wps
[Epoch 50 Batch 90/162] avg loss 0.0032981, throughput 9.35108K wps
[Epoch 50 Batch 120/162] avg loss 0.00342001, throughput 9.50797K wps
[Epoch 50 Batch 150/162] avg loss 0.00315331, throughput 9.31105K wps
Begin Testing...
[Epoch 50] train avg loss 0.00336502, dev acc 0.9100, dev avg loss 0.222865, throughput 9.43515K wps
[Epoch 51 Batch 30/162] avg loss 0.00344844, throughput 9.54461K wps
[Epoch 51 Batch 60/162] avg loss 0.00366597, throughput 9.49713K wps
[Epoch 51 Batch 90/162] avg loss 0.00339648, throughput 9.30367K wps
[Epoch 51 Batch 120/162] avg loss 0.00334348, throughput 9.50288K wps
[Epoch 51 Batch 150/162] avg loss 0.00281499, throughput 9.3962K wps
Begin Testing...
[Epoch 51] train avg loss 0.00332766, dev acc 0.9144, dev avg loss 0.221063, throughput 9.42948K wps
Observed Improvement.
Begin Testing...
[Epoch 52 Batch 30/162] avg loss 0.00330193, throughput 9.59202K wps
[Epoch 52 Batch 60/162] avg loss 0.00308882, throughput 9.34199K wps
[Epoch 52 Batch 90/162] avg loss 0.00306215, throughput 9.32314K wps
[Epoch 52 Batch 120/162] avg loss 0.00336281, throughput 9.34332K wps
[Epoch 52 Batch 150/162] avg loss 0.00334005, throughput 9.27212K wps
Begin Testing...
[Epoch 52] train avg loss 0.00323267, dev acc 0.9122, dev avg loss 0.220731, throughput 9.3833K wps
[Epoch 53 Batch 30/162] avg loss 0.00288676, throughput 9.59984K wps
[Epoch 53 Batch 60/162] avg loss 0.00316525, throughput 9.32294K wps
[Epoch 53 Batch 90/162] avg loss 0.00340108, throughput 9.57318K wps
[Epoch 53 Batch 120/162] avg loss 0.00304465, throughput 9.38061K wps
[Epoch 53 Batch 150/162] avg loss 0.00329722, throughput 9.33921K wps
Begin Testing...
[Epoch 53] train avg loss 0.00317741, dev acc 0.9156, dev avg loss 0.219419, throughput 9.43784K wps
Observed Improvement.
Begin Testing...
[Epoch 54 Batch 30/162] avg loss 0.00352498, throughput 9.59891K wps
[Epoch 54 Batch 60/162] avg loss 0.0031086, throughput 9.22745K wps
[Epoch 54 Batch 90/162] avg loss 0.00299515, throughput 9.28343K wps
[Epoch 54 Batch 120/162] avg loss 0.00326317, throughput 9.53516K wps
[Epoch 54 Batch 150/162] avg loss 0.00288429, throughput 9.26826K wps
Begin Testing...
[Epoch 54] train avg loss 0.00315196, dev acc 0.9133, dev avg loss 0.222463, throughput 9.38464K wps
[Epoch 55 Batch 30/162] avg loss 0.00313675, throughput 9.60211K wps
[Epoch 55 Batch 60/162] avg loss 0.00327937, throughput 9.37876K wps
[Epoch 55 Batch 90/162] avg loss 0.00304455, throughput 9.43512K wps
[Epoch 55 Batch 120/162] avg loss 0.00310149, throughput 9.37944K wps
[Epoch 55 Batch 150/162] avg loss 0.00310378, throughput 9.56915K wps
Begin Testing...
[Epoch 55] train avg loss 0.00313147, dev acc 0.9156, dev avg loss 0.217416, throughput 9.45598K wps
Observed Improvement.
Begin Testing...
[Epoch 56 Batch 30/162] avg loss 0.00308663, throughput 9.63825K wps
[Epoch 56 Batch 60/162] avg loss 0.0031728, throughput 9.46906K wps
[Epoch 56 Batch 90/162] avg loss 0.003047, throughput 9.50369K wps
[Epoch 56 Batch 120/162] avg loss 0.00320278, throughput 9.30888K wps
[Epoch 56 Batch 150/162] avg loss 0.00311601, throughput 9.34739K wps
Begin Testing...
[Epoch 56] train avg loss 0.00311877, dev acc 0.9133, dev avg loss 0.217659, throughput 9.46213K wps
[Epoch 57 Batch 30/162] avg loss 0.00314119, throughput 9.66441K wps
[Epoch 57 Batch 60/162] avg loss 0.0030945, throughput 9.48607K wps
[Epoch 57 Batch 90/162] avg loss 0.00284895, throughput 9.44406K wps
[Epoch 57 Batch 120/162] avg loss 0.00312214, throughput 9.31305K wps
[Epoch 57 Batch 150/162] avg loss 0.00320701, throughput 9.33659K wps
Begin Testing...
[Epoch 57] train avg loss 0.00304471, dev acc 0.9133, dev avg loss 0.21656, throughput 9.43125K wps
[Epoch 58 Batch 30/162] avg loss 0.00289749, throughput 9.5139K wps
[Epoch 58 Batch 60/162] avg loss 0.0030572, throughput 9.2619K wps
[Epoch 58 Batch 90/162] avg loss 0.00319532, throughput 9.17808K wps
[Epoch 58 Batch 120/162] avg loss 0.00298137, throughput 9.35268K wps
[Epoch 58 Batch 150/162] avg loss 0.00273173, throughput 9.3635K wps
Begin Testing...
[Epoch 58] train avg loss 0.0029873, dev acc 0.9133, dev avg loss 0.215808, throughput 9.34364K wps
[Epoch 59 Batch 30/162] avg loss 0.00288885, throughput 9.39518K wps
[Epoch 59 Batch 60/162] avg loss 0.00281376, throughput 9.52286K wps
[Epoch 59 Batch 90/162] avg loss 0.0028003, throughput 9.22134K wps
[Epoch 59 Batch 120/162] avg loss 0.00306146, throughput 9.36954K wps
[Epoch 59 Batch 150/162] avg loss 0.00307794, throughput 9.22466K wps
Begin Testing...
[Epoch 59] train avg loss 0.00294132, dev acc 0.9178, dev avg loss 0.214598, throughput 9.33721K wps
Observed Improvement.
Begin Testing...
[Epoch 60 Batch 30/162] avg loss 0.00281642, throughput 9.33315K wps
[Epoch 60 Batch 60/162] avg loss 0.0027846, throughput 9.3658K wps
[Epoch 60 Batch 90/162] avg loss 0.00288393, throughput 9.3219K wps
[Epoch 60 Batch 120/162] avg loss 0.00260388, throughput 9.1854K wps
[Epoch 60 Batch 150/162] avg loss 0.00311961, throughput 9.28927K wps
Begin Testing...
[Epoch 60] train avg loss 0.00284866, dev acc 0.9178, dev avg loss 0.213585, throughput 9.29819K wps
Observed Improvement.
Begin Testing...
[Epoch 61 Batch 30/162] avg loss 0.0029454, throughput 9.68048K wps
[Epoch 61 Batch 60/162] avg loss 0.0029491, throughput 9.42217K wps
[Epoch 61 Batch 90/162] avg loss 0.00293569, throughput 9.49673K wps
[Epoch 61 Batch 120/162] avg loss 0.00280779, throughput 9.34427K wps
[Epoch 61 Batch 150/162] avg loss 0.00286973, throughput 9.38538K wps
Begin Testing...
[Epoch 61] train avg loss 0.00289454, dev acc 0.9156, dev avg loss 0.214966, throughput 9.45196K wps
[Epoch 62 Batch 30/162] avg loss 0.00307331, throughput 9.67633K wps
[Epoch 62 Batch 60/162] avg loss 0.00260981, throughput 9.38458K wps
[Epoch 62 Batch 90/162] avg loss 0.00287654, throughput 9.23112K wps
[Epoch 62 Batch 120/162] avg loss 0.00253145, throughput 9.33879K wps
[Epoch 62 Batch 150/162] avg loss 0.00286575, throughput 9.36939K wps
Begin Testing...
[Epoch 62] train avg loss 0.00279087, dev acc 0.9178, dev avg loss 0.212548, throughput 9.40694K wps
Observed Improvement.
Begin Testing...
[Epoch 63 Batch 30/162] avg loss 0.00266319, throughput 9.49314K wps
[Epoch 63 Batch 60/162] avg loss 0.00299049, throughput 9.39792K wps
[Epoch 63 Batch 90/162] avg loss 0.0026538, throughput 9.2369K wps
[Epoch 63 Batch 120/162] avg loss 0.00294969, throughput 9.47272K wps
[Epoch 63 Batch 150/162] avg loss 0.00269789, throughput 9.30891K wps
Begin Testing...
[Epoch 63] train avg loss 0.00279171, dev acc 0.9144, dev avg loss 0.212288, throughput 9.37613K wps
[Epoch 64 Batch 30/162] avg loss 0.00279522, throughput 9.50945K wps
[Epoch 64 Batch 60/162] avg loss 0.00290671, throughput 9.3279K wps
[Epoch 64 Batch 90/162] avg loss 0.00285916, throughput 9.3628K wps
[Epoch 64 Batch 120/162] avg loss 0.00260125, throughput 9.44088K wps
[Epoch 64 Batch 150/162] avg loss 0.00258285, throughput 9.28697K wps
Begin Testing...
[Epoch 64] train avg loss 0.00275438, dev acc 0.9156, dev avg loss 0.211617, throughput 9.3893K wps
[Epoch 65 Batch 30/162] avg loss 0.00253041, throughput 9.54755K wps
[Epoch 65 Batch 60/162] avg loss 0.00283382, throughput 9.20343K wps
[Epoch 65 Batch 90/162] avg loss 0.00266519, throughput 9.44997K wps
[Epoch 65 Batch 120/162] avg loss 0.00278815, throughput 9.42333K wps
[Epoch 65 Batch 150/162] avg loss 0.0025763, throughput 9.27003K wps
Begin Testing...
[Epoch 65] train avg loss 0.00271699, dev acc 0.9156, dev avg loss 0.210817, throughput 9.39242K wps
[Epoch 66 Batch 30/162] avg loss 0.00255358, throughput 9.70619K wps
[Epoch 66 Batch 60/162] avg loss 0.00247572, throughput 9.47814K wps
[Epoch 66 Batch 90/162] avg loss 0.0026469, throughput 9.37641K wps
[Epoch 66 Batch 120/162] avg loss 0.002756, throughput 9.32051K wps
[Epoch 66 Batch 150/162] avg loss 0.00253668, throughput 9.39469K wps
Begin Testing...
[Epoch 66] train avg loss 0.00262527, dev acc 0.9144, dev avg loss 0.211205, throughput 9.4388K wps
[Epoch 67 Batch 30/162] avg loss 0.00264357, throughput 9.70131K wps
[Epoch 67 Batch 60/162] avg loss 0.00233298, throughput 9.39527K wps
[Epoch 67 Batch 90/162] avg loss 0.00265754, throughput 9.40917K wps
[Epoch 67 Batch 120/162] avg loss 0.00260187, throughput 9.34168K wps
[Epoch 67 Batch 150/162] avg loss 0.00255836, throughput 9.30964K wps
Begin Testing...
[Epoch 67] train avg loss 0.00256947, dev acc 0.9144, dev avg loss 0.210512, throughput 9.42315K wps
[Epoch 68 Batch 30/162] avg loss 0.00253492, throughput 9.50476K wps
[Epoch 68 Batch 60/162] avg loss 0.0028042, throughput 9.32187K wps
[Epoch 68 Batch 90/162] avg loss 0.00247748, throughput 9.53965K wps
[Epoch 68 Batch 120/162] avg loss 0.00262187, throughput 9.3745K wps
[Epoch 68 Batch 150/162] avg loss 0.00254095, throughput 9.39217K wps
Begin Testing...
[Epoch 68] train avg loss 0.00261883, dev acc 0.9200, dev avg loss 0.209441, throughput 9.40889K wps
Observed Improvement.
Begin Testing...
[Epoch 69 Batch 30/162] avg loss 0.0024845, throughput 9.75069K wps
[Epoch 69 Batch 60/162] avg loss 0.00260505, throughput 9.46023K wps
[Epoch 69 Batch 90/162] avg loss 0.00241479, throughput 9.37886K wps
[Epoch 69 Batch 120/162] avg loss 0.00273826, throughput 9.37237K wps
[Epoch 69 Batch 150/162] avg loss 0.00243985, throughput 9.35268K wps
Begin Testing...
[Epoch 69] train avg loss 0.00254853, dev acc 0.9178, dev avg loss 0.212331, throughput 9.44304K wps
[Epoch 70 Batch 30/162] avg loss 0.00260144, throughput 9.64096K wps
[Epoch 70 Batch 60/162] avg loss 0.00241532, throughput 9.24735K wps
[Epoch 70 Batch 90/162] avg loss 0.00272795, throughput 9.28185K wps
[Epoch 70 Batch 120/162] avg loss 0.00226338, throughput 9.56554K wps
[Epoch 70 Batch 150/162] avg loss 0.00242127, throughput 9.37051K wps
Begin Testing...
[Epoch 70] train avg loss 0.00247929, dev acc 0.9144, dev avg loss 0.209441, throughput 9.41115K wps
[Epoch 71 Batch 30/162] avg loss 0.00251706, throughput 9.63986K wps
[Epoch 71 Batch 60/162] avg loss 0.00224989, throughput 9.48325K wps
[Epoch 71 Batch 90/162] avg loss 0.00259469, throughput 9.32033K wps
[Epoch 71 Batch 120/162] avg loss 0.00230965, throughput 9.41586K wps
[Epoch 71 Batch 150/162] avg loss 0.00266199, throughput 9.39938K wps
Begin Testing...
[Epoch 71] train avg loss 0.00249198, dev acc 0.9167, dev avg loss 0.21003, throughput 9.44658K wps
[Epoch 72 Batch 30/162] avg loss 0.00274066, throughput 9.5333K wps
[Epoch 72 Batch 60/162] avg loss 0.00244282, throughput 9.40396K wps
[Epoch 72 Batch 90/162] avg loss 0.00218038, throughput 9.45801K wps
[Epoch 72 Batch 120/162] avg loss 0.00225924, throughput 9.35715K wps
[Epoch 72 Batch 150/162] avg loss 0.00260395, throughput 9.41767K wps
Begin Testing...
[Epoch 72] train avg loss 0.00244746, dev acc 0.9156, dev avg loss 0.207762, throughput 9.40931K wps
[Epoch 73 Batch 30/162] avg loss 0.00248187, throughput 9.66991K wps
[Epoch 73 Batch 60/162] avg loss 0.00233156, throughput 9.34824K wps
[Epoch 73 Batch 90/162] avg loss 0.00242208, throughput 9.51583K wps
[Epoch 73 Batch 120/162] avg loss 0.0025021, throughput 9.23414K wps
[Epoch 73 Batch 150/162] avg loss 0.00233627, throughput 9.3648K wps
Begin Testing...
[Epoch 73] train avg loss 0.00242023, dev acc 0.9156, dev avg loss 0.210773, throughput 9.4349K wps
[Epoch 74 Batch 30/162] avg loss 0.00240564, throughput 9.71273K wps
[Epoch 74 Batch 60/162] avg loss 0.00210959, throughput 9.31269K wps
[Epoch 74 Batch 90/162] avg loss 0.0022461, throughput 9.38421K wps
[Epoch 74 Batch 120/162] avg loss 0.00245088, throughput 9.49291K wps
[Epoch 74 Batch 150/162] avg loss 0.0023129, throughput 9.33802K wps
Begin Testing...
[Epoch 74] train avg loss 0.00232866, dev acc 0.9144, dev avg loss 0.207504, throughput 9.437K wps
[Epoch 75 Batch 30/162] avg loss 0.00219913, throughput 9.70508K wps
[Epoch 75 Batch 60/162] avg loss 0.00230746, throughput 9.35817K wps
[Epoch 75 Batch 90/162] avg loss 0.00229709, throughput 9.47615K wps
[Epoch 75 Batch 120/162] avg loss 0.00241229, throughput 9.38008K wps
[Epoch 75 Batch 150/162] avg loss 0.00231995, throughput 9.43205K wps
Begin Testing...
[Epoch 75] train avg loss 0.00233382, dev acc 0.9144, dev avg loss 0.2076, throughput 9.46329K wps
[Epoch 76 Batch 30/162] avg loss 0.00269728, throughput 9.52326K wps
[Epoch 76 Batch 60/162] avg loss 0.00233437, throughput 9.4011K wps
[Epoch 76 Batch 90/162] avg loss 0.00212156, throughput 9.50492K wps
[Epoch 76 Batch 120/162] avg loss 0.00245929, throughput 9.32548K wps
[Epoch 76 Batch 150/162] avg loss 0.00199076, throughput 9.44869K wps
Begin Testing...
[Epoch 76] train avg loss 0.00232555, dev acc 0.9156, dev avg loss 0.207058, throughput 9.45389K wps
[Epoch 77 Batch 30/162] avg loss 0.0023459, throughput 9.59286K wps
[Epoch 77 Batch 60/162] avg loss 0.00216819, throughput 9.47974K wps
[Epoch 77 Batch 90/162] avg loss 0.00225828, throughput 9.31954K wps
[Epoch 77 Batch 120/162] avg loss 0.00233275, throughput 9.44167K wps
[Epoch 77 Batch 150/162] avg loss 0.00243333, throughput 9.35746K wps
Begin Testing...
[Epoch 77] train avg loss 0.00230908, dev acc 0.9156, dev avg loss 0.206201, throughput 9.44723K wps
[Epoch 78 Batch 30/162] avg loss 0.00229113, throughput 9.68876K wps
[Epoch 78 Batch 60/162] avg loss 0.00196828, throughput 9.38466K wps
[Epoch 78 Batch 90/162] avg loss 0.00245172, throughput 9.32568K wps
[Epoch 78 Batch 120/162] avg loss 0.00239058, throughput 9.38389K wps
[Epoch 78 Batch 150/162] avg loss 0.00209169, throughput 9.46135K wps
Begin Testing...
[Epoch 78] train avg loss 0.00221657, dev acc 0.9156, dev avg loss 0.206973, throughput 9.44583K wps
[Epoch 79 Batch 30/162] avg loss 0.0022081, throughput 9.57182K wps
[Epoch 79 Batch 60/162] avg loss 0.00244722, throughput 9.41165K wps
[Epoch 79 Batch 90/162] avg loss 0.00207158, throughput 9.38219K wps
[Epoch 79 Batch 120/162] avg loss 0.00234296, throughput 9.37796K wps
[Epoch 79 Batch 150/162] avg loss 0.00220557, throughput 9.45205K wps
Begin Testing...
[Epoch 79] train avg loss 0.00223564, dev acc 0.9167, dev avg loss 0.205641, throughput 9.42664K wps
[Epoch 80 Batch 30/162] avg loss 0.00218495, throughput 9.52886K wps
[Epoch 80 Batch 60/162] avg loss 0.00238412, throughput 9.34769K wps
[Epoch 80 Batch 90/162] avg loss 0.00205708, throughput 9.27918K wps
[Epoch 80 Batch 120/162] avg loss 0.00216558, throughput 9.36922K wps
[Epoch 80 Batch 150/162] avg loss 0.00213911, throughput 9.28894K wps
Begin Testing...
[Epoch 80] train avg loss 0.00216977, dev acc 0.9167, dev avg loss 0.206409, throughput 9.37191K wps
[Epoch 81 Batch 30/162] avg loss 0.00219461, throughput 9.5966K wps
[Epoch 81 Batch 60/162] avg loss 0.00233392, throughput 9.42277K wps
[Epoch 81 Batch 90/162] avg loss 0.0020885, throughput 9.61725K wps
[Epoch 81 Batch 120/162] avg loss 0.00215105, throughput 9.48584K wps
[Epoch 81 Batch 150/162] avg loss 0.00211223, throughput 9.32248K wps
Begin Testing...
[Epoch 81] train avg loss 0.00218494, dev acc 0.9167, dev avg loss 0.204692, throughput 9.50062K wps
[Epoch 82 Batch 30/162] avg loss 0.00209817, throughput 9.67439K wps
[Epoch 82 Batch 60/162] avg loss 0.00222351, throughput 9.3846K wps
[Epoch 82 Batch 90/162] avg loss 0.00210819, throughput 9.42394K wps
[Epoch 82 Batch 120/162] avg loss 0.00205875, throughput 9.3767K wps
[Epoch 82 Batch 150/162] avg loss 0.00206784, throughput 9.54912K wps
Begin Testing...
[Epoch 82] train avg loss 0.0020979, dev acc 0.9178, dev avg loss 0.204249, throughput 9.48434K wps
[Epoch 83 Batch 30/162] avg loss 0.00216141, throughput 9.69198K wps
[Epoch 83 Batch 60/162] avg loss 0.00197263, throughput 9.5139K wps
[Epoch 83 Batch 90/162] avg loss 0.00221968, throughput 9.57626K wps
[Epoch 83 Batch 120/162] avg loss 0.00208259, throughput 9.40631K wps
[Epoch 83 Batch 150/162] avg loss 0.00198753, throughput 9.41953K wps
Begin Testing...
[Epoch 83] train avg loss 0.00209728, dev acc 0.9178, dev avg loss 0.203932, throughput 9.5107K wps
[Epoch 84 Batch 30/162] avg loss 0.00212332, throughput 9.4803K wps
[Epoch 84 Batch 60/162] avg loss 0.00211027, throughput 9.61343K wps
[Epoch 84 Batch 90/162] avg loss 0.00221945, throughput 9.50269K wps
[Epoch 84 Batch 120/162] avg loss 0.00212179, throughput 9.31858K wps
[Epoch 84 Batch 150/162] avg loss 0.00196616, throughput 9.3493K wps
Begin Testing...
[Epoch 84] train avg loss 0.00210958, dev acc 0.9178, dev avg loss 0.203869, throughput 9.46579K wps
[Epoch 85 Batch 30/162] avg loss 0.00208976, throughput 9.55269K wps
[Epoch 85 Batch 60/162] avg loss 0.00201801, throughput 9.46874K wps
[Epoch 85 Batch 90/162] avg loss 0.0019091, throughput 9.35444K wps
[Epoch 85 Batch 120/162] avg loss 0.00200232, throughput 9.48897K wps
[Epoch 85 Batch 150/162] avg loss 0.00218931, throughput 9.43171K wps
Begin Testing...
[Epoch 85] train avg loss 0.00204443, dev acc 0.9178, dev avg loss 0.203934, throughput 9.43316K wps
[Epoch 86 Batch 30/162] avg loss 0.00211727, throughput 9.6974K wps
[Epoch 86 Batch 60/162] avg loss 0.00207113, throughput 9.30964K wps
[Epoch 86 Batch 90/162] avg loss 0.00185112, throughput 9.34813K wps
[Epoch 86 Batch 120/162] avg loss 0.00212271, throughput 9.3216K wps
[Epoch 86 Batch 150/162] avg loss 0.00202072, throughput 9.61068K wps
Begin Testing...
[Epoch 86] train avg loss 0.00202856, dev acc 0.9167, dev avg loss 0.206555, throughput 9.46183K wps
[Epoch 87 Batch 30/162] avg loss 0.00189716, throughput 9.60306K wps
[Epoch 87 Batch 60/162] avg loss 0.00204118, throughput 9.29589K wps
[Epoch 87 Batch 90/162] avg loss 0.00218356, throughput 9.51961K wps
[Epoch 87 Batch 120/162] avg loss 0.00182706, throughput 9.23602K wps
[Epoch 87 Batch 150/162] avg loss 0.00196395, throughput 9.37668K wps
Begin Testing...
[Epoch 87] train avg loss 0.00196859, dev acc 0.9178, dev avg loss 0.202876, throughput 9.41142K wps
[Epoch 88 Batch 30/162] avg loss 0.00204261, throughput 9.57811K wps
[Epoch 88 Batch 60/162] avg loss 0.00207047, throughput 9.47736K wps
[Epoch 88 Batch 90/162] avg loss 0.00178826, throughput 9.45543K wps
[Epoch 88 Batch 120/162] avg loss 0.00192106, throughput 9.35268K wps
[Epoch 88 Batch 150/162] avg loss 0.00213253, throughput 9.48186K wps
Begin Testing...
[Epoch 88] train avg loss 0.00198777, dev acc 0.9178, dev avg loss 0.203035, throughput 9.45533K wps
[Epoch 89 Batch 30/162] avg loss 0.00177821, throughput 9.58372K wps
[Epoch 89 Batch 60/162] avg loss 0.00192296, throughput 9.39022K wps
[Epoch 89 Batch 90/162] avg loss 0.00220993, throughput 9.49686K wps
[Epoch 89 Batch 120/162] avg loss 0.00193389, throughput 9.24021K wps
[Epoch 89 Batch 150/162] avg loss 0.00198891, throughput 9.34758K wps
Begin Testing...
[Epoch 89] train avg loss 0.00194793, dev acc 0.9178, dev avg loss 0.202888, throughput 9.40415K wps
[Epoch 90 Batch 30/162] avg loss 0.00193358, throughput 9.83959K wps
[Epoch 90 Batch 60/162] avg loss 0.00187825, throughput 9.54346K wps
[Epoch 90 Batch 90/162] avg loss 0.00191117, throughput 9.35081K wps
[Epoch 90 Batch 120/162] avg loss 0.00184486, throughput 9.39868K wps
[Epoch 90 Batch 150/162] avg loss 0.00194007, throughput 9.47053K wps
Begin Testing...
[Epoch 90] train avg loss 0.00189864, dev acc 0.9178, dev avg loss 0.202907, throughput 9.51493K wps
[Epoch 91 Batch 30/162] avg loss 0.00207203, throughput 9.56273K wps
[Epoch 91 Batch 60/162] avg loss 0.00165915, throughput 9.22933K wps
[Epoch 91 Batch 90/162] avg loss 0.00207815, throughput 9.42723K wps
[Epoch 91 Batch 120/162] avg loss 0.00176788, throughput 9.50385K wps
[Epoch 91 Batch 150/162] avg loss 0.00194098, throughput 9.26717K wps
Begin Testing...
[Epoch 91] train avg loss 0.00190179, dev acc 0.9189, dev avg loss 0.203068, throughput 9.41169K wps
[Epoch 92 Batch 30/162] avg loss 0.00178526, throughput 9.41938K wps
[Epoch 92 Batch 60/162] avg loss 0.00196837, throughput 9.45556K wps
[Epoch 92 Batch 90/162] avg loss 0.00196454, throughput 9.45827K wps
[Epoch 92 Batch 120/162] avg loss 0.00180546, throughput 9.23448K wps
[Epoch 92 Batch 150/162] avg loss 0.0017621, throughput 9.55964K wps
Begin Testing...
[Epoch 92] train avg loss 0.00184886, dev acc 0.9200, dev avg loss 0.202755, throughput 9.43996K wps
Observed Improvement.
Begin Testing...
[Epoch 93 Batch 30/162] avg loss 0.0018225, throughput 9.5948K wps
[Epoch 93 Batch 60/162] avg loss 0.00184558, throughput 9.23939K wps
[Epoch 93 Batch 90/162] avg loss 0.0019831, throughput 9.37298K wps
[Epoch 93 Batch 120/162] avg loss 0.00195239, throughput 9.61804K wps
[Epoch 93 Batch 150/162] avg loss 0.00172166, throughput 9.40221K wps
Begin Testing...
[Epoch 93] train avg loss 0.0018492, dev acc 0.9189, dev avg loss 0.203956, throughput 9.41678K wps
[Epoch 94 Batch 30/162] avg loss 0.00190348, throughput 9.48548K wps
[Epoch 94 Batch 60/162] avg loss 0.00194797, throughput 9.40303K wps
[Epoch 94 Batch 90/162] avg loss 0.00179099, throughput 9.3183K wps
[Epoch 94 Batch 120/162] avg loss 0.00163815, throughput 9.34593K wps
[Epoch 94 Batch 150/162] avg loss 0.00198078, throughput 9.5256K wps
Begin Testing...
[Epoch 94] train avg loss 0.00182133, dev acc 0.9178, dev avg loss 0.201901, throughput 9.43058K wps
[Epoch 95 Batch 30/162] avg loss 0.0018496, throughput 9.49513K wps
[Epoch 95 Batch 60/162] avg loss 0.00176679, throughput 9.32525K wps
[Epoch 95 Batch 90/162] avg loss 0.0017089, throughput 9.32987K wps
[Epoch 95 Batch 120/162] avg loss 0.00176831, throughput 9.29309K wps
[Epoch 95 Batch 150/162] avg loss 0.00176792, throughput 9.61112K wps
Begin Testing...
[Epoch 95] train avg loss 0.00175803, dev acc 0.9211, dev avg loss 0.202255, throughput 9.40555K wps
Observed Improvement.
Begin Testing...
[Epoch 96 Batch 30/162] avg loss 0.001919, throughput 9.48883K wps
[Epoch 96 Batch 60/162] avg loss 0.00177151, throughput 9.24487K wps
[Epoch 96 Batch 90/162] avg loss 0.00166347, throughput 9.38465K wps
[Epoch 96 Batch 120/162] avg loss 0.00189636, throughput 9.31054K wps
[Epoch 96 Batch 150/162] avg loss 0.00175136, throughput 9.59829K wps
Begin Testing...
[Epoch 96] train avg loss 0.00180056, dev acc 0.9189, dev avg loss 0.202253, throughput 9.40084K wps
[Epoch 97 Batch 30/162] avg loss 0.00179326, throughput 9.46509K wps
[Epoch 97 Batch 60/162] avg loss 0.00173355, throughput 9.37721K wps
[Epoch 97 Batch 90/162] avg loss 0.00154658, throughput 9.30914K wps
[Epoch 97 Batch 120/162] avg loss 0.00171731, throughput 9.29621K wps
[Epoch 97 Batch 150/162] avg loss 0.0017593, throughput 9.56154K wps
Begin Testing...
[Epoch 97] train avg loss 0.00170653, dev acc 0.9189, dev avg loss 0.201612, throughput 9.37815K wps
[Epoch 98 Batch 30/162] avg loss 0.00173192, throughput 9.62262K wps
[Epoch 98 Batch 60/162] avg loss 0.00164598, throughput 9.25017K wps
[Epoch 98 Batch 90/162] avg loss 0.00177925, throughput 9.42925K wps
[Epoch 98 Batch 120/162] avg loss 0.00181834, throughput 9.50292K wps
[Epoch 98 Batch 150/162] avg loss 0.00165632, throughput 9.37837K wps
Begin Testing...
[Epoch 98] train avg loss 0.00174675, dev acc 0.9178, dev avg loss 0.201147, throughput 9.43347K wps
[Epoch 99 Batch 30/162] avg loss 0.00167377, throughput 9.65808K wps
[Epoch 99 Batch 60/162] avg loss 0.00180385, throughput 9.31154K wps
[Epoch 99 Batch 90/162] avg loss 0.00147296, throughput 9.34596K wps
[Epoch 99 Batch 120/162] avg loss 0.00184987, throughput 9.33137K wps
[Epoch 99 Batch 150/162] avg loss 0.00167669, throughput 9.60396K wps
Begin Testing...
[Epoch 99] train avg loss 0.00168801, dev acc 0.9189, dev avg loss 0.202509, throughput 9.44778K wps
[Epoch 100 Batch 30/162] avg loss 0.00160823, throughput 9.46893K wps
[Epoch 100 Batch 60/162] avg loss 0.00167075, throughput 9.36293K wps
[Epoch 100 Batch 90/162] avg loss 0.00162697, throughput 9.53216K wps
[Epoch 100 Batch 120/162] avg loss 0.00166427, throughput 9.23275K wps
[Epoch 100 Batch 150/162] avg loss 0.00160112, throughput 9.37408K wps
Begin Testing...
[Epoch 100] train avg loss 0.00165907, dev acc 0.9200, dev avg loss 0.200766, throughput 9.38997K wps
[Epoch 101 Batch 30/162] avg loss 0.00166235, throughput 9.72319K wps
[Epoch 101 Batch 60/162] avg loss 0.00170181, throughput 9.3252K wps
[Epoch 101 Batch 90/162] avg loss 0.00162537, throughput 9.54429K wps
[Epoch 101 Batch 120/162] avg loss 0.00162783, throughput 9.42488K wps
[Epoch 101 Batch 150/162] avg loss 0.0016265, throughput 9.27651K wps
Begin Testing...
[Epoch 101] train avg loss 0.00165368, dev acc 0.9200, dev avg loss 0.201949, throughput 9.46257K wps
[Epoch 102 Batch 30/162] avg loss 0.00157725, throughput 9.64011K wps
[Epoch 102 Batch 60/162] avg loss 0.0017191, throughput 9.21539K wps
[Epoch 102 Batch 90/162] avg loss 0.00158635, throughput 9.53335K wps
[Epoch 102 Batch 120/162] avg loss 0.00172382, throughput 9.53554K wps
[Epoch 102 Batch 150/162] avg loss 0.00152106, throughput 9.35435K wps
Begin Testing...
[Epoch 102] train avg loss 0.00163151, dev acc 0.9211, dev avg loss 0.201249, throughput 9.44903K wps
Observed Improvement.
Begin Testing...
[Epoch 103 Batch 30/162] avg loss 0.0016471, throughput 9.43055K wps
[Epoch 103 Batch 60/162] avg loss 0.00152364, throughput 9.39166K wps
[Epoch 103 Batch 90/162] avg loss 0.00157243, throughput 9.44942K wps
[Epoch 103 Batch 120/162] avg loss 0.00154024, throughput 9.30904K wps
[Epoch 103 Batch 150/162] avg loss 0.00167516, throughput 9.42171K wps
Begin Testing...
[Epoch 103] train avg loss 0.00160252, dev acc 0.9211, dev avg loss 0.200536, throughput 9.42119K wps
Observed Improvement.
Begin Testing...
[Epoch 104 Batch 30/162] avg loss 0.0014871, throughput 9.7828K wps
[Epoch 104 Batch 60/162] avg loss 0.00166755, throughput 9.30856K wps
[Epoch 104 Batch 90/162] avg loss 0.00146944, throughput 9.30463K wps
[Epoch 104 Batch 120/162] avg loss 0.00148933, throughput 9.47154K wps
[Epoch 104 Batch 150/162] avg loss 0.00166115, throughput 9.49217K wps
Begin Testing...
[Epoch 104] train avg loss 0.0015517, dev acc 0.9222, dev avg loss 0.201214, throughput 9.44257K wps
Observed Improvement.
Begin Testing...
[Epoch 105 Batch 30/162] avg loss 0.00155719, throughput 9.62544K wps
[Epoch 105 Batch 60/162] avg loss 0.00155679, throughput 9.44923K wps
[Epoch 105 Batch 90/162] avg loss 0.00154308, throughput 9.39391K wps
[Epoch 105 Batch 120/162] avg loss 0.00153408, throughput 9.26192K wps
[Epoch 105 Batch 150/162] avg loss 0.00158113, throughput 9.47907K wps
Begin Testing...
[Epoch 105] train avg loss 0.00154243, dev acc 0.9222, dev avg loss 0.201708, throughput 9.43284K wps
Observed Improvement.
Begin Testing...
[Epoch 106 Batch 30/162] avg loss 0.00162147, throughput 9.47069K wps
[Epoch 106 Batch 60/162] avg loss 0.0016228, throughput 9.37276K wps
[Epoch 106 Batch 90/162] avg loss 0.00151918, throughput 9.44175K wps
[Epoch 106 Batch 120/162] avg loss 0.00162879, throughput 9.20823K wps
[Epoch 106 Batch 150/162] avg loss 0.00145652, throughput 9.36163K wps
Begin Testing...
[Epoch 106] train avg loss 0.00154954, dev acc 0.9211, dev avg loss 0.200698, throughput 9.3633K wps
[Epoch 107 Batch 30/162] avg loss 0.00157944, throughput 9.55261K wps
[Epoch 107 Batch 60/162] avg loss 0.00160107, throughput 9.49341K wps
[Epoch 107 Batch 90/162] avg loss 0.0017473, throughput 9.44431K wps
[Epoch 107 Batch 120/162] avg loss 0.0014576, throughput 9.32259K wps
[Epoch 107 Batch 150/162] avg loss 0.00152928, throughput 9.45957K wps
Begin Testing...
[Epoch 107] train avg loss 0.00159164, dev acc 0.9211, dev avg loss 0.200102, throughput 9.43569K wps
[Epoch 108 Batch 30/162] avg loss 0.00141645, throughput 9.65685K wps
[Epoch 108 Batch 60/162] avg loss 0.00141288, throughput 9.24937K wps
[Epoch 108 Batch 90/162] avg loss 0.00132503, throughput 9.52315K wps
[Epoch 108 Batch 120/162] avg loss 0.00155865, throughput 9.36979K wps
[Epoch 108 Batch 150/162] avg loss 0.0016055, throughput 9.32483K wps
Begin Testing...
[Epoch 108] train avg loss 0.0014618, dev acc 0.9200, dev avg loss 0.199801, throughput 9.4027K wps
[Epoch 109 Batch 30/162] avg loss 0.00149596, throughput 9.49883K wps
[Epoch 109 Batch 60/162] avg loss 0.00155767, throughput 9.32054K wps
[Epoch 109 Batch 90/162] avg loss 0.00148827, throughput 9.58784K wps
[Epoch 109 Batch 120/162] avg loss 0.00129597, throughput 9.35212K wps
[Epoch 109 Batch 150/162] avg loss 0.0015284, throughput 9.33439K wps
Begin Testing...
[Epoch 109] train avg loss 0.00148549, dev acc 0.9200, dev avg loss 0.199917, throughput 9.43166K wps
[Epoch 110 Batch 30/162] avg loss 0.00132206, throughput 9.78502K wps
[Epoch 110 Batch 60/162] avg loss 0.00139716, throughput 9.55186K wps
[Epoch 110 Batch 90/162] avg loss 0.00142869, throughput 9.3887K wps
[Epoch 110 Batch 120/162] avg loss 0.00161506, throughput 9.38856K wps
[Epoch 110 Batch 150/162] avg loss 0.00138264, throughput 9.4251K wps
Begin Testing...
[Epoch 110] train avg loss 0.00143514, dev acc 0.9200, dev avg loss 0.199741, throughput 9.5063K wps
[Epoch 111 Batch 30/162] avg loss 0.00146158, throughput 9.54358K wps
[Epoch 111 Batch 60/162] avg loss 0.00139743, throughput 9.41566K wps
[Epoch 111 Batch 90/162] avg loss 0.00149614, throughput 9.30654K wps
[Epoch 111 Batch 120/162] avg loss 0.00135073, throughput 9.27336K wps
[Epoch 111 Batch 150/162] avg loss 0.0014768, throughput 9.37261K wps
Begin Testing...
[Epoch 111] train avg loss 0.00142701, dev acc 0.9211, dev avg loss 0.201406, throughput 9.37322K wps
[Epoch 112 Batch 30/162] avg loss 0.00135556, throughput 9.65498K wps
[Epoch 112 Batch 60/162] avg loss 0.0012309, throughput 9.39201K wps
[Epoch 112 Batch 90/162] avg loss 0.00178642, throughput 9.53596K wps
[Epoch 112 Batch 120/162] avg loss 0.00126391, throughput 9.56475K wps
[Epoch 112 Batch 150/162] avg loss 0.00141663, throughput 9.34482K wps
Begin Testing...
[Epoch 112] train avg loss 0.0014121, dev acc 0.9211, dev avg loss 0.199392, throughput 9.50825K wps
[Epoch 113 Batch 30/162] avg loss 0.0012674, throughput 9.47132K wps
[Epoch 113 Batch 60/162] avg loss 0.0014929, throughput 9.32828K wps
[Epoch 113 Batch 90/162] avg loss 0.00142912, throughput 9.42846K wps
[Epoch 113 Batch 120/162] avg loss 0.00153575, throughput 9.48607K wps
[Epoch 113 Batch 150/162] avg loss 0.00135428, throughput 9.40981K wps
Begin Testing...
[Epoch 113] train avg loss 0.00142419, dev acc 0.9222, dev avg loss 0.200158, throughput 9.43208K wps
Observed Improvement.
Begin Testing...
[Epoch 114 Batch 30/162] avg loss 0.00145002, throughput 9.75467K wps
[Epoch 114 Batch 60/162] avg loss 0.00156375, throughput 9.41808K wps
[Epoch 114 Batch 90/162] avg loss 0.00125475, throughput 9.50728K wps
[Epoch 114 Batch 120/162] avg loss 0.00123433, throughput 9.25453K wps
[Epoch 114 Batch 150/162] avg loss 0.00142943, throughput 9.40689K wps
Begin Testing...
[Epoch 114] train avg loss 0.00136915, dev acc 0.9211, dev avg loss 0.199391, throughput 9.44424K wps
[Epoch 115 Batch 30/162] avg loss 0.00139689, throughput 9.50412K wps
[Epoch 115 Batch 60/162] avg loss 0.00146258, throughput 9.46655K wps
[Epoch 115 Batch 90/162] avg loss 0.0013951, throughput 9.25689K wps
[Epoch 115 Batch 120/162] avg loss 0.0014882, throughput 9.42235K wps
[Epoch 115 Batch 150/162] avg loss 0.0013576, throughput 9.4666K wps
Begin Testing...
[Epoch 115] train avg loss 0.00141653, dev acc 0.9222, dev avg loss 0.199396, throughput 9.40622K wps
Observed Improvement.
Begin Testing...
[Epoch 116 Batch 30/162] avg loss 0.00123126, throughput 9.66187K wps
[Epoch 116 Batch 60/162] avg loss 0.00145207, throughput 9.21584K wps
[Epoch 116 Batch 90/162] avg loss 0.00122372, throughput 9.2379K wps
[Epoch 116 Batch 120/162] avg loss 0.00142571, throughput 9.35462K wps
[Epoch 116 Batch 150/162] avg loss 0.00145025, throughput 9.40324K wps
Begin Testing...
[Epoch 116] train avg loss 0.00135842, dev acc 0.9211, dev avg loss 0.19932, throughput 9.37062K wps
[Epoch 117 Batch 30/162] avg loss 0.00146718, throughput 9.77829K wps
[Epoch 117 Batch 60/162] avg loss 0.00131374, throughput 9.57966K wps
[Epoch 117 Batch 90/162] avg loss 0.00117064, throughput 9.4197K wps
[Epoch 117 Batch 120/162] avg loss 0.00119303, throughput 9.64351K wps
[Epoch 117 Batch 150/162] avg loss 0.00128323, throughput 9.56886K wps
Begin Testing...
[Epoch 117] train avg loss 0.00131644, dev acc 0.9200, dev avg loss 0.200394, throughput 9.57672K wps
[Epoch 118 Batch 30/162] avg loss 0.00129232, throughput 9.63304K wps
[Epoch 118 Batch 60/162] avg loss 0.00127375, throughput 9.46027K wps
[Epoch 118 Batch 90/162] avg loss 0.00132698, throughput 9.48068K wps
[Epoch 118 Batch 120/162] avg loss 0.00141471, throughput 9.32657K wps
[Epoch 118 Batch 150/162] avg loss 0.00138714, throughput 9.57072K wps
Begin Testing...
[Epoch 118] train avg loss 0.00133771, dev acc 0.9200, dev avg loss 0.200515, throughput 9.48687K wps
[Epoch 119 Batch 30/162] avg loss 0.00126112, throughput 9.65556K wps
[Epoch 119 Batch 60/162] avg loss 0.00128316, throughput 9.39931K wps
[Epoch 119 Batch 90/162] avg loss 0.00135514, throughput 9.52061K wps
[Epoch 119 Batch 120/162] avg loss 0.00146752, throughput 9.25823K wps
[Epoch 119 Batch 150/162] avg loss 0.00128907, throughput 9.29855K wps
Begin Testing...
[Epoch 119] train avg loss 0.00131317, dev acc 0.9178, dev avg loss 0.201312, throughput 9.41353K wps
[Epoch 120 Batch 30/162] avg loss 0.00126354, throughput 9.61098K wps
[Epoch 120 Batch 60/162] avg loss 0.00125299, throughput 9.34847K wps
[Epoch 120 Batch 90/162] avg loss 0.00122949, throughput 9.4915K wps
[Epoch 120 Batch 120/162] avg loss 0.00127314, throughput 9.57557K wps
[Epoch 120 Batch 150/162] avg loss 0.00140001, throughput 9.31617K wps
Begin Testing...
[Epoch 120] train avg loss 0.00127869, dev acc 0.9167, dev avg loss 0.202237, throughput 9.47072K wps
[Epoch 121 Batch 30/162] avg loss 0.00120097, throughput 9.84819K wps
[Epoch 121 Batch 60/162] avg loss 0.00128465, throughput 9.50958K wps
[Epoch 121 Batch 90/162] avg loss 0.00135424, throughput 9.34155K wps
[Epoch 121 Batch 120/162] avg loss 0.0013264, throughput 9.47978K wps
[Epoch 121 Batch 150/162] avg loss 0.00123374, throughput 9.45314K wps
Begin Testing...
[Epoch 121] train avg loss 0.00128595, dev acc 0.9189, dev avg loss 0.200314, throughput 9.52004K wps
[Epoch 122 Batch 30/162] avg loss 0.00122095, throughput 9.62941K wps
[Epoch 122 Batch 60/162] avg loss 0.00136059, throughput 9.56211K wps
[Epoch 122 Batch 90/162] avg loss 0.00118195, throughput 9.41465K wps
[Epoch 122 Batch 120/162] avg loss 0.00127616, throughput 9.58194K wps
[Epoch 122 Batch 150/162] avg loss 0.00102895, throughput 9.4896K wps
Begin Testing...
[Epoch 122] train avg loss 0.00122096, dev acc 0.9200, dev avg loss 0.200774, throughput 9.5178K wps
[Epoch 123 Batch 30/162] avg loss 0.00113394, throughput 9.73403K wps
[Epoch 123 Batch 60/162] avg loss 0.00118843, throughput 9.25159K wps
[Epoch 123 Batch 90/162] avg loss 0.00132928, throughput 9.37584K wps
[Epoch 123 Batch 120/162] avg loss 0.00117057, throughput 9.38296K wps
[Epoch 123 Batch 150/162] avg loss 0.00111073, throughput 9.55164K wps
Begin Testing...
[Epoch 123] train avg loss 0.0012109, dev acc 0.9200, dev avg loss 0.200036, throughput 9.4444K wps
[Epoch 124 Batch 30/162] avg loss 0.00116459, throughput 9.56489K wps
[Epoch 124 Batch 60/162] avg loss 0.00114074, throughput 9.4312K wps
[Epoch 124 Batch 90/162] avg loss 0.00109437, throughput 9.36161K wps
[Epoch 124 Batch 120/162] avg loss 0.00143706, throughput 9.50098K wps
[Epoch 124 Batch 150/162] avg loss 0.00128033, throughput 9.27076K wps
Begin Testing...
[Epoch 124] train avg loss 0.00120897, dev acc 0.9200, dev avg loss 0.203903, throughput 9.43607K wps
[Epoch 125 Batch 30/162] avg loss 0.00122595, throughput 9.5241K wps
[Epoch 125 Batch 60/162] avg loss 0.00112253, throughput 9.57219K wps
[Epoch 125 Batch 90/162] avg loss 0.00120417, throughput 9.30604K wps
[Epoch 125 Batch 120/162] avg loss 0.00128179, throughput 9.56216K wps
[Epoch 125 Batch 150/162] avg loss 0.00122073, throughput 9.45244K wps
Begin Testing...
[Epoch 125] train avg loss 0.00122048, dev acc 0.9267, dev avg loss 0.198693, throughput 9.49588K wps
Observed Improvement.
Begin Testing...
[Epoch 126 Batch 30/162] avg loss 0.00122414, throughput 9.83995K wps
[Epoch 126 Batch 60/162] avg loss 0.00101729, throughput 9.44438K wps
[Epoch 126 Batch 90/162] avg loss 0.0011575, throughput 9.40967K wps
[Epoch 126 Batch 120/162] avg loss 0.00117278, throughput 9.48843K wps
[Epoch 126 Batch 150/162] avg loss 0.00139849, throughput 9.42859K wps
Begin Testing...
[Epoch 126] train avg loss 0.00119782, dev acc 0.9222, dev avg loss 0.198676, throughput 9.51163K wps
[Epoch 127 Batch 30/162] avg loss 0.00112447, throughput 9.80328K wps
[Epoch 127 Batch 60/162] avg loss 0.0011851, throughput 9.53782K wps
[Epoch 127 Batch 90/162] avg loss 0.00119993, throughput 9.47315K wps
[Epoch 127 Batch 120/162] avg loss 0.00113144, throughput 9.321K wps
[Epoch 127 Batch 150/162] avg loss 0.00121932, throughput 9.32469K wps
Begin Testing...
[Epoch 127] train avg loss 0.00117533, dev acc 0.9222, dev avg loss 0.199501, throughput 9.47447K wps
[Epoch 128 Batch 30/162] avg loss 0.00123704, throughput 9.54846K wps
[Epoch 128 Batch 60/162] avg loss 0.0011516, throughput 9.36584K wps
[Epoch 128 Batch 90/162] avg loss 0.00111304, throughput 9.24637K wps
[Epoch 128 Batch 120/162] avg loss 0.00114678, throughput 9.42821K wps
[Epoch 128 Batch 150/162] avg loss 0.00111251, throughput 9.3985K wps
Begin Testing...
[Epoch 128] train avg loss 0.00117165, dev acc 0.9244, dev avg loss 0.198568, throughput 9.38661K wps
[Epoch 129 Batch 30/162] avg loss 0.00123391, throughput 9.48081K wps
[Epoch 129 Batch 60/162] avg loss 0.00118635, throughput 9.28781K wps
[Epoch 129 Batch 90/162] avg loss 0.0011579, throughput 9.39547K wps
[Epoch 129 Batch 120/162] avg loss 0.00119239, throughput 9.29944K wps
[Epoch 129 Batch 150/162] avg loss 0.00110653, throughput 9.45271K wps
Begin Testing...
[Epoch 129] train avg loss 0.00118958, dev acc 0.9200, dev avg loss 0.200623, throughput 9.38182K wps
[Epoch 130 Batch 30/162] avg loss 0.00126734, throughput 9.74523K wps
[Epoch 130 Batch 60/162] avg loss 0.00107326, throughput 9.34576K wps
[Epoch 130 Batch 90/162] avg loss 0.00119776, throughput 9.56788K wps
[Epoch 130 Batch 120/162] avg loss 0.00131288, throughput 9.29193K wps
[Epoch 130 Batch 150/162] avg loss 0.00102367, throughput 9.35497K wps
Begin Testing...
[Epoch 130] train avg loss 0.00116492, dev acc 0.9211, dev avg loss 0.199081, throughput 9.4573K wps
[Epoch 131 Batch 30/162] avg loss 0.0011374, throughput 9.54433K wps
[Epoch 131 Batch 60/162] avg loss 0.00115892, throughput 9.40461K wps
[Epoch 131 Batch 90/162] avg loss 0.00112882, throughput 9.45483K wps
[Epoch 131 Batch 120/162] avg loss 0.0012397, throughput 9.36505K wps
[Epoch 131 Batch 150/162] avg loss 0.00119974, throughput 9.50395K wps
Begin Testing...
[Epoch 131] train avg loss 0.0011698, dev acc 0.9222, dev avg loss 0.198727, throughput 9.46904K wps
[Epoch 132 Batch 30/162] avg loss 0.00114839, throughput 9.69259K wps
[Epoch 132 Batch 60/162] avg loss 0.00107325, throughput 9.62718K wps
[Epoch 132 Batch 90/162] avg loss 0.00111775, throughput 9.50188K wps
[Epoch 132 Batch 120/162] avg loss 0.00104873, throughput 9.46268K wps
[Epoch 132 Batch 150/162] avg loss 0.00111144, throughput 9.47854K wps
Begin Testing...
[Epoch 132] train avg loss 0.00111229, dev acc 0.9211, dev avg loss 0.198605, throughput 9.54743K wps
[Epoch 133 Batch 30/162] avg loss 0.00112505, throughput 9.70845K wps
[Epoch 133 Batch 60/162] avg loss 0.00114743, throughput 9.39054K wps
[Epoch 133 Batch 90/162] avg loss 0.0011296, throughput 9.38848K wps
[Epoch 133 Batch 120/162] avg loss 0.00110726, throughput 9.30991K wps
[Epoch 133 Batch 150/162] avg loss 0.00105538, throughput 9.57845K wps
Begin Testing...
[Epoch 133] train avg loss 0.00109864, dev acc 0.9267, dev avg loss 0.198573, throughput 9.45389K wps
Observed Improvement.
Begin Testing...
[Epoch 134 Batch 30/162] avg loss 0.00118438, throughput 9.74429K wps
[Epoch 134 Batch 60/162] avg loss 0.00127878, throughput 9.28595K wps
[Epoch 134 Batch 90/162] avg loss 0.00110584, throughput 9.61692K wps
[Epoch 134 Batch 120/162] avg loss 0.00109656, throughput 9.5527K wps
[Epoch 134 Batch 150/162] avg loss 0.000980213, throughput 9.50213K wps
Begin Testing...
[Epoch 134] train avg loss 0.00111049, dev acc 0.9211, dev avg loss 0.198945, throughput 9.53518K wps
[Epoch 135 Batch 30/162] avg loss 0.000951821, throughput 9.52573K wps
[Epoch 135 Batch 60/162] avg loss 0.00106444, throughput 9.47218K wps
[Epoch 135 Batch 90/162] avg loss 0.00107229, throughput 9.4053K wps
[Epoch 135 Batch 120/162] avg loss 0.00111467, throughput 9.36691K wps
[Epoch 135 Batch 150/162] avg loss 0.00119743, throughput 9.40534K wps
Begin Testing...
[Epoch 135] train avg loss 0.00108335, dev acc 0.9189, dev avg loss 0.199848, throughput 9.44943K wps
[Epoch 136 Batch 30/162] avg loss 0.0011337, throughput 9.64844K wps
[Epoch 136 Batch 60/162] avg loss 0.000968694, throughput 9.52201K wps
[Epoch 136 Batch 90/162] avg loss 0.00094221, throughput 9.31776K wps
[Epoch 136 Batch 120/162] avg loss 0.00118016, throughput 9.40238K wps
[Epoch 136 Batch 150/162] avg loss 0.0010837, throughput 9.55416K wps
Begin Testing...
[Epoch 136] train avg loss 0.00106, dev acc 0.9200, dev avg loss 0.199917, throughput 9.48905K wps
[Epoch 137 Batch 30/162] avg loss 0.0010832, throughput 9.65422K wps
[Epoch 137 Batch 60/162] avg loss 0.00101215, throughput 9.28983K wps
[Epoch 137 Batch 90/162] avg loss 0.00103035, throughput 9.33631K wps
[Epoch 137 Batch 120/162] avg loss 0.00111634, throughput 9.61457K wps
[Epoch 137 Batch 150/162] avg loss 0.000955523, throughput 9.44312K wps
Begin Testing...
[Epoch 137] train avg loss 0.0010401, dev acc 0.9222, dev avg loss 0.198783, throughput 9.45671K wps
[Epoch 138 Batch 30/162] avg loss 0.00104012, throughput 9.75833K wps
[Epoch 138 Batch 60/162] avg loss 0.000960727, throughput 9.51503K wps
[Epoch 138 Batch 90/162] avg loss 0.000979053, throughput 9.2396K wps
[Epoch 138 Batch 120/162] avg loss 0.00114121, throughput 9.39113K wps
[Epoch 138 Batch 150/162] avg loss 0.00099514, throughput 9.32948K wps
Begin Testing...
[Epoch 138] train avg loss 0.00102415, dev acc 0.9200, dev avg loss 0.199828, throughput 9.46137K wps
[Epoch 139 Batch 30/162] avg loss 0.00104314, throughput 9.67382K wps
[Epoch 139 Batch 60/162] avg loss 0.0010679, throughput 9.56573K wps
[Epoch 139 Batch 90/162] avg loss 0.00111646, throughput 9.48577K wps
[Epoch 139 Batch 120/162] avg loss 0.000992503, throughput 9.31018K wps
[Epoch 139 Batch 150/162] avg loss 0.000956073, throughput 9.345K wps
Begin Testing...
[Epoch 139] train avg loss 0.00103061, dev acc 0.9200, dev avg loss 0.200709, throughput 9.47789K wps
[Epoch 140 Batch 30/162] avg loss 0.00101081, throughput 9.74945K wps
[Epoch 140 Batch 60/162] avg loss 0.000925013, throughput 9.51185K wps
[Epoch 140 Batch 90/162] avg loss 0.000924114, throughput 9.341K wps
[Epoch 140 Batch 120/162] avg loss 0.00103509, throughput 9.29989K wps
[Epoch 140 Batch 150/162] avg loss 0.00122123, throughput 9.41744K wps
Begin Testing...
[Epoch 140] train avg loss 0.0010237, dev acc 0.9233, dev avg loss 0.198594, throughput 9.45118K wps
[Epoch 141 Batch 30/162] avg loss 0.00112801, throughput 9.50258K wps
[Epoch 141 Batch 60/162] avg loss 0.000966827, throughput 9.44862K wps
[Epoch 141 Batch 90/162] avg loss 0.00099807, throughput 9.40835K wps
[Epoch 141 Batch 120/162] avg loss 0.00112632, throughput 9.32449K wps
[Epoch 141 Batch 150/162] avg loss 0.00104741, throughput 9.46399K wps
Begin Testing...
[Epoch 141] train avg loss 0.00104734, dev acc 0.9200, dev avg loss 0.199433, throughput 9.43847K wps
[Epoch 142 Batch 30/162] avg loss 0.00100774, throughput 9.60706K wps
[Epoch 142 Batch 60/162] avg loss 0.00091134, throughput 9.55926K wps
[Epoch 142 Batch 90/162] avg loss 0.00103602, throughput 9.42383K wps
[Epoch 142 Batch 120/162] avg loss 0.000892126, throughput 9.48672K wps
[Epoch 142 Batch 150/162] avg loss 0.00096457, throughput 9.31282K wps
Begin Testing...
[Epoch 142] train avg loss 0.000960899, dev acc 0.9256, dev avg loss 0.198792, throughput 9.47258K wps
[Epoch 143 Batch 30/162] avg loss 0.000911537, throughput 9.50119K wps
[Epoch 143 Batch 60/162] avg loss 0.000990997, throughput 9.38806K wps
[Epoch 143 Batch 90/162] avg loss 0.00103406, throughput 9.37189K wps
[Epoch 143 Batch 120/162] avg loss 0.00105439, throughput 9.42889K wps
[Epoch 143 Batch 150/162] avg loss 0.00104742, throughput 9.55173K wps
Begin Testing...
[Epoch 143] train avg loss 0.00101029, dev acc 0.9244, dev avg loss 0.198735, throughput 9.43379K wps
[Epoch 144 Batch 30/162] avg loss 0.00103161, throughput 9.49657K wps
[Epoch 144 Batch 60/162] avg loss 0.000987322, throughput 9.46036K wps
[Epoch 144 Batch 90/162] avg loss 0.0009654, throughput 9.35749K wps
[Epoch 144 Batch 120/162] avg loss 0.000876967, throughput 9.33762K wps
[Epoch 144 Batch 150/162] avg loss 0.00102326, throughput 9.45841K wps
Begin Testing...
[Epoch 144] train avg loss 0.00096817, dev acc 0.9200, dev avg loss 0.199545, throughput 9.4321K wps
[Epoch 145 Batch 30/162] avg loss 0.000942425, throughput 9.5046K wps
[Epoch 145 Batch 60/162] avg loss 0.000880348, throughput 9.50288K wps
[Epoch 145 Batch 90/162] avg loss 0.000934699, throughput 9.46826K wps
[Epoch 145 Batch 120/162] avg loss 0.000941132, throughput 9.54349K wps
[Epoch 145 Batch 150/162] avg loss 0.00103938, throughput 9.44504K wps
Begin Testing...
[Epoch 145] train avg loss 0.000965215, dev acc 0.9222, dev avg loss 0.199497, throughput 9.47077K wps
[Epoch 146 Batch 30/162] avg loss 0.00108623, throughput 9.54697K wps
[Epoch 146 Batch 60/162] avg loss 0.00104724, throughput 9.41657K wps
[Epoch 146 Batch 90/162] avg loss 0.00088306, throughput 9.63263K wps
[Epoch 146 Batch 120/162] avg loss 0.000965227, throughput 9.30892K wps
[Epoch 146 Batch 150/162] avg loss 0.00106214, throughput 9.38229K wps
Begin Testing...
[Epoch 146] train avg loss 0.00101385, dev acc 0.9211, dev avg loss 0.199102, throughput 9.45293K wps
[Epoch 147 Batch 30/162] avg loss 0.000953613, throughput 9.47463K wps
[Epoch 147 Batch 60/162] avg loss 0.00100259, throughput 9.40584K wps
[Epoch 147 Batch 90/162] avg loss 0.000996786, throughput 9.3569K wps
[Epoch 147 Batch 120/162] avg loss 0.000976257, throughput 9.49522K wps
[Epoch 147 Batch 150/162] avg loss 0.00110328, throughput 9.53369K wps
Begin Testing...
[Epoch 147] train avg loss 0.0010046, dev acc 0.9244, dev avg loss 0.198703, throughput 9.45931K wps
[Epoch 148 Batch 30/162] avg loss 0.000987521, throughput 9.53456K wps
[Epoch 148 Batch 60/162] avg loss 0.000952668, throughput 9.50629K wps
[Epoch 148 Batch 90/162] avg loss 0.000900614, throughput 9.47786K wps
[Epoch 148 Batch 120/162] avg loss 0.000957141, throughput 9.27526K wps
[Epoch 148 Batch 150/162] avg loss 0.000949938, throughput 9.41687K wps
Begin Testing...
[Epoch 148] train avg loss 0.000948086, dev acc 0.9200, dev avg loss 0.199635, throughput 9.42668K wps
[Epoch 149 Batch 30/162] avg loss 0.000996697, throughput 9.43098K wps
[Epoch 149 Batch 60/162] avg loss 0.000812088, throughput 9.33715K wps
[Epoch 149 Batch 90/162] avg loss 0.000928236, throughput 9.47858K wps
[Epoch 149 Batch 120/162] avg loss 0.000960914, throughput 9.3957K wps
[Epoch 149 Batch 150/162] avg loss 0.000935762, throughput 9.32519K wps
Begin Testing...
[Epoch 149] train avg loss 0.000925015, dev acc 0.9211, dev avg loss 0.199426, throughput 9.39277K wps
[Epoch 150 Batch 30/162] avg loss 0.000882992, throughput 9.68211K wps
[Epoch 150 Batch 60/162] avg loss 0.000961667, throughput 9.41373K wps
[Epoch 150 Batch 90/162] avg loss 0.000888893, throughput 9.45114K wps
[Epoch 150 Batch 120/162] avg loss 0.000955521, throughput 9.46768K wps
[Epoch 150 Batch 150/162] avg loss 0.000985954, throughput 9.45305K wps
Begin Testing...
[Epoch 150] train avg loss 0.000928807, dev acc 0.9222, dev avg loss 0.199245, throughput 9.49367K wps
[Epoch 151 Batch 30/162] avg loss 0.000802809, throughput 9.68546K wps
[Epoch 151 Batch 60/162] avg loss 0.000962354, throughput 9.50235K wps
[Epoch 151 Batch 90/162] avg loss 0.00102501, throughput 9.34244K wps
[Epoch 151 Batch 120/162] avg loss 0.000856025, throughput 9.65015K wps
[Epoch 151 Batch 150/162] avg loss 0.000875561, throughput 9.57495K wps
Begin Testing...
[Epoch 151] train avg loss 0.000897011, dev acc 0.9244, dev avg loss 0.199281, throughput 9.52615K wps
[Epoch 152 Batch 30/162] avg loss 0.000878993, throughput 9.82813K wps
[Epoch 152 Batch 60/162] avg loss 0.000899776, throughput 9.34696K wps
[Epoch 152 Batch 90/162] avg loss 0.000976171, throughput 9.39521K wps
[Epoch 152 Batch 120/162] avg loss 0.000839825, throughput 9.39748K wps
[Epoch 152 Batch 150/162] avg loss 0.000951247, throughput 9.40701K wps
Begin Testing...
[Epoch 152] train avg loss 0.000909035, dev acc 0.9211, dev avg loss 0.200356, throughput 9.49282K wps
[Epoch 153 Batch 30/162] avg loss 0.000922641, throughput 9.49795K wps
[Epoch 153 Batch 60/162] avg loss 0.000936908, throughput 9.36141K wps
[Epoch 153 Batch 90/162] avg loss 0.00084419, throughput 9.49577K wps
[Epoch 153 Batch 120/162] avg loss 0.000923579, throughput 9.415K wps
[Epoch 153 Batch 150/162] avg loss 0.000858425, throughput 9.3123K wps
Begin Testing...
[Epoch 153] train avg loss 0.000897268, dev acc 0.9233, dev avg loss 0.19942, throughput 9.41189K wps
[Epoch 154 Batch 30/162] avg loss 0.000847193, throughput 9.50315K wps
[Epoch 154 Batch 60/162] avg loss 0.000931955, throughput 9.40797K wps
[Epoch 154 Batch 90/162] avg loss 0.000811779, throughput 9.32301K wps
[Epoch 154 Batch 120/162] avg loss 0.000976393, throughput 9.35001K wps
[Epoch 154 Batch 150/162] avg loss 0.000965315, throughput 9.42706K wps
Begin Testing...
[Epoch 154] train avg loss 0.000906859, dev acc 0.9211, dev avg loss 0.199789, throughput 9.41613K wps
[Epoch 155 Batch 30/162] avg loss 0.000817911, throughput 9.55407K wps
[Epoch 155 Batch 60/162] avg loss 0.00074818, throughput 9.28836K wps
[Epoch 155 Batch 90/162] avg loss 0.000833582, throughput 9.3469K wps
[Epoch 155 Batch 120/162] avg loss 0.00101265, throughput 9.48918K wps
[Epoch 155 Batch 150/162] avg loss 0.000799282, throughput 9.45709K wps
Begin Testing...
[Epoch 155] train avg loss 0.00085837, dev acc 0.9222, dev avg loss 0.199819, throughput 9.4363K wps
[Epoch 156 Batch 30/162] avg loss 0.000858828, throughput 9.78387K wps
[Epoch 156 Batch 60/162] avg loss 0.000903185, throughput 9.44814K wps
[Epoch 156 Batch 90/162] avg loss 0.000908508, throughput 9.38627K wps
[Epoch 156 Batch 120/162] avg loss 0.00084631, throughput 9.44584K wps
[Epoch 156 Batch 150/162] avg loss 0.000833211, throughput 9.44619K wps
Begin Testing...
[Epoch 156] train avg loss 0.00086402, dev acc 0.9211, dev avg loss 0.200072, throughput 9.47459K wps
[Epoch 157 Batch 30/162] avg loss 0.00086372, throughput 9.57542K wps
[Epoch 157 Batch 60/162] avg loss 0.00089777, throughput 9.47412K wps
[Epoch 157 Batch 90/162] avg loss 0.000891196, throughput 9.38632K wps
[Epoch 157 Batch 120/162] avg loss 0.000950652, throughput 9.42596K wps
[Epoch 157 Batch 150/162] avg loss 0.000803284, throughput 9.35706K wps
Begin Testing...
[Epoch 157] train avg loss 0.000883327, dev acc 0.9222, dev avg loss 0.199339, throughput 9.44445K wps
[Epoch 158 Batch 30/162] avg loss 0.00085479, throughput 9.60167K wps
[Epoch 158 Batch 60/162] avg loss 0.00084292, throughput 9.54283K wps
[Epoch 158 Batch 90/162] avg loss 0.000748971, throughput 9.67261K wps
[Epoch 158 Batch 120/162] avg loss 0.000813794, throughput 9.48769K wps
[Epoch 158 Batch 150/162] avg loss 0.000905792, throughput 9.44029K wps
Begin Testing...
[Epoch 158] train avg loss 0.000840809, dev acc 0.9189, dev avg loss 0.202241, throughput 9.54886K wps
[Epoch 159 Batch 30/162] avg loss 0.0007115, throughput 9.51376K wps
[Epoch 159 Batch 60/162] avg loss 0.000844186, throughput 9.4224K wps
[Epoch 159 Batch 90/162] avg loss 0.000883228, throughput 9.57975K wps
[Epoch 159 Batch 120/162] avg loss 0.000791884, throughput 9.27564K wps
[Epoch 159 Batch 150/162] avg loss 0.000965188, throughput 9.34714K wps
Begin Testing...
[Epoch 159] train avg loss 0.000828883, dev acc 0.9256, dev avg loss 0.199401, throughput 9.41943K wps
[Epoch 160 Batch 30/162] avg loss 0.000787447, throughput 9.59446K wps
[Epoch 160 Batch 60/162] avg loss 0.000758259, throughput 9.42134K wps
[Epoch 160 Batch 90/162] avg loss 0.000919504, throughput 9.43197K wps
[Epoch 160 Batch 120/162] avg loss 0.00084813, throughput 9.41282K wps
[Epoch 160 Batch 150/162] avg loss 0.000853881, throughput 9.59901K wps
Begin Testing...
[Epoch 160] train avg loss 0.000824844, dev acc 0.9211, dev avg loss 0.200824, throughput 9.46523K wps
[Epoch 161 Batch 30/162] avg loss 0.000744598, throughput 9.73636K wps
[Epoch 161 Batch 60/162] avg loss 0.000782382, throughput 9.49515K wps
[Epoch 161 Batch 90/162] avg loss 0.000900285, throughput 9.42103K wps
[Epoch 161 Batch 120/162] avg loss 0.000794527, throughput 9.4669K wps
[Epoch 161 Batch 150/162] avg loss 0.000820995, throughput 9.38735K wps
Begin Testing...
[Epoch 161] train avg loss 0.000806818, dev acc 0.9233, dev avg loss 0.199672, throughput 9.50153K wps
[Epoch 162 Batch 30/162] avg loss 0.000882261, throughput 9.63547K wps
[Epoch 162 Batch 60/162] avg loss 0.000727078, throughput 9.40971K wps
[Epoch 162 Batch 90/162] avg loss 0.000777933, throughput 9.4112K wps
[Epoch 162 Batch 120/162] avg loss 0.000857384, throughput 9.38293K wps
[Epoch 162 Batch 150/162] avg loss 0.000815616, throughput 9.47637K wps
Begin Testing...
[Epoch 162] train avg loss 0.00080976, dev acc 0.9267, dev avg loss 0.199484, throughput 9.45265K wps
Observed Improvement.
Begin Testing...
[Epoch 163 Batch 30/162] avg loss 0.00072524, throughput 9.63014K wps
[Epoch 163 Batch 60/162] avg loss 0.000830317, throughput 9.30393K wps
[Epoch 163 Batch 90/162] avg loss 0.000898723, throughput 9.34679K wps
[Epoch 163 Batch 120/162] avg loss 0.000770703, throughput 9.53268K wps
[Epoch 163 Batch 150/162] avg loss 0.000850974, throughput 9.41522K wps
Begin Testing...
[Epoch 163] train avg loss 0.000806451, dev acc 0.9256, dev avg loss 0.199551, throughput 9.42915K wps
[Epoch 164 Batch 30/162] avg loss 0.000694249, throughput 9.47889K wps
[Epoch 164 Batch 60/162] avg loss 0.000774463, throughput 9.40038K wps
[Epoch 164 Batch 90/162] avg loss 0.000716219, throughput 9.57985K wps
[Epoch 164 Batch 120/162] avg loss 0.000825106, throughput 9.50963K wps
[Epoch 164 Batch 150/162] avg loss 0.000835781, throughput 9.39779K wps
Begin Testing...
[Epoch 164] train avg loss 0.000773821, dev acc 0.9233, dev avg loss 0.200097, throughput 9.4674K wps
[Epoch 165 Batch 30/162] avg loss 0.000749146, throughput 9.60754K wps
[Epoch 165 Batch 60/162] avg loss 0.000748908, throughput 9.31985K wps
[Epoch 165 Batch 90/162] avg loss 0.00078026, throughput 9.39373K wps
[Epoch 165 Batch 120/162] avg loss 0.000801425, throughput 9.49352K wps
[Epoch 165 Batch 150/162] avg loss 0.000868468, throughput 9.24188K wps
Begin Testing...
[Epoch 165] train avg loss 0.000793269, dev acc 0.9244, dev avg loss 0.199631, throughput 9.40458K wps
[Epoch 166 Batch 30/162] avg loss 0.00075902, throughput 9.6779K wps
[Epoch 166 Batch 60/162] avg loss 0.000728, throughput 9.53675K wps
[Epoch 166 Batch 90/162] avg loss 0.000790671, throughput 9.31952K wps
[Epoch 166 Batch 120/162] avg loss 0.00074247, throughput 9.25501K wps
[Epoch 166 Batch 150/162] avg loss 0.000810725, throughput 9.31378K wps
Begin Testing...
[Epoch 166] train avg loss 0.000772062, dev acc 0.9256, dev avg loss 0.199701, throughput 9.41588K wps
[Epoch 167 Batch 30/162] avg loss 0.000766148, throughput 9.55133K wps
[Epoch 167 Batch 60/162] avg loss 0.000831309, throughput 9.56734K wps
[Epoch 167 Batch 90/162] avg loss 0.000840688, throughput 9.39538K wps
[Epoch 167 Batch 120/162] avg loss 0.00081, throughput 9.43078K wps
[Epoch 167 Batch 150/162] avg loss 0.000752378, throughput 9.50785K wps
Begin Testing...
[Epoch 167] train avg loss 0.000789774, dev acc 0.9244, dev avg loss 0.2001, throughput 9.50093K wps
[Epoch 168 Batch 30/162] avg loss 0.000789058, throughput 9.74463K wps
[Epoch 168 Batch 60/162] avg loss 0.000760805, throughput 9.40672K wps
[Epoch 168 Batch 90/162] avg loss 0.000727514, throughput 9.39256K wps
[Epoch 168 Batch 120/162] avg loss 0.000677497, throughput 9.49565K wps
[Epoch 168 Batch 150/162] avg loss 0.000727855, throughput 9.24238K wps
Begin Testing...
[Epoch 168] train avg loss 0.000739517, dev acc 0.9222, dev avg loss 0.201245, throughput 9.45834K wps
[Epoch 169 Batch 30/162] avg loss 0.000833044, throughput 9.7503K wps
[Epoch 169 Batch 60/162] avg loss 0.000916314, throughput 9.30996K wps
[Epoch 169 Batch 90/162] avg loss 0.000838302, throughput 9.4012K wps
[Epoch 169 Batch 120/162] avg loss 0.000727336, throughput 9.62459K wps
[Epoch 169 Batch 150/162] avg loss 0.000876009, throughput 9.38613K wps
Begin Testing...
[Epoch 169] train avg loss 0.000830419, dev acc 0.9233, dev avg loss 0.199889, throughput 9.48211K wps
[Epoch 170 Batch 30/162] avg loss 0.000787692, throughput 9.48405K wps
[Epoch 170 Batch 60/162] avg loss 0.00073441, throughput 9.37826K wps
[Epoch 170 Batch 90/162] avg loss 0.000827956, throughput 9.29571K wps
[Epoch 170 Batch 120/162] avg loss 0.000768396, throughput 9.28597K wps
[Epoch 170 Batch 150/162] avg loss 0.000679814, throughput 9.23183K wps
Begin Testing...
[Epoch 170] train avg loss 0.00076412, dev acc 0.9244, dev avg loss 0.200318, throughput 9.33321K wps
[Epoch 171 Batch 30/162] avg loss 0.000718235, throughput 9.53152K wps
[Epoch 171 Batch 60/162] avg loss 0.000772861, throughput 9.37812K wps
[Epoch 171 Batch 90/162] avg loss 0.000739895, throughput 9.58111K wps
[Epoch 171 Batch 120/162] avg loss 0.000677051, throughput 9.4181K wps
[Epoch 171 Batch 150/162] avg loss 0.000809539, throughput 9.59466K wps
Begin Testing...
[Epoch 171] train avg loss 0.000745079, dev acc 0.9244, dev avg loss 0.200677, throughput 9.47914K wps
[Epoch 172 Batch 30/162] avg loss 0.000685703, throughput 9.58633K wps
[Epoch 172 Batch 60/162] avg loss 0.000790037, throughput 9.53276K wps
[Epoch 172 Batch 90/162] avg loss 0.00071011, throughput 9.44027K wps
[Epoch 172 Batch 120/162] avg loss 0.000718413, throughput 9.50613K wps
[Epoch 172 Batch 150/162] avg loss 0.000678016, throughput 9.16429K wps
Begin Testing...
[Epoch 172] train avg loss 0.000719593, dev acc 0.9233, dev avg loss 0.201643, throughput 9.44835K wps
[Epoch 173 Batch 30/162] avg loss 0.000608755, throughput 9.7805K wps
[Epoch 173 Batch 60/162] avg loss 0.000754511, throughput 9.51477K wps
[Epoch 173 Batch 90/162] avg loss 0.000763572, throughput 9.37372K wps
[Epoch 173 Batch 120/162] avg loss 0.000708415, throughput 9.585K wps
[Epoch 173 Batch 150/162] avg loss 0.000661871, throughput 9.45717K wps
Begin Testing...
[Epoch 173] train avg loss 0.000708725, dev acc 0.9222, dev avg loss 0.201403, throughput 9.53569K wps
[Epoch 174 Batch 30/162] avg loss 0.000659929, throughput 9.76009K wps
[Epoch 174 Batch 60/162] avg loss 0.000719012, throughput 9.32172K wps
[Epoch 174 Batch 90/162] avg loss 0.000779537, throughput 9.46468K wps
[Epoch 174 Batch 120/162] avg loss 0.000703371, throughput 9.57703K wps
[Epoch 174 Batch 150/162] avg loss 0.000869526, throughput 9.44914K wps
Begin Testing...
[Epoch 174] train avg loss 0.0007496, dev acc 0.9233, dev avg loss 0.201453, throughput 9.49778K wps
[Epoch 175 Batch 30/162] avg loss 0.000886255, throughput 9.45866K wps
[Epoch 175 Batch 60/162] avg loss 0.000728094, throughput 9.31166K wps
[Epoch 175 Batch 90/162] avg loss 0.000722166, throughput 9.33195K wps
[Epoch 175 Batch 120/162] avg loss 0.000631821, throughput 9.29576K wps
[Epoch 175 Batch 150/162] avg loss 0.000671513, throughput 9.35121K wps
Begin Testing...
[Epoch 175] train avg loss 0.000724313, dev acc 0.9244, dev avg loss 0.20069, throughput 9.34115K wps
[Epoch 176 Batch 30/162] avg loss 0.000694296, throughput 9.51019K wps
[Epoch 176 Batch 60/162] avg loss 0.000753325, throughput 9.4155K wps
[Epoch 176 Batch 90/162] avg loss 0.000705295, throughput 9.45234K wps
[Epoch 176 Batch 120/162] avg loss 0.000713599, throughput 9.5297K wps
[Epoch 176 Batch 150/162] avg loss 0.000662685, throughput 9.28363K wps
Begin Testing...
[Epoch 176] train avg loss 0.000704863, dev acc 0.9256, dev avg loss 0.20059, throughput 9.42809K wps
[Epoch 177 Batch 30/162] avg loss 0.000700139, throughput 9.54879K wps
[Epoch 177 Batch 60/162] avg loss 0.000716132, throughput 9.48013K wps
[Epoch 177 Batch 90/162] avg loss 0.00070663, throughput 9.46375K wps
[Epoch 177 Batch 120/162] avg loss 0.0007473, throughput 9.55563K wps
[Epoch 177 Batch 150/162] avg loss 0.00075779, throughput 9.37742K wps
Begin Testing...
[Epoch 177] train avg loss 0.000720801, dev acc 0.9244, dev avg loss 0.200838, throughput 9.48429K wps
[Epoch 178 Batch 30/162] avg loss 0.000730782, throughput 9.51219K wps
[Epoch 178 Batch 60/162] avg loss 0.000641114, throughput 9.61443K wps
[Epoch 178 Batch 90/162] avg loss 0.000798168, throughput 9.44486K wps
[Epoch 178 Batch 120/162] avg loss 0.000659751, throughput 9.41798K wps
[Epoch 178 Batch 150/162] avg loss 0.000647889, throughput 9.54781K wps
Begin Testing...
[Epoch 178] train avg loss 0.000695505, dev acc 0.9244, dev avg loss 0.201376, throughput 9.49313K wps
[Epoch 179 Batch 30/162] avg loss 0.000699846, throughput 9.71583K wps
[Epoch 179 Batch 60/162] avg loss 0.000599674, throughput 9.2814K wps
[Epoch 179 Batch 90/162] avg loss 0.000644186, throughput 9.34344K wps
[Epoch 179 Batch 120/162] avg loss 0.000631521, throughput 9.3782K wps
[Epoch 179 Batch 150/162] avg loss 0.000763138, throughput 9.61594K wps
Begin Testing...
[Epoch 179] train avg loss 0.000667105, dev acc 0.9233, dev avg loss 0.202039, throughput 9.44612K wps
[Epoch 180 Batch 30/162] avg loss 0.000653926, throughput 9.71603K wps
[Epoch 180 Batch 60/162] avg loss 0.000611203, throughput 9.34385K wps
[Epoch 180 Batch 90/162] avg loss 0.0007032, throughput 9.4824K wps
[Epoch 180 Batch 120/162] avg loss 0.000777208, throughput 9.26851K wps
[Epoch 180 Batch 150/162] avg loss 0.000718971, throughput 9.38206K wps
Begin Testing...
[Epoch 180] train avg loss 0.000688757, dev acc 0.9278, dev avg loss 0.200694, throughput 9.45252K wps
Observed Improvement.
Begin Testing...
[Epoch 181 Batch 30/162] avg loss 0.000720283, throughput 9.44112K wps
[Epoch 181 Batch 60/162] avg loss 0.000691045, throughput 9.38024K wps
[Epoch 181 Batch 90/162] avg loss 0.000639149, throughput 9.49512K wps
[Epoch 181 Batch 120/162] avg loss 0.000694108, throughput 9.45787K wps
[Epoch 181 Batch 150/162] avg loss 0.000657934, throughput 9.37029K wps
Begin Testing...
[Epoch 181] train avg loss 0.000679076, dev acc 0.9244, dev avg loss 0.201125, throughput 9.41363K wps
[Epoch 182 Batch 30/162] avg loss 0.000634448, throughput 9.67381K wps
[Epoch 182 Batch 60/162] avg loss 0.000579765, throughput 9.20359K wps
[Epoch 182 Batch 90/162] avg loss 0.000721314, throughput 9.61781K wps
[Epoch 182 Batch 120/162] avg loss 0.000640318, throughput 9.39113K wps
[Epoch 182 Batch 150/162] avg loss 0.000698275, throughput 9.60175K wps
Begin Testing...
[Epoch 182] train avg loss 0.000649868, dev acc 0.9244, dev avg loss 0.201647, throughput 9.46895K wps
[Epoch 183 Batch 30/162] avg loss 0.000713188, throughput 9.67285K wps
[Epoch 183 Batch 60/162] avg loss 0.000607393, throughput 9.33653K wps
[Epoch 183 Batch 90/162] avg loss 0.000618886, throughput 9.35772K wps
[Epoch 183 Batch 120/162] avg loss 0.000684018, throughput 9.61838K wps
[Epoch 183 Batch 150/162] avg loss 0.000736899, throughput 9.45701K wps
Begin Testing...
[Epoch 183] train avg loss 0.000669464, dev acc 0.9233, dev avg loss 0.201248, throughput 9.46556K wps
[Epoch 184 Batch 30/162] avg loss 0.000650741, throughput 9.69982K wps
[Epoch 184 Batch 60/162] avg loss 0.000657959, throughput 9.32394K wps
[Epoch 184 Batch 90/162] avg loss 0.000664686, throughput 9.43441K wps
[Epoch 184 Batch 120/162] avg loss 0.000655762, throughput 9.27351K wps
[Epoch 184 Batch 150/162] avg loss 0.000570726, throughput 9.63358K wps
Begin Testing...
[Epoch 184] train avg loss 0.000641091, dev acc 0.9244, dev avg loss 0.201742, throughput 9.43354K wps
[Epoch 185 Batch 30/162] avg loss 0.000676742, throughput 9.79134K wps
[Epoch 185 Batch 60/162] avg loss 0.000551051, throughput 9.28449K wps
[Epoch 185 Batch 90/162] avg loss 0.000685884, throughput 9.48081K wps
[Epoch 185 Batch 120/162] avg loss 0.000626526, throughput 9.3924K wps
[Epoch 185 Batch 150/162] avg loss 0.000610767, throughput 9.46619K wps
Begin Testing...
[Epoch 185] train avg loss 0.000627161, dev acc 0.9244, dev avg loss 0.201371, throughput 9.47731K wps
[Epoch 186 Batch 30/162] avg loss 0.00059824, throughput 9.65117K wps
[Epoch 186 Batch 60/162] avg loss 0.000748581, throughput 9.31883K wps
[Epoch 186 Batch 90/162] avg loss 0.000641093, throughput 9.42148K wps
[Epoch 186 Batch 120/162] avg loss 0.000579003, throughput 9.55746K wps
[Epoch 186 Batch 150/162] avg loss 0.000697524, throughput 9.43642K wps
Begin Testing...
[Epoch 186] train avg loss 0.000647043, dev acc 0.9244, dev avg loss 0.201418, throughput 9.45515K wps
[Epoch 187 Batch 30/162] avg loss 0.000665172, throughput 9.6867K wps
[Epoch 187 Batch 60/162] avg loss 0.000687576, throughput 9.3475K wps
[Epoch 187 Batch 90/162] avg loss 0.00061594, throughput 9.41102K wps
[Epoch 187 Batch 120/162] avg loss 0.000662239, throughput 9.53247K wps
[Epoch 187 Batch 150/162] avg loss 0.00060423, throughput 9.55254K wps
Begin Testing...
[Epoch 187] train avg loss 0.000640293, dev acc 0.9244, dev avg loss 0.201405, throughput 9.48942K wps
[Epoch 188 Batch 30/162] avg loss 0.000632196, throughput 9.37958K wps
[Epoch 188 Batch 60/162] avg loss 0.000648961, throughput 9.29563K wps
[Epoch 188 Batch 90/162] avg loss 0.000603649, throughput 9.54751K wps
[Epoch 188 Batch 120/162] avg loss 0.000651942, throughput 9.39886K wps
[Epoch 188 Batch 150/162] avg loss 0.000603888, throughput 9.40534K wps
Begin Testing...
[Epoch 188] train avg loss 0.000627734, dev acc 0.9267, dev avg loss 0.20126, throughput 9.413K wps
[Epoch 189 Batch 30/162] avg loss 0.00068028, throughput 9.71252K wps
[Epoch 189 Batch 60/162] avg loss 0.000610199, throughput 9.24677K wps
[Epoch 189 Batch 90/162] avg loss 0.000640021, throughput 9.33184K wps
[Epoch 189 Batch 120/162] avg loss 0.000619203, throughput 9.34236K wps
[Epoch 189 Batch 150/162] avg loss 0.000750619, throughput 9.28939K wps
Begin Testing...
[Epoch 189] train avg loss 0.000648586, dev acc 0.9233, dev avg loss 0.202591, throughput 9.38351K wps
[Epoch 190 Batch 30/162] avg loss 0.00057054, throughput 9.71518K wps
[Epoch 190 Batch 60/162] avg loss 0.00059983, throughput 9.41088K wps
[Epoch 190 Batch 90/162] avg loss 0.000577179, throughput 9.41536K wps
[Epoch 190 Batch 120/162] avg loss 0.00068698, throughput 9.57264K wps
[Epoch 190 Batch 150/162] avg loss 0.000657615, throughput 9.49624K wps
Begin Testing...
[Epoch 190] train avg loss 0.000610008, dev acc 0.9233, dev avg loss 0.202553, throughput 9.52627K wps
[Epoch 191 Batch 30/162] avg loss 0.000670352, throughput 9.43396K wps
[Epoch 191 Batch 60/162] avg loss 0.000624464, throughput 9.27549K wps
[Epoch 191 Batch 90/162] avg loss 0.000614721, throughput 9.43273K wps
[Epoch 191 Batch 120/162] avg loss 0.000663615, throughput 9.24872K wps
[Epoch 191 Batch 150/162] avg loss 0.00053307, throughput 9.42324K wps
Begin Testing...
[Epoch 191] train avg loss 0.00062837, dev acc 0.9222, dev avg loss 0.203794, throughput 9.35788K wps
[Epoch 192 Batch 30/162] avg loss 0.000656959, throughput 9.48227K wps
[Epoch 192 Batch 60/162] avg loss 0.00063398, throughput 9.46438K wps
[Epoch 192 Batch 90/162] avg loss 0.000536834, throughput 9.37685K wps
[Epoch 192 Batch 120/162] avg loss 0.000597453, throughput 9.3603K wps
[Epoch 192 Batch 150/162] avg loss 0.000661628, throughput 9.34268K wps
Begin Testing...
[Epoch 192] train avg loss 0.000610057, dev acc 0.9244, dev avg loss 0.202969, throughput 9.396K wps
[Epoch 193 Batch 30/162] avg loss 0.000604095, throughput 9.51263K wps
[Epoch 193 Batch 60/162] avg loss 0.000572264, throughput 9.39626K wps
[Epoch 193 Batch 90/162] avg loss 0.000620527, throughput 9.21785K wps
[Epoch 193 Batch 120/162] avg loss 0.000616928, throughput 9.36852K wps
[Epoch 193 Batch 150/162] avg loss 0.00061055, throughput 9.42404K wps
Begin Testing...
[Epoch 193] train avg loss 0.000610511, dev acc 0.9256, dev avg loss 0.20311, throughput 9.37992K wps
[Epoch 194 Batch 30/162] avg loss 0.000638084, throughput 9.62626K wps
[Epoch 194 Batch 60/162] avg loss 0.000640144, throughput 9.30426K wps
[Epoch 194 Batch 90/162] avg loss 0.000637773, throughput 9.46358K wps
[Epoch 194 Batch 120/162] avg loss 0.000662084, throughput 9.27644K wps
[Epoch 194 Batch 150/162] avg loss 0.000666145, throughput 9.38359K wps
Begin Testing...
[Epoch 194] train avg loss 0.00063943, dev acc 0.9244, dev avg loss 0.203001, throughput 9.41595K wps
[Epoch 195 Batch 30/162] avg loss 0.000561682, throughput 9.62772K wps
[Epoch 195 Batch 60/162] avg loss 0.000591931, throughput 9.23939K wps
[Epoch 195 Batch 90/162] avg loss 0.000728222, throughput 9.34173K wps
[Epoch 195 Batch 120/162] avg loss 0.000604459, throughput 9.41145K wps
[Epoch 195 Batch 150/162] avg loss 0.00060469, throughput 9.42237K wps
Begin Testing...
[Epoch 195] train avg loss 0.000622859, dev acc 0.9244, dev avg loss 0.202507, throughput 9.40891K wps
[Epoch 196 Batch 30/162] avg loss 0.000575882, throughput 9.61991K wps
[Epoch 196 Batch 60/162] avg loss 0.000620248, throughput 9.47817K wps
[Epoch 196 Batch 90/162] avg loss 0.000659196, throughput 9.33449K wps
[Epoch 196 Batch 120/162] avg loss 0.000501042, throughput 9.31091K wps
[Epoch 196 Batch 150/162] avg loss 0.0005818, throughput 9.49988K wps
Begin Testing...
[Epoch 196] train avg loss 0.000588459, dev acc 0.9256, dev avg loss 0.20278, throughput 9.45574K wps
[Epoch 197 Batch 30/162] avg loss 0.000525616, throughput 9.72667K wps
[Epoch 197 Batch 60/162] avg loss 0.000638361, throughput 9.59348K wps
[Epoch 197 Batch 90/162] avg loss 0.000572691, throughput 9.36773K wps
[Epoch 197 Batch 120/162] avg loss 0.000603652, throughput 9.53931K wps
[Epoch 197 Batch 150/162] avg loss 0.000628934, throughput 9.34071K wps
Begin Testing...
[Epoch 197] train avg loss 0.00059507, dev acc 0.9256, dev avg loss 0.202141, throughput 9.5038K wps
[Epoch 198 Batch 30/162] avg loss 0.000562714, throughput 9.51776K wps
[Epoch 198 Batch 60/162] avg loss 0.000542279, throughput 9.30895K wps
[Epoch 198 Batch 90/162] avg loss 0.00053567, throughput 9.37371K wps
[Epoch 198 Batch 120/162] avg loss 0.000517381, throughput 9.31743K wps
[Epoch 198 Batch 150/162] avg loss 0.00063553, throughput 9.4898K wps
Begin Testing...
[Epoch 198] train avg loss 0.000564634, dev acc 0.9256, dev avg loss 0.20227, throughput 9.39752K wps
[Epoch 199 Batch 30/162] avg loss 0.000563386, throughput 9.6886K wps
[Epoch 199 Batch 60/162] avg loss 0.00053364, throughput 9.23429K wps
[Epoch 199 Batch 90/162] avg loss 0.000577254, throughput 9.34716K wps
[Epoch 199 Batch 120/162] avg loss 0.000624464, throughput 9.46508K wps
[Epoch 199 Batch 150/162] avg loss 0.000591925, throughput 9.35146K wps
Begin Testing...
[Epoch 199] train avg loss 0.000568744, dev acc 0.9244, dev avg loss 0.20317, throughput 9.437K wps
Test loss 0.211503, test acc 0.9150
Total time cost 442.42s
[Epoch 0 Batch 30/162] avg loss 0.0140422, throughput 7.18286K wps
[Epoch 0 Batch 60/162] avg loss 0.0137961, throughput 9.2935K wps
[Epoch 0 Batch 90/162] avg loss 0.0136954, throughput 9.31835K wps
[Epoch 0 Batch 120/162] avg loss 0.0137202, throughput 9.50034K wps
[Epoch 0 Batch 150/162] avg loss 0.0134267, throughput 9.43002K wps
Begin Testing...
[Epoch 0] train avg loss 0.0137218, dev acc 0.7244, dev avg loss 0.667148, throughput 8.89527K wps
Observed Improvement.
Begin Testing...
[Epoch 1 Batch 30/162] avg loss 0.0134143, throughput 9.4478K wps
[Epoch 1 Batch 60/162] avg loss 0.0132198, throughput 9.40328K wps
[Epoch 1 Batch 90/162] avg loss 0.0131083, throughput 9.30334K wps
[Epoch 1 Batch 120/162] avg loss 0.0130728, throughput 9.37577K wps
[Epoch 1 Batch 150/162] avg loss 0.0129009, throughput 9.34432K wps
Begin Testing...
[Epoch 1] train avg loss 0.0131175, dev acc 0.8033, dev avg loss 0.638434, throughput 9.37413K wps
Observed Improvement.
Begin Testing...
[Epoch 2 Batch 30/162] avg loss 0.012919, throughput 9.62073K wps
[Epoch 2 Batch 60/162] avg loss 0.012632, throughput 9.37491K wps
[Epoch 2 Batch 90/162] avg loss 0.0125039, throughput 9.47085K wps
[Epoch 2 Batch 120/162] avg loss 0.012385, throughput 9.34322K wps
[Epoch 2 Batch 150/162] avg loss 0.0123177, throughput 9.51948K wps
Begin Testing...
[Epoch 2] train avg loss 0.0125301, dev acc 0.8233, dev avg loss 0.607795, throughput 9.46806K wps
Observed Improvement.
Begin Testing...
[Epoch 3 Batch 30/162] avg loss 0.0122125, throughput 9.50629K wps
[Epoch 3 Batch 60/162] avg loss 0.0118965, throughput 9.42116K wps
[Epoch 3 Batch 90/162] avg loss 0.0118693, throughput 9.31045K wps
[Epoch 3 Batch 120/162] avg loss 0.0116506, throughput 9.39055K wps
[Epoch 3 Batch 150/162] avg loss 0.0115358, throughput 9.46591K wps
Begin Testing...
[Epoch 3] train avg loss 0.0118186, dev acc 0.8322, dev avg loss 0.572309, throughput 9.42468K wps
Observed Improvement.
Begin Testing...
[Epoch 4 Batch 30/162] avg loss 0.0114715, throughput 9.5694K wps
[Epoch 4 Batch 60/162] avg loss 0.0111401, throughput 9.40126K wps
[Epoch 4 Batch 90/162] avg loss 0.0109132, throughput 9.49637K wps
[Epoch 4 Batch 120/162] avg loss 0.0109563, throughput 9.34319K wps
[Epoch 4 Batch 150/162] avg loss 0.0108461, throughput 9.36171K wps
Begin Testing...
[Epoch 4] train avg loss 0.0110386, dev acc 0.8467, dev avg loss 0.531816, throughput 9.44777K wps
Observed Improvement.
Begin Testing...
[Epoch 5 Batch 30/162] avg loss 0.0104467, throughput 9.7223K wps
[Epoch 5 Batch 60/162] avg loss 0.0104472, throughput 9.36977K wps
[Epoch 5 Batch 90/162] avg loss 0.0103339, throughput 9.41997K wps
[Epoch 5 Batch 120/162] avg loss 0.010211, throughput 9.46882K wps
[Epoch 5 Batch 150/162] avg loss 0.00991035, throughput 9.3333K wps
Begin Testing...
[Epoch 5] train avg loss 0.0102499, dev acc 0.8544, dev avg loss 0.491442, throughput 9.46384K wps
Observed Improvement.
Begin Testing...
[Epoch 6 Batch 30/162] avg loss 0.00973124, throughput 9.63358K wps
[Epoch 6 Batch 60/162] avg loss 0.00976641, throughput 9.24886K wps
[Epoch 6 Batch 90/162] avg loss 0.00948293, throughput 9.55239K wps
[Epoch 6 Batch 120/162] avg loss 0.00937961, throughput 9.33743K wps
[Epoch 6 Batch 150/162] avg loss 0.0092074, throughput 9.38974K wps
Begin Testing...
[Epoch 6] train avg loss 0.00948418, dev acc 0.8600, dev avg loss 0.454412, throughput 9.4168K wps
Observed Improvement.
Begin Testing...
[Epoch 7 Batch 30/162] avg loss 0.00870234, throughput 9.58873K wps
[Epoch 7 Batch 60/162] avg loss 0.00898772, throughput 9.28199K wps
[Epoch 7 Batch 90/162] avg loss 0.00875899, throughput 9.39097K wps
[Epoch 7 Batch 120/162] avg loss 0.00878208, throughput 9.50732K wps
[Epoch 7 Batch 150/162] avg loss 0.00854325, throughput 9.43702K wps
Begin Testing...
[Epoch 7] train avg loss 0.00874892, dev acc 0.8544, dev avg loss 0.425357, throughput 9.42555K wps
[Epoch 8 Batch 30/162] avg loss 0.00843247, throughput 9.63573K wps
[Epoch 8 Batch 60/162] avg loss 0.0082087, throughput 9.25327K wps
[Epoch 8 Batch 90/162] avg loss 0.0082209, throughput 9.42846K wps
[Epoch 8 Batch 120/162] avg loss 0.00817901, throughput 9.37656K wps
[Epoch 8 Batch 150/162] avg loss 0.00783102, throughput 9.48734K wps
Begin Testing...
[Epoch 8] train avg loss 0.00812788, dev acc 0.8567, dev avg loss 0.397434, throughput 9.44121K wps
[Epoch 9 Batch 30/162] avg loss 0.00805778, throughput 9.70555K wps
[Epoch 9 Batch 60/162] avg loss 0.00751204, throughput 9.45242K wps
[Epoch 9 Batch 90/162] avg loss 0.007871, throughput 9.47295K wps
[Epoch 9 Batch 120/162] avg loss 0.0076041, throughput 9.5428K wps
[Epoch 9 Batch 150/162] avg loss 0.00743874, throughput 9.35692K wps
Begin Testing...
[Epoch 9] train avg loss 0.00770218, dev acc 0.8633, dev avg loss 0.37823, throughput 9.48747K wps
Observed Improvement.
Begin Testing...
[Epoch 10 Batch 30/162] avg loss 0.00741356, throughput 9.67433K wps
[Epoch 10 Batch 60/162] avg loss 0.00719929, throughput 9.29259K wps
[Epoch 10 Batch 90/162] avg loss 0.00758783, throughput 9.57224K wps
[Epoch 10 Batch 120/162] avg loss 0.00736084, throughput 9.34534K wps
[Epoch 10 Batch 150/162] avg loss 0.00705627, throughput 9.42105K wps
Begin Testing...
[Epoch 10] train avg loss 0.00731359, dev acc 0.8600, dev avg loss 0.363265, throughput 9.44943K wps
[Epoch 11 Batch 30/162] avg loss 0.00733745, throughput 9.47408K wps
[Epoch 11 Batch 60/162] avg loss 0.00684512, throughput 9.36401K wps
[Epoch 11 Batch 90/162] avg loss 0.00712059, throughput 9.59248K wps
[Epoch 11 Batch 120/162] avg loss 0.00664513, throughput 9.51544K wps
[Epoch 11 Batch 150/162] avg loss 0.00696167, throughput 9.3919K wps
Begin Testing...
[Epoch 11] train avg loss 0.00698996, dev acc 0.8611, dev avg loss 0.349674, throughput 9.48157K wps
[Epoch 12 Batch 30/162] avg loss 0.00675449, throughput 9.71543K wps
[Epoch 12 Batch 60/162] avg loss 0.00678367, throughput 9.36683K wps
[Epoch 12 Batch 90/162] avg loss 0.00648105, throughput 9.457K wps
[Epoch 12 Batch 120/162] avg loss 0.00675374, throughput 9.36225K wps
[Epoch 12 Batch 150/162] avg loss 0.00689043, throughput 9.45292K wps
Begin Testing...
[Epoch 12] train avg loss 0.00673004, dev acc 0.8700, dev avg loss 0.338352, throughput 9.46229K wps
Observed Improvement.
Begin Testing...
[Epoch 13 Batch 30/162] avg loss 0.00676001, throughput 9.5901K wps
[Epoch 13 Batch 60/162] avg loss 0.00653463, throughput 9.45419K wps
[Epoch 13 Batch 90/162] avg loss 0.00618469, throughput 9.56463K wps
[Epoch 13 Batch 120/162] avg loss 0.0062944, throughput 9.37753K wps
[Epoch 13 Batch 150/162] avg loss 0.00664608, throughput 9.53421K wps
Begin Testing...
[Epoch 13] train avg loss 0.00646771, dev acc 0.8700, dev avg loss 0.332173, throughput 9.51703K wps
Observed Improvement.
Begin Testing...
[Epoch 14 Batch 30/162] avg loss 0.00585164, throughput 9.60878K wps
[Epoch 14 Batch 60/162] avg loss 0.00643956, throughput 9.26558K wps
[Epoch 14 Batch 90/162] avg loss 0.00634607, throughput 9.28868K wps
[Epoch 14 Batch 120/162] avg loss 0.00642644, throughput 9.3199K wps
[Epoch 14 Batch 150/162] avg loss 0.00621378, throughput 9.13756K wps
Begin Testing...
[Epoch 14] train avg loss 0.00630384, dev acc 0.8700, dev avg loss 0.323646, throughput 9.35257K wps
Observed Improvement.
Begin Testing...
[Epoch 15 Batch 30/162] avg loss 0.00609484, throughput 9.54582K wps
[Epoch 15 Batch 60/162] avg loss 0.00606797, throughput 9.25482K wps
[Epoch 15 Batch 90/162] avg loss 0.00605409, throughput 9.44378K wps
[Epoch 15 Batch 120/162] avg loss 0.00635318, throughput 9.45723K wps
[Epoch 15 Batch 150/162] avg loss 0.00600655, throughput 9.56357K wps
Begin Testing...
[Epoch 15] train avg loss 0.00611124, dev acc 0.8733, dev avg loss 0.315087, throughput 9.4606K wps
Observed Improvement.
Begin Testing...
[Epoch 16 Batch 30/162] avg loss 0.00600658, throughput 9.71563K wps
[Epoch 16 Batch 60/162] avg loss 0.00592784, throughput 9.44696K wps
[Epoch 16 Batch 90/162] avg loss 0.00576932, throughput 9.5335K wps
[Epoch 16 Batch 120/162] avg loss 0.00575565, throughput 9.53378K wps
[Epoch 16 Batch 150/162] avg loss 0.00609783, throughput 9.46803K wps
Begin Testing...
[Epoch 16] train avg loss 0.00594063, dev acc 0.8744, dev avg loss 0.309239, throughput 9.53321K wps
Observed Improvement.
Begin Testing...
[Epoch 17 Batch 30/162] avg loss 0.00597555, throughput 9.42816K wps
[Epoch 17 Batch 60/162] avg loss 0.00582374, throughput 9.43046K wps
[Epoch 17 Batch 90/162] avg loss 0.0058241, throughput 9.53879K wps
[Epoch 17 Batch 120/162] avg loss 0.00561988, throughput 9.60495K wps
[Epoch 17 Batch 150/162] avg loss 0.0059051, throughput 9.48598K wps
Begin Testing...
[Epoch 17] train avg loss 0.00580733, dev acc 0.8756, dev avg loss 0.305182, throughput 9.50095K wps
Observed Improvement.
Begin Testing...
[Epoch 18 Batch 30/162] avg loss 0.00554669, throughput 9.52198K wps
[Epoch 18 Batch 60/162] avg loss 0.00561088, throughput 9.41543K wps
[Epoch 18 Batch 90/162] avg loss 0.00572496, throughput 9.52233K wps
[Epoch 18 Batch 120/162] avg loss 0.00577037, throughput 9.38488K wps
[Epoch 18 Batch 150/162] avg loss 0.00561514, throughput 9.4988K wps
Begin Testing...
[Epoch 18] train avg loss 0.00567987, dev acc 0.8800, dev avg loss 0.3005, throughput 9.47503K wps
Observed Improvement.
Begin Testing...
[Epoch 19 Batch 30/162] avg loss 0.00532531, throughput 9.46315K wps
[Epoch 19 Batch 60/162] avg loss 0.00570328, throughput 9.48452K wps
[Epoch 19 Batch 90/162] avg loss 0.00548202, throughput 9.4839K wps
[Epoch 19 Batch 120/162] avg loss 0.00563729, throughput 9.3246K wps
[Epoch 19 Batch 150/162] avg loss 0.00567315, throughput 9.33802K wps
Begin Testing...
[Epoch 19] train avg loss 0.00555215, dev acc 0.8822, dev avg loss 0.297874, throughput 9.40932K wps
Observed Improvement.
Begin Testing...
[Epoch 20 Batch 30/162] avg loss 0.00560229, throughput 9.72466K wps
[Epoch 20 Batch 60/162] avg loss 0.00535057, throughput 9.27526K wps
[Epoch 20 Batch 90/162] avg loss 0.00537177, throughput 9.6006K wps
[Epoch 20 Batch 120/162] avg loss 0.00548621, throughput 9.30199K wps
[Epoch 20 Batch 150/162] avg loss 0.00533787, throughput 9.60057K wps
Begin Testing...
[Epoch 20] train avg loss 0.00540368, dev acc 0.8800, dev avg loss 0.293108, throughput 9.50344K wps
[Epoch 21 Batch 30/162] avg loss 0.00494786, throughput 9.63363K wps
[Epoch 21 Batch 60/162] avg loss 0.00588408, throughput 9.5877K wps
[Epoch 21 Batch 90/162] avg loss 0.00506719, throughput 9.42665K wps
[Epoch 21 Batch 120/162] avg loss 0.00551935, throughput 9.44661K wps
[Epoch 21 Batch 150/162] avg loss 0.00528133, throughput 9.29483K wps
Begin Testing...
[Epoch 21] train avg loss 0.00535581, dev acc 0.8844, dev avg loss 0.289087, throughput 9.44779K wps
Observed Improvement.
Begin Testing...
[Epoch 22 Batch 30/162] avg loss 0.00519827, throughput 9.57579K wps
[Epoch 22 Batch 60/162] avg loss 0.00502786, throughput 9.33691K wps
[Epoch 22 Batch 90/162] avg loss 0.00532887, throughput 9.34608K wps
[Epoch 22 Batch 120/162] avg loss 0.00517759, throughput 9.35607K wps
[Epoch 22 Batch 150/162] avg loss 0.00521812, throughput 9.30894K wps
Begin Testing...
[Epoch 22] train avg loss 0.00520637, dev acc 0.8822, dev avg loss 0.286546, throughput 9.36888K wps
[Epoch 23 Batch 30/162] avg loss 0.00523582, throughput 9.65187K wps
[Epoch 23 Batch 60/162] avg loss 0.00522421, throughput 9.59958K wps
[Epoch 23 Batch 90/162] avg loss 0.00541408, throughput 9.48092K wps
[Epoch 23 Batch 120/162] avg loss 0.00481433, throughput 9.41624K wps
[Epoch 23 Batch 150/162] avg loss 0.00455338, throughput 9.31324K wps
Begin Testing...
[Epoch 23] train avg loss 0.00503948, dev acc 0.8833, dev avg loss 0.283101, throughput 9.49033K wps
[Epoch 24 Batch 30/162] avg loss 0.00487639, throughput 9.51307K wps
[Epoch 24 Batch 60/162] avg loss 0.0049673, throughput 9.3953K wps
[Epoch 24 Batch 90/162] avg loss 0.00492247, throughput 9.34049K wps
[Epoch 24 Batch 120/162] avg loss 0.00517346, throughput 9.40292K wps
[Epoch 24 Batch 150/162] avg loss 0.00499024, throughput 9.54821K wps
Begin Testing...
[Epoch 24] train avg loss 0.00497388, dev acc 0.8844, dev avg loss 0.280353, throughput 9.4196K wps
Observed Improvement.
Begin Testing...
[Epoch 25 Batch 30/162] avg loss 0.0049576, throughput 9.59562K wps
[Epoch 25 Batch 60/162] avg loss 0.00465047, throughput 9.22504K wps
[Epoch 25 Batch 90/162] avg loss 0.00483766, throughput 9.39814K wps
[Epoch 25 Batch 120/162] avg loss 0.00471379, throughput 9.48716K wps
[Epoch 25 Batch 150/162] avg loss 0.00490665, throughput 9.66107K wps
Begin Testing...
[Epoch 25] train avg loss 0.0048676, dev acc 0.8911, dev avg loss 0.282609, throughput 9.47451K wps
Observed Improvement.
Begin Testing...
[Epoch 26 Batch 30/162] avg loss 0.0051962, throughput 9.62728K wps
[Epoch 26 Batch 60/162] avg loss 0.00428586, throughput 9.2791K wps
[Epoch 26 Batch 90/162] avg loss 0.00498549, throughput 9.39475K wps
[Epoch 26 Batch 120/162] avg loss 0.00464043, throughput 9.59874K wps
[Epoch 26 Batch 150/162] avg loss 0.00494002, throughput 9.23763K wps
Begin Testing...
[Epoch 26] train avg loss 0.00478803, dev acc 0.8878, dev avg loss 0.275678, throughput 9.42442K wps
[Epoch 27 Batch 30/162] avg loss 0.00517158, throughput 9.5025K wps
[Epoch 27 Batch 60/162] avg loss 0.00476277, throughput 9.39688K wps
[Epoch 27 Batch 90/162] avg loss 0.00462116, throughput 9.29664K wps
[Epoch 27 Batch 120/162] avg loss 0.00485151, throughput 9.42347K wps
[Epoch 27 Batch 150/162] avg loss 0.00439272, throughput 9.27614K wps
Begin Testing...
[Epoch 27] train avg loss 0.0047656, dev acc 0.8856, dev avg loss 0.273667, throughput 9.37752K wps
[Epoch 28 Batch 30/162] avg loss 0.00469557, throughput 9.60443K wps
[Epoch 28 Batch 60/162] avg loss 0.00497813, throughput 9.23676K wps
[Epoch 28 Batch 90/162] avg loss 0.0047982, throughput 9.31416K wps
[Epoch 28 Batch 120/162] avg loss 0.00473131, throughput 9.35819K wps
[Epoch 28 Batch 150/162] avg loss 0.00444541, throughput 9.49076K wps
Begin Testing...
[Epoch 28] train avg loss 0.00468474, dev acc 0.8878, dev avg loss 0.272319, throughput 9.38611K wps
[Epoch 29 Batch 30/162] avg loss 0.00445117, throughput 9.61171K wps
[Epoch 29 Batch 60/162] avg loss 0.00442433, throughput 9.40294K wps
[Epoch 29 Batch 90/162] avg loss 0.00469758, throughput 9.47023K wps
[Epoch 29 Batch 120/162] avg loss 0.00458825, throughput 9.50263K wps
[Epoch 29 Batch 150/162] avg loss 0.00455407, throughput 9.26628K wps
Begin Testing...
[Epoch 29] train avg loss 0.00454631, dev acc 0.8889, dev avg loss 0.270602, throughput 9.45641K wps
[Epoch 30 Batch 30/162] avg loss 0.0043112, throughput 9.5011K wps
[Epoch 30 Batch 60/162] avg loss 0.00436722, throughput 9.39486K wps
[Epoch 30 Batch 90/162] avg loss 0.00481834, throughput 9.2512K wps
[Epoch 30 Batch 120/162] avg loss 0.00437984, throughput 9.44235K wps
[Epoch 30 Batch 150/162] avg loss 0.00472403, throughput 9.45661K wps
Begin Testing...
[Epoch 30] train avg loss 0.0044957, dev acc 0.8900, dev avg loss 0.268137, throughput 9.41925K wps
[Epoch 31 Batch 30/162] avg loss 0.00438802, throughput 9.73689K wps
[Epoch 31 Batch 60/162] avg loss 0.00425037, throughput 9.44483K wps
[Epoch 31 Batch 90/162] avg loss 0.00452522, throughput 9.40789K wps
[Epoch 31 Batch 120/162] avg loss 0.00458714, throughput 9.44848K wps
[Epoch 31 Batch 150/162] avg loss 0.00439272, throughput 9.4871K wps
Begin Testing...
[Epoch 31] train avg loss 0.00444202, dev acc 0.8878, dev avg loss 0.267325, throughput 9.48109K wps
[Epoch 32 Batch 30/162] avg loss 0.00435684, throughput 9.50634K wps
[Epoch 32 Batch 60/162] avg loss 0.00437578, throughput 9.45396K wps
[Epoch 32 Batch 90/162] avg loss 0.00436963, throughput 9.2586K wps
[Epoch 32 Batch 120/162] avg loss 0.00443165, throughput 9.43108K wps
[Epoch 32 Batch 150/162] avg loss 0.00425455, throughput 9.45919K wps
Begin Testing...
[Epoch 32] train avg loss 0.00434496, dev acc 0.8889, dev avg loss 0.265612, throughput 9.41242K wps
[Epoch 33 Batch 30/162] avg loss 0.00421435, throughput 9.62867K wps
[Epoch 33 Batch 60/162] avg loss 0.00414752, throughput 9.53543K wps
[Epoch 33 Batch 90/162] avg loss 0.00394169, throughput 9.42258K wps
[Epoch 33 Batch 120/162] avg loss 0.00454907, throughput 9.34645K wps
[Epoch 33 Batch 150/162] avg loss 0.0042142, throughput 9.32992K wps
Begin Testing...
[Epoch 33] train avg loss 0.00423132, dev acc 0.8956, dev avg loss 0.266501, throughput 9.44129K wps
Observed Improvement.
Begin Testing...
[Epoch 34 Batch 30/162] avg loss 0.00445449, throughput 9.48963K wps
[Epoch 34 Batch 60/162] avg loss 0.00429841, throughput 9.34195K wps
[Epoch 34 Batch 90/162] avg loss 0.00405799, throughput 9.27272K wps
[Epoch 34 Batch 120/162] avg loss 0.00412002, throughput 9.43628K wps
[Epoch 34 Batch 150/162] avg loss 0.00403201, throughput 9.30384K wps
Begin Testing...
[Epoch 34] train avg loss 0.00417717, dev acc 0.8944, dev avg loss 0.261844, throughput 9.38732K wps
[Epoch 35 Batch 30/162] avg loss 0.00420647, throughput 9.63541K wps
[Epoch 35 Batch 60/162] avg loss 0.004408, throughput 9.45227K wps
[Epoch 35 Batch 90/162] avg loss 0.00400692, throughput 9.40473K wps
[Epoch 35 Batch 120/162] avg loss 0.00417339, throughput 9.27338K wps
[Epoch 35 Batch 150/162] avg loss 0.00413619, throughput 9.35546K wps
Begin Testing...
[Epoch 35] train avg loss 0.00417271, dev acc 0.8944, dev avg loss 0.260326, throughput 9.41143K wps
[Epoch 36 Batch 30/162] avg loss 0.00395964, throughput 9.56937K wps
[Epoch 36 Batch 60/162] avg loss 0.00399158, throughput 9.35261K wps
[Epoch 36 Batch 90/162] avg loss 0.00420302, throughput 9.43645K wps
[Epoch 36 Batch 120/162] avg loss 0.00396143, throughput 9.24966K wps
[Epoch 36 Batch 150/162] avg loss 0.00425394, throughput 9.56149K wps
Begin Testing...
[Epoch 36] train avg loss 0.00404366, dev acc 0.8933, dev avg loss 0.25902, throughput 9.44254K wps
[Epoch 37 Batch 30/162] avg loss 0.00406676, throughput 9.5057K wps
[Epoch 37 Batch 60/162] avg loss 0.00391117, throughput 9.40041K wps
[Epoch 37 Batch 90/162] avg loss 0.00395022, throughput 9.57996K wps
[Epoch 37 Batch 120/162] avg loss 0.00407816, throughput 9.48388K wps
[Epoch 37 Batch 150/162] avg loss 0.00380932, throughput 9.32681K wps
Begin Testing...
[Epoch 37] train avg loss 0.003973, dev acc 0.8967, dev avg loss 0.261769, throughput 9.44663K wps
Observed Improvement.
Begin Testing...
[Epoch 38 Batch 30/162] avg loss 0.00389545, throughput 9.64064K wps
[Epoch 38 Batch 60/162] avg loss 0.0036025, throughput 9.51721K wps
[Epoch 38 Batch 90/162] avg loss 0.00366782, throughput 9.47687K wps
[Epoch 38 Batch 120/162] avg loss 0.00431908, throughput 9.19411K wps
[Epoch 38 Batch 150/162] avg loss 0.00397264, throughput 9.36635K wps
Begin Testing...
[Epoch 38] train avg loss 0.00386134, dev acc 0.8978, dev avg loss 0.256381, throughput 9.45109K wps
Observed Improvement.
Begin Testing...
[Epoch 39 Batch 30/162] avg loss 0.00391837, throughput 9.49239K wps
[Epoch 39 Batch 60/162] avg loss 0.00379664, throughput 9.28214K wps
[Epoch 39 Batch 90/162] avg loss 0.00383174, throughput 9.38839K wps
[Epoch 39 Batch 120/162] avg loss 0.00367619, throughput 9.30268K wps
[Epoch 39 Batch 150/162] avg loss 0.00398415, throughput 9.302K wps
Begin Testing...
[Epoch 39] train avg loss 0.00384752, dev acc 0.8944, dev avg loss 0.257191, throughput 9.3572K wps
[Epoch 40 Batch 30/162] avg loss 0.00372589, throughput 9.43286K wps
[Epoch 40 Batch 60/162] avg loss 0.00387268, throughput 9.334K wps
[Epoch 40 Batch 90/162] avg loss 0.00398563, throughput 9.44921K wps
[Epoch 40 Batch 120/162] avg loss 0.00371893, throughput 9.44634K wps
[Epoch 40 Batch 150/162] avg loss 0.00373852, throughput 9.48827K wps
Begin Testing...
[Epoch 40] train avg loss 0.00379574, dev acc 0.8967, dev avg loss 0.254443, throughput 9.44179K wps
[Epoch 41 Batch 30/162] avg loss 0.0037101, throughput 9.59218K wps
[Epoch 41 Batch 60/162] avg loss 0.00367523, throughput 9.51539K wps
[Epoch 41 Batch 90/162] avg loss 0.00349439, throughput 9.39221K wps
[Epoch 41 Batch 120/162] avg loss 0.00369703, throughput 9.34787K wps
[Epoch 41 Batch 150/162] avg loss 0.00375333, throughput 9.5061K wps
Begin Testing...
[Epoch 41] train avg loss 0.00373394, dev acc 0.8978, dev avg loss 0.254466, throughput 9.47083K wps
Observed Improvement.
Begin Testing...
[Epoch 42 Batch 30/162] avg loss 0.00359386, throughput 9.65703K wps
[Epoch 42 Batch 60/162] avg loss 0.00379183, throughput 9.41004K wps
[Epoch 42 Batch 90/162] avg loss 0.00404414, throughput 9.38425K wps
[Epoch 42 Batch 120/162] avg loss 0.00370032, throughput 9.63672K wps
[Epoch 42 Batch 150/162] avg loss 0.00374099, throughput 9.44825K wps
Begin Testing...
[Epoch 42] train avg loss 0.00375165, dev acc 0.8978, dev avg loss 0.252598, throughput 9.51299K wps
Observed Improvement.
Begin Testing...
[Epoch 43 Batch 30/162] avg loss 0.00374416, throughput 9.62092K wps
[Epoch 43 Batch 60/162] avg loss 0.00345905, throughput 9.53708K wps
[Epoch 43 Batch 90/162] avg loss 0.00406457, throughput 9.36483K wps
[Epoch 43 Batch 120/162] avg loss 0.00343715, throughput 9.33002K wps
[Epoch 43 Batch 150/162] avg loss 0.00338111, throughput 9.52288K wps
Begin Testing...
[Epoch 43] train avg loss 0.00357953, dev acc 0.8956, dev avg loss 0.251364, throughput 9.48252K wps
[Epoch 44 Batch 30/162] avg loss 0.00367875, throughput 9.63737K wps
[Epoch 44 Batch 60/162] avg loss 0.00366548, throughput 9.48517K wps
[Epoch 44 Batch 90/162] avg loss 0.00357258, throughput 9.41066K wps
[Epoch 44 Batch 120/162] avg loss 0.00353021, throughput 9.30851K wps
[Epoch 44 Batch 150/162] avg loss 0.00390197, throughput 9.54336K wps
Begin Testing...
[Epoch 44] train avg loss 0.00363544, dev acc 0.8967, dev avg loss 0.250476, throughput 9.46243K wps
[Epoch 45 Batch 30/162] avg loss 0.00364818, throughput 9.75346K wps
[Epoch 45 Batch 60/162] avg loss 0.00331761, throughput 9.58772K wps
[Epoch 45 Batch 90/162] avg loss 0.00334907, throughput 9.34858K wps
[Epoch 45 Batch 120/162] avg loss 0.00352415, throughput 9.67175K wps
[Epoch 45 Batch 150/162] avg loss 0.00360875, throughput 9.30611K wps
Begin Testing...
[Epoch 45] train avg loss 0.0034996, dev acc 0.8956, dev avg loss 0.249962, throughput 9.53333K wps
[Epoch 46 Batch 30/162] avg loss 0.00312446, throughput 9.56149K wps
[Epoch 46 Batch 60/162] avg loss 0.00349117, throughput 9.49442K wps
[Epoch 46 Batch 90/162] avg loss 0.00343554, throughput 9.4851K wps
[Epoch 46 Batch 120/162] avg loss 0.00364097, throughput 9.40872K wps
[Epoch 46 Batch 150/162] avg loss 0.0033347, throughput 9.49952K wps
Begin Testing...
[Epoch 46] train avg loss 0.00340239, dev acc 0.8989, dev avg loss 0.249271, throughput 9.46805K wps
Observed Improvement.
Begin Testing...
[Epoch 47 Batch 30/162] avg loss 0.00339269, throughput 9.69227K wps
[Epoch 47 Batch 60/162] avg loss 0.00357179, throughput 9.46511K wps
[Epoch 47 Batch 90/162] avg loss 0.00352139, throughput 9.29475K wps
[Epoch 47 Batch 120/162] avg loss 0.00313853, throughput 9.4109K wps
[Epoch 47 Batch 150/162] avg loss 0.00347206, throughput 9.25436K wps
Begin Testing...
[Epoch 47] train avg loss 0.00343104, dev acc 0.9011, dev avg loss 0.250189, throughput 9.42363K wps
Observed Improvement.
Begin Testing...
[Epoch 48 Batch 30/162] avg loss 0.00323179, throughput 9.60564K wps
[Epoch 48 Batch 60/162] avg loss 0.00352659, throughput 9.52944K wps
[Epoch 48 Batch 90/162] avg loss 0.0032965, throughput 9.26361K wps
[Epoch 48 Batch 120/162] avg loss 0.00327278, throughput 9.59944K wps
[Epoch 48 Batch 150/162] avg loss 0.00338295, throughput 9.46972K wps
Begin Testing...
[Epoch 48] train avg loss 0.00334496, dev acc 0.8989, dev avg loss 0.24789, throughput 9.47588K wps
[Epoch 49 Batch 30/162] avg loss 0.00326888, throughput 9.65504K wps
[Epoch 49 Batch 60/162] avg loss 0.00299243, throughput 9.24112K wps
[Epoch 49 Batch 90/162] avg loss 0.0033398, throughput 9.54241K wps
[Epoch 49 Batch 120/162] avg loss 0.00318589, throughput 9.49066K wps
[Epoch 49 Batch 150/162] avg loss 0.00360925, throughput 9.27282K wps
Begin Testing...
[Epoch 49] train avg loss 0.00328679, dev acc 0.9000, dev avg loss 0.246548, throughput 9.4436K wps
[Epoch 50 Batch 30/162] avg loss 0.00344233, throughput 9.58007K wps
[Epoch 50 Batch 60/162] avg loss 0.00337216, throughput 9.43527K wps
[Epoch 50 Batch 90/162] avg loss 0.0035342, throughput 9.48181K wps
[Epoch 50 Batch 120/162] avg loss 0.00342062, throughput 9.40291K wps
[Epoch 50 Batch 150/162] avg loss 0.00315227, throughput 9.5977K wps
Begin Testing...
[Epoch 50] train avg loss 0.00333785, dev acc 0.8978, dev avg loss 0.247292, throughput 9.50373K wps
[Epoch 51 Batch 30/162] avg loss 0.00306886, throughput 9.55899K wps
[Epoch 51 Batch 60/162] avg loss 0.00324111, throughput 9.48492K wps
[Epoch 51 Batch 90/162] avg loss 0.00347053, throughput 9.37515K wps
[Epoch 51 Batch 120/162] avg loss 0.0029426, throughput 9.36046K wps
[Epoch 51 Batch 150/162] avg loss 0.00350203, throughput 9.33628K wps
Begin Testing...
[Epoch 51] train avg loss 0.00324066, dev acc 0.8989, dev avg loss 0.246441, throughput 9.41466K wps
[Epoch 52 Batch 30/162] avg loss 0.00329651, throughput 9.48975K wps
[Epoch 52 Batch 60/162] avg loss 0.00308484, throughput 9.47111K wps
[Epoch 52 Batch 90/162] avg loss 0.00315949, throughput 9.30795K wps
[Epoch 52 Batch 120/162] avg loss 0.00327119, throughput 9.34289K wps
[Epoch 52 Batch 150/162] avg loss 0.00302739, throughput 9.52501K wps
Begin Testing...
[Epoch 52] train avg loss 0.00317249, dev acc 0.9000, dev avg loss 0.246482, throughput 9.43292K wps
[Epoch 53 Batch 30/162] avg loss 0.00347673, throughput 9.60939K wps
[Epoch 53 Batch 60/162] avg loss 0.00315038, throughput 9.44682K wps
[Epoch 53 Batch 90/162] avg loss 0.0034084, throughput 9.38566K wps
[Epoch 53 Batch 120/162] avg loss 0.00308647, throughput 9.31445K wps
[Epoch 53 Batch 150/162] avg loss 0.00300444, throughput 9.41105K wps
Begin Testing...
[Epoch 53] train avg loss 0.00318953, dev acc 0.9000, dev avg loss 0.244645, throughput 9.44013K wps
[Epoch 54 Batch 30/162] avg loss 0.0030505, throughput 9.8296K wps
[Epoch 54 Batch 60/162] avg loss 0.00349957, throughput 9.45981K wps
[Epoch 54 Batch 90/162] avg loss 0.00301865, throughput 9.39532K wps
[Epoch 54 Batch 120/162] avg loss 0.00326751, throughput 9.46381K wps
[Epoch 54 Batch 150/162] avg loss 0.0029384, throughput 9.41587K wps
Begin Testing...
[Epoch 54] train avg loss 0.00313741, dev acc 0.9011, dev avg loss 0.243344, throughput 9.49904K wps
Observed Improvement.
Begin Testing...
[Epoch 55 Batch 30/162] avg loss 0.00321449, throughput 9.59532K wps
[Epoch 55 Batch 60/162] avg loss 0.00297857, throughput 9.31674K wps
[Epoch 55 Batch 90/162] avg loss 0.00301541, throughput 9.44021K wps
[Epoch 55 Batch 120/162] avg loss 0.0031074, throughput 9.49669K wps
[Epoch 55 Batch 150/162] avg loss 0.00323747, throughput 9.39376K wps
Begin Testing...
[Epoch 55] train avg loss 0.00307783, dev acc 0.9022, dev avg loss 0.242915, throughput 9.45558K wps
Observed Improvement.
Begin Testing...
[Epoch 56 Batch 30/162] avg loss 0.00287106, throughput 9.60062K wps
[Epoch 56 Batch 60/162] avg loss 0.00312535, throughput 9.51523K wps
[Epoch 56 Batch 90/162] avg loss 0.00294015, throughput 9.22676K wps
[Epoch 56 Batch 120/162] avg loss 0.00319855, throughput 9.41634K wps
[Epoch 56 Batch 150/162] avg loss 0.00300817, throughput 9.39078K wps
Begin Testing...
[Epoch 56] train avg loss 0.00300322, dev acc 0.9011, dev avg loss 0.243321, throughput 9.43001K wps
[Epoch 57 Batch 30/162] avg loss 0.00306217, throughput 9.70008K wps
[Epoch 57 Batch 60/162] avg loss 0.00314334, throughput 9.4909K wps
[Epoch 57 Batch 90/162] avg loss 0.00271299, throughput 9.56445K wps
[Epoch 57 Batch 120/162] avg loss 0.00292442, throughput 9.45238K wps
[Epoch 57 Batch 150/162] avg loss 0.00289458, throughput 9.52011K wps
Begin Testing...
[Epoch 57] train avg loss 0.00294767, dev acc 0.9033, dev avg loss 0.242022, throughput 9.54036K wps
Observed Improvement.
Begin Testing...
[Epoch 58 Batch 30/162] avg loss 0.00295128, throughput 9.75491K wps
[Epoch 58 Batch 60/162] avg loss 0.00306336, throughput 9.57329K wps
[Epoch 58 Batch 90/162] avg loss 0.00279825, throughput 9.36341K wps
[Epoch 58 Batch 120/162] avg loss 0.0029306, throughput 9.41688K wps
[Epoch 58 Batch 150/162] avg loss 0.00281792, throughput 9.57169K wps
Begin Testing...
[Epoch 58] train avg loss 0.00291493, dev acc 0.9000, dev avg loss 0.245195, throughput 9.51285K wps
[Epoch 59 Batch 30/162] avg loss 0.00276599, throughput 9.6669K wps
[Epoch 59 Batch 60/162] avg loss 0.00296356, throughput 9.42621K wps
[Epoch 59 Batch 90/162] avg loss 0.00274941, throughput 9.43503K wps
[Epoch 59 Batch 120/162] avg loss 0.00289481, throughput 9.36225K wps
[Epoch 59 Batch 150/162] avg loss 0.00309269, throughput 9.49276K wps
Begin Testing...
[Epoch 59] train avg loss 0.00288158, dev acc 0.9033, dev avg loss 0.240755, throughput 9.45282K wps
Observed Improvement.
Begin Testing...
[Epoch 60 Batch 30/162] avg loss 0.00284223, throughput 9.42737K wps
[Epoch 60 Batch 60/162] avg loss 0.00267461, throughput 9.41869K wps
[Epoch 60 Batch 90/162] avg loss 0.00305755, throughput 9.56125K wps
[Epoch 60 Batch 120/162] avg loss 0.00273709, throughput 9.44577K wps
[Epoch 60 Batch 150/162] avg loss 0.0029957, throughput 9.38255K wps
Begin Testing...
[Epoch 60] train avg loss 0.00285618, dev acc 0.9033, dev avg loss 0.240525, throughput 9.4366K wps
Observed Improvement.
Begin Testing...
[Epoch 61 Batch 30/162] avg loss 0.00279332, throughput 9.73054K wps
[Epoch 61 Batch 60/162] avg loss 0.0031886, throughput 9.51809K wps
[Epoch 61 Batch 90/162] avg loss 0.00281929, throughput 9.51197K wps
[Epoch 61 Batch 120/162] avg loss 0.00275081, throughput 9.47342K wps
[Epoch 61 Batch 150/162] avg loss 0.00280328, throughput 9.34612K wps
Begin Testing...
[Epoch 61] train avg loss 0.00282389, dev acc 0.9022, dev avg loss 0.240432, throughput 9.50405K wps
[Epoch 62 Batch 30/162] avg loss 0.00282216, throughput 9.6559K wps
[Epoch 62 Batch 60/162] avg loss 0.0026385, throughput 9.51348K wps
[Epoch 62 Batch 90/162] avg loss 0.00276593, throughput 9.33238K wps
[Epoch 62 Batch 120/162] avg loss 0.00278765, throughput 9.58793K wps
[Epoch 62 Batch 150/162] avg loss 0.00284948, throughput 9.37325K wps
Begin Testing...
[Epoch 62] train avg loss 0.00274284, dev acc 0.9033, dev avg loss 0.238991, throughput 9.5033K wps
Observed Improvement.
Begin Testing...
[Epoch 63 Batch 30/162] avg loss 0.0026844, throughput 9.48358K wps
[Epoch 63 Batch 60/162] avg loss 0.00269869, throughput 9.33006K wps
[Epoch 63 Batch 90/162] avg loss 0.00275951, throughput 9.31025K wps
[Epoch 63 Batch 120/162] avg loss 0.0028551, throughput 9.354K wps
[Epoch 63 Batch 150/162] avg loss 0.00271709, throughput 9.33153K wps
Begin Testing...
[Epoch 63] train avg loss 0.00275588, dev acc 0.9033, dev avg loss 0.238678, throughput 9.35671K wps
Observed Improvement.
Begin Testing...
[Epoch 64 Batch 30/162] avg loss 0.00294694, throughput 9.75233K wps
[Epoch 64 Batch 60/162] avg loss 0.00270183, throughput 9.56395K wps
[Epoch 64 Batch 90/162] avg loss 0.00273751, throughput 9.50389K wps
[Epoch 64 Batch 120/162] avg loss 0.00264511, throughput 9.41426K wps
[Epoch 64 Batch 150/162] avg loss 0.00275478, throughput 9.4041K wps
Begin Testing...
[Epoch 64] train avg loss 0.00273998, dev acc 0.9044, dev avg loss 0.237897, throughput 9.5357K wps
Observed Improvement.
Begin Testing...
[Epoch 65 Batch 30/162] avg loss 0.00258117, throughput 9.61083K wps
[Epoch 65 Batch 60/162] avg loss 0.00268167, throughput 9.44652K wps
[Epoch 65 Batch 90/162] avg loss 0.00240897, throughput 9.39331K wps
[Epoch 65 Batch 120/162] avg loss 0.00270209, throughput 9.54052K wps
[Epoch 65 Batch 150/162] avg loss 0.00257115, throughput 9.46396K wps
Begin Testing...
[Epoch 65] train avg loss 0.00259717, dev acc 0.9022, dev avg loss 0.238362, throughput 9.45924K wps
[Epoch 66 Batch 30/162] avg loss 0.00263231, throughput 9.52731K wps
[Epoch 66 Batch 60/162] avg loss 0.00249942, throughput 9.30884K wps
[Epoch 66 Batch 90/162] avg loss 0.0027734, throughput 9.59533K wps
[Epoch 66 Batch 120/162] avg loss 0.00268559, throughput 9.45862K wps
[Epoch 66 Batch 150/162] avg loss 0.00257004, throughput 9.66081K wps
Begin Testing...
[Epoch 66] train avg loss 0.00261388, dev acc 0.9033, dev avg loss 0.23702, throughput 9.4881K wps
[Epoch 67 Batch 30/162] avg loss 0.00237505, throughput 9.63899K wps
[Epoch 67 Batch 60/162] avg loss 0.00249864, throughput 9.54036K wps
[Epoch 67 Batch 90/162] avg loss 0.00257215, throughput 9.40725K wps
[Epoch 67 Batch 120/162] avg loss 0.00268806, throughput 9.3513K wps
[Epoch 67 Batch 150/162] avg loss 0.00259846, throughput 9.42018K wps
Begin Testing...
[Epoch 67] train avg loss 0.00254609, dev acc 0.9022, dev avg loss 0.239451, throughput 9.46395K wps
[Epoch 68 Batch 30/162] avg loss 0.00249419, throughput 9.59641K wps
[Epoch 68 Batch 60/162] avg loss 0.0028501, throughput 9.57869K wps
[Epoch 68 Batch 90/162] avg loss 0.00247481, throughput 9.41839K wps
[Epoch 68 Batch 120/162] avg loss 0.00236332, throughput 9.41563K wps
[Epoch 68 Batch 150/162] avg loss 0.00242012, throughput 9.57505K wps
Begin Testing...
[Epoch 68] train avg loss 0.00253412, dev acc 0.9022, dev avg loss 0.236552, throughput 9.51405K wps
[Epoch 69 Batch 30/162] avg loss 0.00238946, throughput 9.75349K wps
[Epoch 69 Batch 60/162] avg loss 0.00264082, throughput 9.26767K wps
[Epoch 69 Batch 90/162] avg loss 0.00277535, throughput 9.53607K wps
[Epoch 69 Batch 120/162] avg loss 0.00221092, throughput 9.42635K wps
[Epoch 69 Batch 150/162] avg loss 0.00247828, throughput 9.37498K wps
Begin Testing...
[Epoch 69] train avg loss 0.00249504, dev acc 0.9056, dev avg loss 0.236737, throughput 9.47323K wps
Observed Improvement.
Begin Testing...
[Epoch 70 Batch 30/162] avg loss 0.00247951, throughput 9.60875K wps
[Epoch 70 Batch 60/162] avg loss 0.00248456, throughput 9.54031K wps
[Epoch 70 Batch 90/162] avg loss 0.00235864, throughput 9.57694K wps
[Epoch 70 Batch 120/162] avg loss 0.00244573, throughput 9.34954K wps
[Epoch 70 Batch 150/162] avg loss 0.00237443, throughput 9.41655K wps
Begin Testing...
[Epoch 70] train avg loss 0.00242211, dev acc 0.9033, dev avg loss 0.236613, throughput 9.50208K wps
[Epoch 71 Batch 30/162] avg loss 0.00226784, throughput 9.67299K wps
[Epoch 71 Batch 60/162] avg loss 0.00257066, throughput 9.56568K wps
[Epoch 71 Batch 90/162] avg loss 0.00234259, throughput 9.48808K wps
[Epoch 71 Batch 120/162] avg loss 0.00235208, throughput 9.5727K wps
[Epoch 71 Batch 150/162] avg loss 0.00263418, throughput 9.28417K wps
Begin Testing...
[Epoch 71] train avg loss 0.00241692, dev acc 0.9056, dev avg loss 0.238223, throughput 9.51111K wps
Observed Improvement.
Begin Testing...
[Epoch 72 Batch 30/162] avg loss 0.00238829, throughput 9.49263K wps
[Epoch 72 Batch 60/162] avg loss 0.00234113, throughput 9.34611K wps
[Epoch 72 Batch 90/162] avg loss 0.00241332, throughput 9.34415K wps
[Epoch 72 Batch 120/162] avg loss 0.00222798, throughput 9.35776K wps
[Epoch 72 Batch 150/162] avg loss 0.00248393, throughput 9.39048K wps
Begin Testing...
[Epoch 72] train avg loss 0.00234986, dev acc 0.9033, dev avg loss 0.237199, throughput 9.38379K wps
[Epoch 73 Batch 30/162] avg loss 0.00228942, throughput 9.71984K wps
[Epoch 73 Batch 60/162] avg loss 0.00233471, throughput 9.3723K wps
[Epoch 73 Batch 90/162] avg loss 0.00257221, throughput 9.4512K wps
[Epoch 73 Batch 120/162] avg loss 0.00230979, throughput 9.4457K wps
[Epoch 73 Batch 150/162] avg loss 0.00246272, throughput 9.27616K wps
Begin Testing...
[Epoch 73] train avg loss 0.00240399, dev acc 0.9033, dev avg loss 0.23896, throughput 9.44189K wps
[Epoch 74 Batch 30/162] avg loss 0.0022164, throughput 9.83079K wps
[Epoch 74 Batch 60/162] avg loss 0.00265901, throughput 9.53465K wps
[Epoch 74 Batch 90/162] avg loss 0.0024422, throughput 9.55773K wps
[Epoch 74 Batch 120/162] avg loss 0.00215701, throughput 9.48549K wps
[Epoch 74 Batch 150/162] avg loss 0.00227401, throughput 9.48824K wps
Begin Testing...
[Epoch 74] train avg loss 0.00233734, dev acc 0.9044, dev avg loss 0.235007, throughput 9.55812K wps
[Epoch 75 Batch 30/162] avg loss 0.00239308, throughput 9.55573K wps
[Epoch 75 Batch 60/162] avg loss 0.00215275, throughput 9.40644K wps
[Epoch 75 Batch 90/162] avg loss 0.002381, throughput 9.44793K wps
[Epoch 75 Batch 120/162] avg loss 0.00203749, throughput 9.57618K wps
[Epoch 75 Batch 150/162] avg loss 0.00265913, throughput 9.5684K wps
Begin Testing...
[Epoch 75] train avg loss 0.00231668, dev acc 0.9033, dev avg loss 0.236346, throughput 9.49573K wps
[Epoch 76 Batch 30/162] avg loss 0.00197726, throughput 9.54696K wps
[Epoch 76 Batch 60/162] avg loss 0.00239426, throughput 9.41229K wps
[Epoch 76 Batch 90/162] avg loss 0.00236232, throughput 9.50419K wps
[Epoch 76 Batch 120/162] avg loss 0.00218201, throughput 9.4093K wps
[Epoch 76 Batch 150/162] avg loss 0.00228679, throughput 9.45347K wps
Begin Testing...
[Epoch 76] train avg loss 0.00224537, dev acc 0.9044, dev avg loss 0.234583, throughput 9.46506K wps
[Epoch 77 Batch 30/162] avg loss 0.00233235, throughput 9.51829K wps
[Epoch 77 Batch 60/162] avg loss 0.00216376, throughput 9.47046K wps
[Epoch 77 Batch 90/162] avg loss 0.00230498, throughput 9.39725K wps
[Epoch 77 Batch 120/162] avg loss 0.00208656, throughput 9.28508K wps
[Epoch 77 Batch 150/162] avg loss 0.00220753, throughput 9.40677K wps
Begin Testing...
[Epoch 77] train avg loss 0.00222706, dev acc 0.9044, dev avg loss 0.234966, throughput 9.41241K wps
[Epoch 78 Batch 30/162] avg loss 0.00208551, throughput 9.56143K wps
[Epoch 78 Batch 60/162] avg loss 0.00228781, throughput 9.47817K wps
[Epoch 78 Batch 90/162] avg loss 0.00230539, throughput 9.27502K wps
[Epoch 78 Batch 120/162] avg loss 0.00220393, throughput 9.38894K wps
[Epoch 78 Batch 150/162] avg loss 0.00211793, throughput 9.41664K wps
Begin Testing...
[Epoch 78] train avg loss 0.00220011, dev acc 0.9044, dev avg loss 0.236272, throughput 9.43068K wps
[Epoch 79 Batch 30/162] avg loss 0.00207843, throughput 9.70164K wps
[Epoch 79 Batch 60/162] avg loss 0.00205188, throughput 9.46972K wps
[Epoch 79 Batch 90/162] avg loss 0.00244698, throughput 9.45774K wps
[Epoch 79 Batch 120/162] avg loss 0.0021048, throughput 9.45725K wps
[Epoch 79 Batch 150/162] avg loss 0.00213487, throughput 9.4479K wps
Begin Testing...
[Epoch 79] train avg loss 0.0021697, dev acc 0.9044, dev avg loss 0.233864, throughput 9.51349K wps
[Epoch 80 Batch 30/162] avg loss 0.00228443, throughput 9.59293K wps
[Epoch 80 Batch 60/162] avg loss 0.00189993, throughput 9.29173K wps
[Epoch 80 Batch 90/162] avg loss 0.00210619, throughput 9.59028K wps
[Epoch 80 Batch 120/162] avg loss 0.00214281, throughput 9.33783K wps
[Epoch 80 Batch 150/162] avg loss 0.00221075, throughput 9.61755K wps
Begin Testing...
[Epoch 80] train avg loss 0.00212459, dev acc 0.9056, dev avg loss 0.233096, throughput 9.4854K wps
Observed Improvement.
Begin Testing...
[Epoch 81 Batch 30/162] avg loss 0.00230807, throughput 9.75581K wps
[Epoch 81 Batch 60/162] avg loss 0.00186999, throughput 9.44939K wps
[Epoch 81 Batch 90/162] avg loss 0.00214232, throughput 9.27759K wps
[Epoch 81 Batch 120/162] avg loss 0.00208468, throughput 9.37237K wps
[Epoch 81 Batch 150/162] avg loss 0.0021095, throughput 9.49207K wps
Begin Testing...
[Epoch 81] train avg loss 0.00210686, dev acc 0.9067, dev avg loss 0.2347, throughput 9.47855K wps
Observed Improvement.
Begin Testing...
[Epoch 82 Batch 30/162] avg loss 0.0020782, throughput 9.76253K wps
[Epoch 82 Batch 60/162] avg loss 0.00214071, throughput 9.40875K wps
[Epoch 82 Batch 90/162] avg loss 0.00198098, throughput 9.49509K wps
[Epoch 82 Batch 120/162] avg loss 0.00222756, throughput 9.34594K wps
[Epoch 82 Batch 150/162] avg loss 0.00197796, throughput 9.35926K wps
Begin Testing...
[Epoch 82] train avg loss 0.00210545, dev acc 0.9089, dev avg loss 0.234337, throughput 9.48323K wps
Observed Improvement.
Begin Testing...
[Epoch 83 Batch 30/162] avg loss 0.00218615, throughput 9.59741K wps
[Epoch 83 Batch 60/162] avg loss 0.00191836, throughput 9.27804K wps
[Epoch 83 Batch 90/162] avg loss 0.00186948, throughput 9.28552K wps
[Epoch 83 Batch 120/162] avg loss 0.00220125, throughput 9.52811K wps
[Epoch 83 Batch 150/162] avg loss 0.00209541, throughput 9.40279K wps
Begin Testing...
[Epoch 83] train avg loss 0.00205084, dev acc 0.9089, dev avg loss 0.233926, throughput 9.41617K wps
Observed Improvement.
Begin Testing...
[Epoch 84 Batch 30/162] avg loss 0.00205982, throughput 9.60268K wps
[Epoch 84 Batch 60/162] avg loss 0.00204798, throughput 9.2878K wps
[Epoch 84 Batch 90/162] avg loss 0.00208939, throughput 9.52452K wps
[Epoch 84 Batch 120/162] avg loss 0.00185644, throughput 9.34449K wps
[Epoch 84 Batch 150/162] avg loss 0.00199965, throughput 9.45096K wps
Begin Testing...
[Epoch 84] train avg loss 0.00201818, dev acc 0.9044, dev avg loss 0.233123, throughput 9.43699K wps
[Epoch 85 Batch 30/162] avg loss 0.00206186, throughput 9.63452K wps
[Epoch 85 Batch 60/162] avg loss 0.00209902, throughput 9.32211K wps
[Epoch 85 Batch 90/162] avg loss 0.00203478, throughput 9.45279K wps
[Epoch 85 Batch 120/162] avg loss 0.00196974, throughput 9.31092K wps
[Epoch 85 Batch 150/162] avg loss 0.00224166, throughput 9.35085K wps
Begin Testing...
[Epoch 85] train avg loss 0.0020691, dev acc 0.9033, dev avg loss 0.23324, throughput 9.41544K wps
[Epoch 86 Batch 30/162] avg loss 0.00190311, throughput 9.7988K wps
[Epoch 86 Batch 60/162] avg loss 0.00194061, throughput 9.32938K wps
[Epoch 86 Batch 90/162] avg loss 0.00174652, throughput 9.53536K wps
[Epoch 86 Batch 120/162] avg loss 0.0016922, throughput 9.31871K wps
[Epoch 86 Batch 150/162] avg loss 0.00225328, throughput 9.55641K wps
Begin Testing...
[Epoch 86] train avg loss 0.00193955, dev acc 0.9056, dev avg loss 0.23375, throughput 9.49839K wps
[Epoch 87 Batch 30/162] avg loss 0.00195882, throughput 9.42401K wps
[Epoch 87 Batch 60/162] avg loss 0.00200846, throughput 9.39364K wps
[Epoch 87 Batch 90/162] avg loss 0.0017587, throughput 9.42066K wps
[Epoch 87 Batch 120/162] avg loss 0.00191597, throughput 9.45118K wps
[Epoch 87 Batch 150/162] avg loss 0.00207078, throughput 9.44183K wps
Begin Testing...
[Epoch 87] train avg loss 0.00192623, dev acc 0.9056, dev avg loss 0.232297, throughput 9.43566K wps
[Epoch 88 Batch 30/162] avg loss 0.00174324, throughput 9.63214K wps
[Epoch 88 Batch 60/162] avg loss 0.00171938, throughput 9.33051K wps
[Epoch 88 Batch 90/162] avg loss 0.00223408, throughput 9.46136K wps
[Epoch 88 Batch 120/162] avg loss 0.00185931, throughput 9.4871K wps
[Epoch 88 Batch 150/162] avg loss 0.0020512, throughput 9.29013K wps
Begin Testing...
[Epoch 88] train avg loss 0.0019084, dev acc 0.9067, dev avg loss 0.232617, throughput 9.42424K wps
[Epoch 89 Batch 30/162] avg loss 0.00189569, throughput 9.55458K wps
[Epoch 89 Batch 60/162] avg loss 0.00185958, throughput 9.41293K wps
[Epoch 89 Batch 90/162] avg loss 0.00198439, throughput 9.55019K wps
[Epoch 89 Batch 120/162] avg loss 0.00203424, throughput 9.60004K wps
[Epoch 89 Batch 150/162] avg loss 0.00172255, throughput 9.3776K wps
Begin Testing...
[Epoch 89] train avg loss 0.00189921, dev acc 0.9078, dev avg loss 0.233586, throughput 9.48071K wps
[Epoch 90 Batch 30/162] avg loss 0.0019206, throughput 9.63871K wps
[Epoch 90 Batch 60/162] avg loss 0.00186736, throughput 9.49226K wps
[Epoch 90 Batch 90/162] avg loss 0.00188893, throughput 9.43891K wps
[Epoch 90 Batch 120/162] avg loss 0.00202942, throughput 9.34139K wps
[Epoch 90 Batch 150/162] avg loss 0.0017351, throughput 9.48383K wps
Begin Testing...
[Epoch 90] train avg loss 0.00189004, dev acc 0.9056, dev avg loss 0.232773, throughput 9.47545K wps
[Epoch 91 Batch 30/162] avg loss 0.00175208, throughput 9.71315K wps
[Epoch 91 Batch 60/162] avg loss 0.00192566, throughput 9.32663K wps
[Epoch 91 Batch 90/162] avg loss 0.00187736, throughput 9.33437K wps
[Epoch 91 Batch 120/162] avg loss 0.00183564, throughput 9.39247K wps
[Epoch 91 Batch 150/162] avg loss 0.00194891, throughput 9.39723K wps
Begin Testing...
[Epoch 91] train avg loss 0.00185106, dev acc 0.9078, dev avg loss 0.231651, throughput 9.41548K wps
[Epoch 92 Batch 30/162] avg loss 0.0016962, throughput 9.49928K wps
[Epoch 92 Batch 60/162] avg loss 0.00200907, throughput 9.24739K wps
[Epoch 92 Batch 90/162] avg loss 0.00169897, throughput 9.35697K wps
[Epoch 92 Batch 120/162] avg loss 0.00192679, throughput 9.34989K wps
[Epoch 92 Batch 150/162] avg loss 0.00193032, throughput 9.33131K wps
Begin Testing...
[Epoch 92] train avg loss 0.00183702, dev acc 0.9067, dev avg loss 0.231005, throughput 9.36476K wps
[Epoch 93 Batch 30/162] avg loss 0.00198151, throughput 9.72899K wps
[Epoch 93 Batch 60/162] avg loss 0.00189333, throughput 9.29571K wps
[Epoch 93 Batch 90/162] avg loss 0.00191072, throughput 9.36736K wps
[Epoch 93 Batch 120/162] avg loss 0.00193699, throughput 9.48983K wps
[Epoch 93 Batch 150/162] avg loss 0.00177089, throughput 9.54946K wps
Begin Testing...
[Epoch 93] train avg loss 0.00188866, dev acc 0.9089, dev avg loss 0.231763, throughput 9.48971K wps
Observed Improvement.
Begin Testing...
[Epoch 94 Batch 30/162] avg loss 0.00167929, throughput 9.59623K wps
[Epoch 94 Batch 60/162] avg loss 0.00176237, throughput 9.26253K wps
[Epoch 94 Batch 90/162] avg loss 0.00176733, throughput 9.48907K wps
[Epoch 94 Batch 120/162] avg loss 0.00201417, throughput 9.22864K wps
[Epoch 94 Batch 150/162] avg loss 0.00164876, throughput 9.46725K wps
Begin Testing...
[Epoch 94] train avg loss 0.00176616, dev acc 0.9067, dev avg loss 0.231269, throughput 9.40727K wps
[Epoch 95 Batch 30/162] avg loss 0.00187361, throughput 9.61019K wps
[Epoch 95 Batch 60/162] avg loss 0.00170376, throughput 9.21268K wps
[Epoch 95 Batch 90/162] avg loss 0.0019063, throughput 9.52444K wps
[Epoch 95 Batch 120/162] avg loss 0.00161044, throughput 9.52817K wps
[Epoch 95 Batch 150/162] avg loss 0.00168726, throughput 9.53738K wps
Begin Testing...
[Epoch 95] train avg loss 0.00177911, dev acc 0.9100, dev avg loss 0.231809, throughput 9.48744K wps
Observed Improvement.
Begin Testing...
[Epoch 96 Batch 30/162] avg loss 0.00164973, throughput 9.64667K wps
[Epoch 96 Batch 60/162] avg loss 0.0018261, throughput 9.60098K wps
[Epoch 96 Batch 90/162] avg loss 0.00185781, throughput 9.34885K wps
[Epoch 96 Batch 120/162] avg loss 0.0016571, throughput 9.5405K wps
[Epoch 96 Batch 150/162] avg loss 0.00172072, throughput 9.32558K wps
Begin Testing...
[Epoch 96] train avg loss 0.00174566, dev acc 0.9078, dev avg loss 0.231038, throughput 9.49135K wps
[Epoch 97 Batch 30/162] avg loss 0.00172349, throughput 9.6806K wps
[Epoch 97 Batch 60/162] avg loss 0.0019015, throughput 9.31626K wps
[Epoch 97 Batch 90/162] avg loss 0.00173132, throughput 9.64392K wps
[Epoch 97 Batch 120/162] avg loss 0.00177573, throughput 9.49568K wps
[Epoch 97 Batch 150/162] avg loss 0.00161562, throughput 9.43681K wps
Begin Testing...
[Epoch 97] train avg loss 0.00174106, dev acc 0.9067, dev avg loss 0.23071, throughput 9.51176K wps
[Epoch 98 Batch 30/162] avg loss 0.00175057, throughput 9.67126K wps
[Epoch 98 Batch 60/162] avg loss 0.00174134, throughput 9.55369K wps
[Epoch 98 Batch 90/162] avg loss 0.00159539, throughput 9.3399K wps
[Epoch 98 Batch 120/162] avg loss 0.00155096, throughput 9.33833K wps
[Epoch 98 Batch 150/162] avg loss 0.001698, throughput 9.40679K wps
Begin Testing...
[Epoch 98] train avg loss 0.00167764, dev acc 0.9089, dev avg loss 0.232567, throughput 9.45498K wps
[Epoch 99 Batch 30/162] avg loss 0.00173709, throughput 9.76721K wps
[Epoch 99 Batch 60/162] avg loss 0.00165767, throughput 9.34398K wps
[Epoch 99 Batch 90/162] avg loss 0.00164185, throughput 9.29646K wps
[Epoch 99 Batch 120/162] avg loss 0.00159075, throughput 9.36776K wps
[Epoch 99 Batch 150/162] avg loss 0.00167799, throughput 9.48964K wps
Begin Testing...
[Epoch 99] train avg loss 0.00166708, dev acc 0.9100, dev avg loss 0.230981, throughput 9.4546K wps
Observed Improvement.
Begin Testing...
[Epoch 100 Batch 30/162] avg loss 0.00177946, throughput 9.88289K wps
[Epoch 100 Batch 60/162] avg loss 0.00156994, throughput 9.55573K wps
[Epoch 100 Batch 90/162] avg loss 0.0017641, throughput 9.40304K wps
[Epoch 100 Batch 120/162] avg loss 0.00154562, throughput 9.62457K wps
[Epoch 100 Batch 150/162] avg loss 0.00156198, throughput 9.43729K wps
Begin Testing...
[Epoch 100] train avg loss 0.00165192, dev acc 0.9089, dev avg loss 0.231392, throughput 9.56662K wps
[Epoch 101 Batch 30/162] avg loss 0.0017023, throughput 9.61379K wps
[Epoch 101 Batch 60/162] avg loss 0.00155896, throughput 9.44196K wps
[Epoch 101 Batch 90/162] avg loss 0.00161112, throughput 9.27076K wps
[Epoch 101 Batch 120/162] avg loss 0.00157599, throughput 9.5093K wps
[Epoch 101 Batch 150/162] avg loss 0.00168535, throughput 9.50622K wps
Begin Testing...
[Epoch 101] train avg loss 0.00163828, dev acc 0.9100, dev avg loss 0.231236, throughput 9.45619K wps
Observed Improvement.
Begin Testing...
[Epoch 102 Batch 30/162] avg loss 0.00158592, throughput 9.68284K wps
[Epoch 102 Batch 60/162] avg loss 0.00169051, throughput 9.34457K wps
[Epoch 102 Batch 90/162] avg loss 0.00162776, throughput 9.35417K wps
[Epoch 102 Batch 120/162] avg loss 0.00161411, throughput 9.5591K wps
[Epoch 102 Batch 150/162] avg loss 0.0015784, throughput 9.2364K wps
Begin Testing...
[Epoch 102] train avg loss 0.00161341, dev acc 0.9078, dev avg loss 0.231198, throughput 9.44406K wps
[Epoch 103 Batch 30/162] avg loss 0.00161369, throughput 9.44827K wps
[Epoch 103 Batch 60/162] avg loss 0.00165274, throughput 9.3009K wps
[Epoch 103 Batch 90/162] avg loss 0.00148703, throughput 9.3854K wps
[Epoch 103 Batch 120/162] avg loss 0.00158074, throughput 9.48193K wps
[Epoch 103 Batch 150/162] avg loss 0.00174653, throughput 9.23574K wps
Begin Testing...
[Epoch 103] train avg loss 0.00162123, dev acc 0.9089, dev avg loss 0.232362, throughput 9.38859K wps
[Epoch 104 Batch 30/162] avg loss 0.00157465, throughput 9.66848K wps
[Epoch 104 Batch 60/162] avg loss 0.00168573, throughput 9.45382K wps
[Epoch 104 Batch 90/162] avg loss 0.0014762, throughput 9.3264K wps
[Epoch 104 Batch 120/162] avg loss 0.00162264, throughput 9.36418K wps
[Epoch 104 Batch 150/162] avg loss 0.00143148, throughput 9.35938K wps
Begin Testing...
[Epoch 104] train avg loss 0.00157206, dev acc 0.9089, dev avg loss 0.230843, throughput 9.4422K wps
[Epoch 105 Batch 30/162] avg loss 0.00138914, throughput 9.70474K wps
[Epoch 105 Batch 60/162] avg loss 0.00157334, throughput 9.35212K wps
[Epoch 105 Batch 90/162] avg loss 0.00164756, throughput 9.34464K wps
[Epoch 105 Batch 120/162] avg loss 0.00157804, throughput 9.53363K wps
[Epoch 105 Batch 150/162] avg loss 0.00156536, throughput 9.30213K wps
Begin Testing...
[Epoch 105] train avg loss 0.00156017, dev acc 0.9078, dev avg loss 0.230947, throughput 9.44452K wps
[Epoch 106 Batch 30/162] avg loss 0.00167607, throughput 9.50789K wps
[Epoch 106 Batch 60/162] avg loss 0.00149497, throughput 9.23811K wps
[Epoch 106 Batch 90/162] avg loss 0.0015804, throughput 9.38465K wps
[Epoch 106 Batch 120/162] avg loss 0.00142475, throughput 9.46722K wps
[Epoch 106 Batch 150/162] avg loss 0.00137981, throughput 9.24434K wps
Begin Testing...
[Epoch 106] train avg loss 0.00150915, dev acc 0.9067, dev avg loss 0.231185, throughput 9.36391K wps
[Epoch 107 Batch 30/162] avg loss 0.00154697, throughput 9.48331K wps
[Epoch 107 Batch 60/162] avg loss 0.00140119, throughput 9.48517K wps
[Epoch 107 Batch 90/162] avg loss 0.00153385, throughput 9.30068K wps
[Epoch 107 Batch 120/162] avg loss 0.00131722, throughput 9.57513K wps
[Epoch 107 Batch 150/162] avg loss 0.00153148, throughput 9.60153K wps
Begin Testing...
[Epoch 107] train avg loss 0.00147933, dev acc 0.9089, dev avg loss 0.23296, throughput 9.49533K wps
[Epoch 108 Batch 30/162] avg loss 0.00152636, throughput 9.5273K wps
[Epoch 108 Batch 60/162] avg loss 0.00133818, throughput 9.44082K wps
[Epoch 108 Batch 90/162] avg loss 0.00144891, throughput 9.57743K wps
[Epoch 108 Batch 120/162] avg loss 0.00154193, throughput 9.47475K wps
[Epoch 108 Batch 150/162] avg loss 0.00137311, throughput 9.5091K wps
Begin Testing...
[Epoch 108] train avg loss 0.00146118, dev acc 0.9100, dev avg loss 0.231511, throughput 9.49119K wps
Observed Improvement.
Begin Testing...
[Epoch 109 Batch 30/162] avg loss 0.00145889, throughput 9.53668K wps
[Epoch 109 Batch 60/162] avg loss 0.0014654, throughput 9.42453K wps
[Epoch 109 Batch 90/162] avg loss 0.00148896, throughput 9.42709K wps
[Epoch 109 Batch 120/162] avg loss 0.00136276, throughput 9.547K wps
[Epoch 109 Batch 150/162] avg loss 0.00145581, throughput 9.27671K wps
Begin Testing...
[Epoch 109] train avg loss 0.00145121, dev acc 0.9100, dev avg loss 0.230901, throughput 9.438K wps
Observed Improvement.
Begin Testing...
[Epoch 110 Batch 30/162] avg loss 0.00124118, throughput 9.67691K wps
[Epoch 110 Batch 60/162] avg loss 0.00145392, throughput 9.37659K wps
[Epoch 110 Batch 90/162] avg loss 0.00139357, throughput 9.36294K wps
[Epoch 110 Batch 120/162] avg loss 0.00156652, throughput 9.36546K wps
[Epoch 110 Batch 150/162] avg loss 0.00140168, throughput 9.32295K wps
Begin Testing...
[Epoch 110] train avg loss 0.00141125, dev acc 0.9100, dev avg loss 0.231061, throughput 9.41061K wps
Observed Improvement.
Begin Testing...
[Epoch 111 Batch 30/162] avg loss 0.0014223, throughput 9.7503K wps
[Epoch 111 Batch 60/162] avg loss 0.00142897, throughput 9.45155K wps
[Epoch 111 Batch 90/162] avg loss 0.00140558, throughput 9.29849K wps
[Epoch 111 Batch 120/162] avg loss 0.00143581, throughput 9.40522K wps
[Epoch 111 Batch 150/162] avg loss 0.00136527, throughput 9.43094K wps
Begin Testing...
[Epoch 111] train avg loss 0.00142264, dev acc 0.9078, dev avg loss 0.233955, throughput 9.47329K wps
[Epoch 112 Batch 30/162] avg loss 0.00142264, throughput 9.65195K wps
[Epoch 112 Batch 60/162] avg loss 0.00138868, throughput 9.54812K wps
[Epoch 112 Batch 90/162] avg loss 0.00143521, throughput 9.60013K wps
[Epoch 112 Batch 120/162] avg loss 0.00136962, throughput 9.32394K wps
[Epoch 112 Batch 150/162] avg loss 0.00144492, throughput 9.38876K wps
Begin Testing...
[Epoch 112] train avg loss 0.00141386, dev acc 0.9100, dev avg loss 0.231494, throughput 9.51177K wps
Observed Improvement.
Begin Testing...
[Epoch 113 Batch 30/162] avg loss 0.00143093, throughput 9.50098K wps
[Epoch 113 Batch 60/162] avg loss 0.0015202, throughput 9.2477K wps
[Epoch 113 Batch 90/162] avg loss 0.00131997, throughput 9.39283K wps
[Epoch 113 Batch 120/162] avg loss 0.00150868, throughput 9.43789K wps
[Epoch 113 Batch 150/162] avg loss 0.00129257, throughput 9.27451K wps
Begin Testing...
[Epoch 113] train avg loss 0.00141651, dev acc 0.9089, dev avg loss 0.231303, throughput 9.37897K wps
[Epoch 114 Batch 30/162] avg loss 0.00128209, throughput 9.76906K wps
[Epoch 114 Batch 60/162] avg loss 0.00131826, throughput 9.35038K wps
[Epoch 114 Batch 90/162] avg loss 0.00137351, throughput 9.57132K wps
[Epoch 114 Batch 120/162] avg loss 0.0014918, throughput 9.49405K wps
[Epoch 114 Batch 150/162] avg loss 0.00132746, throughput 9.50537K wps
Begin Testing...
[Epoch 114] train avg loss 0.00136385, dev acc 0.9111, dev avg loss 0.231004, throughput 9.5073K wps
Observed Improvement.
Begin Testing...
[Epoch 115 Batch 30/162] avg loss 0.00142524, throughput 9.51886K wps
[Epoch 115 Batch 60/162] avg loss 0.00117217, throughput 9.41475K wps
[Epoch 115 Batch 90/162] avg loss 0.00134475, throughput 9.41614K wps
[Epoch 115 Batch 120/162] avg loss 0.00148198, throughput 9.47668K wps
[Epoch 115 Batch 150/162] avg loss 0.00130951, throughput 9.27547K wps
Begin Testing...
[Epoch 115] train avg loss 0.00135318, dev acc 0.9100, dev avg loss 0.232282, throughput 9.42365K wps
[Epoch 116 Batch 30/162] avg loss 0.00129882, throughput 9.58521K wps
[Epoch 116 Batch 60/162] avg loss 0.0013387, throughput 9.56154K wps
[Epoch 116 Batch 90/162] avg loss 0.00124751, throughput 9.48153K wps
[Epoch 116 Batch 120/162] avg loss 0.00124682, throughput 9.41098K wps
[Epoch 116 Batch 150/162] avg loss 0.00132552, throughput 9.27498K wps
Begin Testing...
[Epoch 116] train avg loss 0.00130091, dev acc 0.9089, dev avg loss 0.233897, throughput 9.44995K wps
[Epoch 117 Batch 30/162] avg loss 0.00135398, throughput 9.65705K wps
[Epoch 117 Batch 60/162] avg loss 0.00136195, throughput 9.20803K wps
[Epoch 117 Batch 90/162] avg loss 0.00135955, throughput 9.59577K wps
[Epoch 117 Batch 120/162] avg loss 0.00126117, throughput 9.49448K wps
[Epoch 117 Batch 150/162] avg loss 0.00129931, throughput 9.39629K wps
Begin Testing...
[Epoch 117] train avg loss 0.00132423, dev acc 0.9111, dev avg loss 0.231874, throughput 9.46994K wps
Observed Improvement.
Begin Testing...
[Epoch 118 Batch 30/162] avg loss 0.0012014, throughput 9.79438K wps
[Epoch 118 Batch 60/162] avg loss 0.0013299, throughput 9.41035K wps
[Epoch 118 Batch 90/162] avg loss 0.00131252, throughput 9.50088K wps
[Epoch 118 Batch 120/162] avg loss 0.00148068, throughput 9.33535K wps
[Epoch 118 Batch 150/162] avg loss 0.00123757, throughput 9.35919K wps
Begin Testing...
[Epoch 118] train avg loss 0.00133295, dev acc 0.9100, dev avg loss 0.233623, throughput 9.48917K wps
[Epoch 119 Batch 30/162] avg loss 0.00139257, throughput 8.64105K wps
[Epoch 119 Batch 60/162] avg loss 0.00127229, throughput 9.63241K wps
[Epoch 119 Batch 90/162] avg loss 0.00127384, throughput 9.42764K wps
[Epoch 119 Batch 120/162] avg loss 0.0013477, throughput 9.42333K wps
[Epoch 119 Batch 150/162] avg loss 0.00130624, throughput 9.3275K wps
Begin Testing...
[Epoch 119] train avg loss 0.00131437, dev acc 0.9089, dev avg loss 0.231863, throughput 9.28311K wps
[Epoch 120 Batch 30/162] avg loss 0.00119028, throughput 9.52502K wps
[Epoch 120 Batch 60/162] avg loss 0.00119499, throughput 9.49375K wps
[Epoch 120 Batch 90/162] avg loss 0.00138944, throughput 9.56341K wps
[Epoch 120 Batch 120/162] avg loss 0.00127802, throughput 9.49594K wps
[Epoch 120 Batch 150/162] avg loss 0.00142661, throughput 9.57499K wps
Begin Testing...
[Epoch 120] train avg loss 0.00128959, dev acc 0.9100, dev avg loss 0.231116, throughput 9.51842K wps
[Epoch 121 Batch 30/162] avg loss 0.00123836, throughput 9.79153K wps
[Epoch 121 Batch 60/162] avg loss 0.0013712, throughput 9.31197K wps
[Epoch 121 Batch 90/162] avg loss 0.00130673, throughput 9.459K wps
[Epoch 121 Batch 120/162] avg loss 0.00117524, throughput 9.32896K wps
[Epoch 121 Batch 150/162] avg loss 0.00136416, throughput 9.47054K wps
Begin Testing...
[Epoch 121] train avg loss 0.00129137, dev acc 0.9089, dev avg loss 0.231597, throughput 9.47461K wps
[Epoch 122 Batch 30/162] avg loss 0.00125466, throughput 9.5769K wps
[Epoch 122 Batch 60/162] avg loss 0.00130391, throughput 9.49063K wps
[Epoch 122 Batch 90/162] avg loss 0.00123786, throughput 9.52744K wps
[Epoch 122 Batch 120/162] avg loss 0.00113567, throughput 9.36725K wps
[Epoch 122 Batch 150/162] avg loss 0.00127073, throughput 9.44143K wps
Begin Testing...
[Epoch 122] train avg loss 0.00124037, dev acc 0.9089, dev avg loss 0.231677, throughput 9.47443K wps
[Epoch 123 Batch 30/162] avg loss 0.00124063, throughput 9.55169K wps
[Epoch 123 Batch 60/162] avg loss 0.00119596, throughput 9.51159K wps
[Epoch 123 Batch 90/162] avg loss 0.00113664, throughput 9.5278K wps
[Epoch 123 Batch 120/162] avg loss 0.00113098, throughput 9.33944K wps
[Epoch 123 Batch 150/162] avg loss 0.00123898, throughput 9.45991K wps
Begin Testing...
[Epoch 123] train avg loss 0.001194, dev acc 0.9100, dev avg loss 0.23152, throughput 9.46472K wps
[Epoch 124 Batch 30/162] avg loss 0.00120816, throughput 9.66889K wps
[Epoch 124 Batch 60/162] avg loss 0.00119266, throughput 9.55849K wps
[Epoch 124 Batch 90/162] avg loss 0.00128684, throughput 9.52755K wps
[Epoch 124 Batch 120/162] avg loss 0.00123652, throughput 9.41114K wps
[Epoch 124 Batch 150/162] avg loss 0.00114845, throughput 9.3479K wps
Begin Testing...
[Epoch 124] train avg loss 0.00120558, dev acc 0.9100, dev avg loss 0.232895, throughput 9.50534K wps
[Epoch 125 Batch 30/162] avg loss 0.00129345, throughput 9.51511K wps
[Epoch 125 Batch 60/162] avg loss 0.0012511, throughput 9.4542K wps
[Epoch 125 Batch 90/162] avg loss 0.00121979, throughput 9.44023K wps
[Epoch 125 Batch 120/162] avg loss 0.00109882, throughput 9.37668K wps
[Epoch 125 Batch 150/162] avg loss 0.00113827, throughput 9.55714K wps
Begin Testing...
[Epoch 125] train avg loss 0.00119768, dev acc 0.9111, dev avg loss 0.231234, throughput 9.46811K wps
Observed Improvement.
Begin Testing...
[Epoch 126 Batch 30/162] avg loss 0.00131378, throughput 9.72791K wps
[Epoch 126 Batch 60/162] avg loss 0.00119958, throughput 9.22228K wps
[Epoch 126 Batch 90/162] avg loss 0.00107723, throughput 9.31649K wps
[Epoch 126 Batch 120/162] avg loss 0.00117262, throughput 9.45298K wps
[Epoch 126 Batch 150/162] avg loss 0.00113459, throughput 9.44983K wps
Begin Testing...
[Epoch 126] train avg loss 0.0011848, dev acc 0.9111, dev avg loss 0.231473, throughput 9.43481K wps
Observed Improvement.
Begin Testing...
[Epoch 127 Batch 30/162] avg loss 0.00100859, throughput 9.74467K wps
[Epoch 127 Batch 60/162] avg loss 0.00117626, throughput 9.57037K wps
[Epoch 127 Batch 90/162] avg loss 0.00121317, throughput 9.60485K wps
[Epoch 127 Batch 120/162] avg loss 0.00120067, throughput 9.44181K wps
[Epoch 127 Batch 150/162] avg loss 0.00110055, throughput 9.37651K wps
Begin Testing...
[Epoch 127] train avg loss 0.00113945, dev acc 0.9111, dev avg loss 0.233417, throughput 9.51204K wps
Observed Improvement.
Begin Testing...
[Epoch 128 Batch 30/162] avg loss 0.00117919, throughput 9.56179K wps
[Epoch 128 Batch 60/162] avg loss 0.0012265, throughput 9.40013K wps
[Epoch 128 Batch 90/162] avg loss 0.00114917, throughput 9.53671K wps
[Epoch 128 Batch 120/162] avg loss 0.0011057, throughput 9.413K wps
[Epoch 128 Batch 150/162] avg loss 0.00106979, throughput 9.42141K wps
Begin Testing...
[Epoch 128] train avg loss 0.0011455, dev acc 0.9100, dev avg loss 0.232148, throughput 9.45762K wps
[Epoch 129 Batch 30/162] avg loss 0.00109544, throughput 9.58115K wps
[Epoch 129 Batch 60/162] avg loss 0.00117612, throughput 9.46285K wps
[Epoch 129 Batch 90/162] avg loss 0.00107821, throughput 9.50812K wps
[Epoch 129 Batch 120/162] avg loss 0.00107582, throughput 9.30331K wps
[Epoch 129 Batch 150/162] avg loss 0.00119815, throughput 9.54827K wps
Begin Testing...
[Epoch 129] train avg loss 0.00112006, dev acc 0.9089, dev avg loss 0.232068, throughput 9.46008K wps
[Epoch 130 Batch 30/162] avg loss 0.00112453, throughput 9.72985K wps
[Epoch 130 Batch 60/162] avg loss 0.00112876, throughput 9.4569K wps
[Epoch 130 Batch 90/162] avg loss 0.00113899, throughput 9.60967K wps
[Epoch 130 Batch 120/162] avg loss 0.00118462, throughput 9.58598K wps
[Epoch 130 Batch 150/162] avg loss 0.00108666, throughput 9.45877K wps
Begin Testing...
[Epoch 130] train avg loss 0.00111997, dev acc 0.9100, dev avg loss 0.232153, throughput 9.56395K wps
[Epoch 131 Batch 30/162] avg loss 0.00106128, throughput 9.50651K wps
[Epoch 131 Batch 60/162] avg loss 0.0012039, throughput 9.44674K wps
[Epoch 131 Batch 90/162] avg loss 0.00110875, throughput 9.40225K wps
[Epoch 131 Batch 120/162] avg loss 0.00116618, throughput 9.31515K wps
[Epoch 131 Batch 150/162] avg loss 0.0010405, throughput 9.46946K wps
Begin Testing...
[Epoch 131] train avg loss 0.001138, dev acc 0.9111, dev avg loss 0.231968, throughput 9.44279K wps
Observed Improvement.
Begin Testing...
[Epoch 132 Batch 30/162] avg loss 0.0010035, throughput 9.38632K wps
[Epoch 132 Batch 60/162] avg loss 0.00102871, throughput 9.24564K wps
[Epoch 132 Batch 90/162] avg loss 0.00118289, throughput 9.45862K wps
[Epoch 132 Batch 120/162] avg loss 0.00121869, throughput 9.52244K wps
[Epoch 132 Batch 150/162] avg loss 0.00115543, throughput 9.40658K wps
Begin Testing...
[Epoch 132] train avg loss 0.00112453, dev acc 0.9111, dev avg loss 0.232104, throughput 9.40687K wps
Observed Improvement.
Begin Testing...
[Epoch 133 Batch 30/162] avg loss 0.00117966, throughput 9.67746K wps
[Epoch 133 Batch 60/162] avg loss 0.00116795, throughput 9.26679K wps
[Epoch 133 Batch 90/162] avg loss 0.00108767, throughput 9.37785K wps
[Epoch 133 Batch 120/162] avg loss 0.00106559, throughput 9.34623K wps
[Epoch 133 Batch 150/162] avg loss 0.00102137, throughput 9.65554K wps
Begin Testing...
[Epoch 133] train avg loss 0.00111938, dev acc 0.9100, dev avg loss 0.234602, throughput 9.443K wps
[Epoch 134 Batch 30/162] avg loss 0.00101383, throughput 9.71491K wps
[Epoch 134 Batch 60/162] avg loss 0.000972838, throughput 9.56949K wps
[Epoch 134 Batch 90/162] avg loss 0.00104077, throughput 9.50926K wps
[Epoch 134 Batch 120/162] avg loss 0.0011692, throughput 9.63077K wps
[Epoch 134 Batch 150/162] avg loss 0.00119298, throughput 9.41511K wps
Begin Testing...
[Epoch 134] train avg loss 0.0010807, dev acc 0.9100, dev avg loss 0.233982, throughput 9.55638K wps
[Epoch 135 Batch 30/162] avg loss 0.00106688, throughput 9.57799K wps
[Epoch 135 Batch 60/162] avg loss 0.0010689, throughput 9.38195K wps
[Epoch 135 Batch 90/162] avg loss 0.00114211, throughput 9.5792K wps
[Epoch 135 Batch 120/162] avg loss 0.0010689, throughput 9.25316K wps
[Epoch 135 Batch 150/162] avg loss 0.00110241, throughput 9.43573K wps
Begin Testing...
[Epoch 135] train avg loss 0.00108973, dev acc 0.9111, dev avg loss 0.23218, throughput 9.43231K wps
Observed Improvement.
Begin Testing...
[Epoch 136 Batch 30/162] avg loss 0.000980985, throughput 9.68996K wps
[Epoch 136 Batch 60/162] avg loss 0.00104199, throughput 9.30821K wps
[Epoch 136 Batch 90/162] avg loss 0.00107606, throughput 9.67117K wps
[Epoch 136 Batch 120/162] avg loss 0.00114017, throughput 9.3432K wps
[Epoch 136 Batch 150/162] avg loss 0.00102443, throughput 9.4603K wps
Begin Testing...
[Epoch 136] train avg loss 0.00106263, dev acc 0.9111, dev avg loss 0.232865, throughput 9.4947K wps
Observed Improvement.
Begin Testing...
[Epoch 137 Batch 30/162] avg loss 0.00103074, throughput 9.68907K wps
[Epoch 137 Batch 60/162] avg loss 0.00107339, throughput 9.56503K wps
[Epoch 137 Batch 90/162] avg loss 0.00102165, throughput 9.41558K wps
[Epoch 137 Batch 120/162] avg loss 0.00108368, throughput 9.38937K wps
[Epoch 137 Batch 150/162] avg loss 0.000993215, throughput 9.39766K wps
Begin Testing...
[Epoch 137] train avg loss 0.00103855, dev acc 0.9111, dev avg loss 0.232873, throughput 9.49411K wps
Observed Improvement.
Begin Testing...
[Epoch 138 Batch 30/162] avg loss 0.000972933, throughput 9.42512K wps
[Epoch 138 Batch 60/162] avg loss 0.00115791, throughput 9.28952K wps
[Epoch 138 Batch 90/162] avg loss 0.00104724, throughput 9.39467K wps
[Epoch 138 Batch 120/162] avg loss 0.00105297, throughput 9.50163K wps
[Epoch 138 Batch 150/162] avg loss 0.000917481, throughput 9.35466K wps
Begin Testing...
[Epoch 138] train avg loss 0.00103118, dev acc 0.9144, dev avg loss 0.234059, throughput 9.39211K wps
Observed Improvement.
Begin Testing...
[Epoch 139 Batch 30/162] avg loss 0.00109139, throughput 9.79376K wps
[Epoch 139 Batch 60/162] avg loss 0.00105411, throughput 9.36916K wps
[Epoch 139 Batch 90/162] avg loss 0.000969425, throughput 9.49438K wps
[Epoch 139 Batch 120/162] avg loss 0.00111097, throughput 9.57678K wps
[Epoch 139 Batch 150/162] avg loss 0.000949307, throughput 9.45279K wps
Begin Testing...
[Epoch 139] train avg loss 0.00104557, dev acc 0.9122, dev avg loss 0.23201, throughput 9.51594K wps
[Epoch 140 Batch 30/162] avg loss 0.000927044, throughput 9.58159K wps
[Epoch 140 Batch 60/162] avg loss 0.00105573, throughput 9.37793K wps
[Epoch 140 Batch 90/162] avg loss 0.000932308, throughput 9.39764K wps
[Epoch 140 Batch 120/162] avg loss 0.00102175, throughput 9.23341K wps
[Epoch 140 Batch 150/162] avg loss 0.00100474, throughput 9.4418K wps
Begin Testing...
[Epoch 140] train avg loss 0.000986946, dev acc 0.9122, dev avg loss 0.232969, throughput 9.40564K wps
[Epoch 141 Batch 30/162] avg loss 0.00101291, throughput 9.62014K wps
[Epoch 141 Batch 60/162] avg loss 0.000887653, throughput 9.34667K wps
[Epoch 141 Batch 90/162] avg loss 0.00109438, throughput 9.57717K wps
[Epoch 141 Batch 120/162] avg loss 0.000916129, throughput 9.41817K wps
[Epoch 141 Batch 150/162] avg loss 0.00102618, throughput 9.62246K wps
Begin Testing...
[Epoch 141] train avg loss 0.000984817, dev acc 0.9111, dev avg loss 0.233115, throughput 9.52045K wps
[Epoch 142 Batch 30/162] avg loss 0.00105309, throughput 9.58606K wps
[Epoch 142 Batch 60/162] avg loss 0.000869127, throughput 9.54023K wps
[Epoch 142 Batch 90/162] avg loss 0.000880154, throughput 9.3614K wps
[Epoch 142 Batch 120/162] avg loss 0.00109837, throughput 9.44246K wps
[Epoch 142 Batch 150/162] avg loss 0.000986854, throughput 9.34643K wps
Begin Testing...
[Epoch 142] train avg loss 0.000978664, dev acc 0.9122, dev avg loss 0.232578, throughput 9.44319K wps
[Epoch 143 Batch 30/162] avg loss 0.00096577, throughput 9.5553K wps
[Epoch 143 Batch 60/162] avg loss 0.00106719, throughput 9.48256K wps
[Epoch 143 Batch 90/162] avg loss 0.00104406, throughput 9.37857K wps
[Epoch 143 Batch 120/162] avg loss 0.000808912, throughput 9.44979K wps
[Epoch 143 Batch 150/162] avg loss 0.000977087, throughput 9.27848K wps
Begin Testing...
[Epoch 143] train avg loss 0.000988134, dev acc 0.9111, dev avg loss 0.233524, throughput 9.42721K wps
[Epoch 144 Batch 30/162] avg loss 0.0010138, throughput 9.7637K wps
[Epoch 144 Batch 60/162] avg loss 0.000986695, throughput 9.37004K wps
[Epoch 144 Batch 90/162] avg loss 0.00100836, throughput 9.57314K wps
[Epoch 144 Batch 120/162] avg loss 0.000890342, throughput 9.5427K wps
[Epoch 144 Batch 150/162] avg loss 0.000948808, throughput 9.35804K wps
Begin Testing...
[Epoch 144] train avg loss 0.000971972, dev acc 0.9122, dev avg loss 0.233347, throughput 9.49839K wps
[Epoch 145 Batch 30/162] avg loss 0.00093396, throughput 9.53849K wps
[Epoch 145 Batch 60/162] avg loss 0.00100694, throughput 9.40178K wps
[Epoch 145 Batch 90/162] avg loss 0.000979731, throughput 9.33688K wps
[Epoch 145 Batch 120/162] avg loss 0.00101926, throughput 9.34463K wps
[Epoch 145 Batch 150/162] avg loss 0.000884236, throughput 9.41048K wps
Begin Testing...
[Epoch 145] train avg loss 0.00095896, dev acc 0.9133, dev avg loss 0.232439, throughput 9.40754K wps
[Epoch 146 Batch 30/162] avg loss 0.000932533, throughput 9.55042K wps
[Epoch 146 Batch 60/162] avg loss 0.000911497, throughput 9.33268K wps
[Epoch 146 Batch 90/162] avg loss 0.0008691, throughput 9.37202K wps
[Epoch 146 Batch 120/162] avg loss 0.00104923, throughput 9.25895K wps
[Epoch 146 Batch 150/162] avg loss 0.000867212, throughput 9.63283K wps
Begin Testing...
[Epoch 146] train avg loss 0.00093366, dev acc 0.9133, dev avg loss 0.232828, throughput 9.4156K wps
[Epoch 147 Batch 30/162] avg loss 0.00100815, throughput 9.67484K wps
[Epoch 147 Batch 60/162] avg loss 0.000990884, throughput 9.30786K wps
[Epoch 147 Batch 90/162] avg loss 0.00102449, throughput 9.36622K wps
[Epoch 147 Batch 120/162] avg loss 0.000910834, throughput 9.5053K wps
[Epoch 147 Batch 150/162] avg loss 0.000845734, throughput 9.33684K wps
Begin Testing...
[Epoch 147] train avg loss 0.000962246, dev acc 0.9144, dev avg loss 0.234393, throughput 9.41962K wps
Observed Improvement.
Begin Testing...
[Epoch 148 Batch 30/162] avg loss 0.00100481, throughput 9.59471K wps
[Epoch 148 Batch 60/162] avg loss 0.000963306, throughput 9.36321K wps
[Epoch 148 Batch 90/162] avg loss 0.000889151, throughput 9.43894K wps
[Epoch 148 Batch 120/162] avg loss 0.00089818, throughput 9.2641K wps
[Epoch 148 Batch 150/162] avg loss 0.000901616, throughput 9.32115K wps
Begin Testing...
[Epoch 148] train avg loss 0.000942506, dev acc 0.9144, dev avg loss 0.234285, throughput 9.38761K wps
Observed Improvement.
Begin Testing...
[Epoch 149 Batch 30/162] avg loss 0.000757587, throughput 9.76641K wps
[Epoch 149 Batch 60/162] avg loss 0.0010248, throughput 9.53268K wps
[Epoch 149 Batch 90/162] avg loss 0.00105774, throughput 9.45029K wps
[Epoch 149 Batch 120/162] avg loss 0.000885354, throughput 9.43846K wps
[Epoch 149 Batch 150/162] avg loss 0.00110434, throughput 9.30674K wps
Begin Testing...
[Epoch 149] train avg loss 0.000955154, dev acc 0.9100, dev avg loss 0.234515, throughput 9.48186K wps
[Epoch 150 Batch 30/162] avg loss 0.000906463, throughput 9.63746K wps
[Epoch 150 Batch 60/162] avg loss 0.00093466, throughput 9.31724K wps
[Epoch 150 Batch 90/162] avg loss 0.0010139, throughput 9.28901K wps
[Epoch 150 Batch 120/162] avg loss 0.000973083, throughput 9.59441K wps
[Epoch 150 Batch 150/162] avg loss 0.000830205, throughput 9.40035K wps
Begin Testing...
[Epoch 150] train avg loss 0.000934547, dev acc 0.9133, dev avg loss 0.23411, throughput 9.45929K wps
[Epoch 151 Batch 30/162] avg loss 0.000884044, throughput 9.72137K wps
[Epoch 151 Batch 60/162] avg loss 0.000929398, throughput 9.27369K wps
[Epoch 151 Batch 90/162] avg loss 0.00085073, throughput 9.61833K wps
[Epoch 151 Batch 120/162] avg loss 0.00093405, throughput 9.39246K wps
[Epoch 151 Batch 150/162] avg loss 0.000894171, throughput 9.43711K wps
Begin Testing...
[Epoch 151] train avg loss 0.000899789, dev acc 0.9111, dev avg loss 0.235084, throughput 9.47939K wps
[Epoch 152 Batch 30/162] avg loss 0.000924412, throughput 9.74035K wps
[Epoch 152 Batch 60/162] avg loss 0.00089018, throughput 9.30607K wps
[Epoch 152 Batch 90/162] avg loss 0.000843307, throughput 9.52359K wps
[Epoch 152 Batch 120/162] avg loss 0.000936487, throughput 9.20592K wps
[Epoch 152 Batch 150/162] avg loss 0.000921904, throughput 9.48166K wps
Begin Testing...
[Epoch 152] train avg loss 0.000903886, dev acc 0.9111, dev avg loss 0.236854, throughput 9.43963K wps
[Epoch 153 Batch 30/162] avg loss 0.00080144, throughput 9.64852K wps
[Epoch 153 Batch 60/162] avg loss 0.000926004, throughput 9.48433K wps
[Epoch 153 Batch 90/162] avg loss 0.000785056, throughput 9.5037K wps
[Epoch 153 Batch 120/162] avg loss 0.000929932, throughput 9.45341K wps
[Epoch 153 Batch 150/162] avg loss 0.000898757, throughput 9.5055K wps
Begin Testing...
[Epoch 153] train avg loss 0.000852379, dev acc 0.9144, dev avg loss 0.236154, throughput 9.50263K wps
Observed Improvement.
Begin Testing...
[Epoch 154 Batch 30/162] avg loss 0.000809562, throughput 9.675K wps
[Epoch 154 Batch 60/162] avg loss 0.000874498, throughput 9.55421K wps
[Epoch 154 Batch 90/162] avg loss 0.000827335, throughput 9.32593K wps
[Epoch 154 Batch 120/162] avg loss 0.000794934, throughput 9.47148K wps
[Epoch 154 Batch 150/162] avg loss 0.000845916, throughput 9.46778K wps
Begin Testing...
[Epoch 154] train avg loss 0.000836829, dev acc 0.9144, dev avg loss 0.236225, throughput 9.49423K wps
Observed Improvement.
Begin Testing...
[Epoch 155 Batch 30/162] avg loss 0.000914239, throughput 9.53265K wps
[Epoch 155 Batch 60/162] avg loss 0.000759744, throughput 9.49944K wps
[Epoch 155 Batch 90/162] avg loss 0.000816025, throughput 9.47016K wps
[Epoch 155 Batch 120/162] avg loss 0.000905486, throughput 9.3818K wps
[Epoch 155 Batch 150/162] avg loss 0.000783079, throughput 9.37823K wps
Begin Testing...
[Epoch 155] train avg loss 0.000838449, dev acc 0.9122, dev avg loss 0.235383, throughput 9.44621K wps
[Epoch 156 Batch 30/162] avg loss 0.000902768, throughput 9.49009K wps
[Epoch 156 Batch 60/162] avg loss 0.000830216, throughput 9.24046K wps
[Epoch 156 Batch 90/162] avg loss 0.000890692, throughput 9.56665K wps
[Epoch 156 Batch 120/162] avg loss 0.00100251, throughput 9.58509K wps
[Epoch 156 Batch 150/162] avg loss 0.000777453, throughput 9.4326K wps
Begin Testing...
[Epoch 156] train avg loss 0.000885556, dev acc 0.9111, dev avg loss 0.237453, throughput 9.44524K wps
[Epoch 157 Batch 30/162] avg loss 0.000874128, throughput 9.66153K wps
[Epoch 157 Batch 60/162] avg loss 0.000829836, throughput 9.69994K wps
[Epoch 157 Batch 90/162] avg loss 0.000917275, throughput 9.41887K wps
[Epoch 157 Batch 120/162] avg loss 0.000751024, throughput 9.21542K wps
[Epoch 157 Batch 150/162] avg loss 0.000810742, throughput 9.49258K wps
Begin Testing...
[Epoch 157] train avg loss 0.000832229, dev acc 0.9144, dev avg loss 0.237423, throughput 9.50007K wps
Observed Improvement.
Begin Testing...
[Epoch 158 Batch 30/162] avg loss 0.000753415, throughput 9.76525K wps
[Epoch 158 Batch 60/162] avg loss 0.000817794, throughput 9.5279K wps
[Epoch 158 Batch 90/162] avg loss 0.000844788, throughput 9.38867K wps
[Epoch 158 Batch 120/162] avg loss 0.000863531, throughput 9.40778K wps
[Epoch 158 Batch 150/162] avg loss 0.000852283, throughput 9.35071K wps
Begin Testing...
[Epoch 158] train avg loss 0.000827078, dev acc 0.9133, dev avg loss 0.236164, throughput 9.47564K wps
[Epoch 159 Batch 30/162] avg loss 0.00079698, throughput 9.56514K wps
[Epoch 159 Batch 60/162] avg loss 0.000874239, throughput 9.34371K wps
[Epoch 159 Batch 90/162] avg loss 0.000780831, throughput 9.22309K wps
[Epoch 159 Batch 120/162] avg loss 0.000815178, throughput 9.43224K wps
[Epoch 159 Batch 150/162] avg loss 0.000859414, throughput 9.60454K wps
Begin Testing...
[Epoch 159] train avg loss 0.000824637, dev acc 0.9144, dev avg loss 0.23536, throughput 9.43062K wps
Observed Improvement.
Begin Testing...
[Epoch 160 Batch 30/162] avg loss 0.000864249, throughput 9.82922K wps
[Epoch 160 Batch 60/162] avg loss 0.000864273, throughput 9.25349K wps
[Epoch 160 Batch 90/162] avg loss 0.000894273, throughput 9.20198K wps
[Epoch 160 Batch 120/162] avg loss 0.000729374, throughput 9.41466K wps
[Epoch 160 Batch 150/162] avg loss 0.000769394, throughput 9.41177K wps
Begin Testing...
[Epoch 160] train avg loss 0.000823009, dev acc 0.9133, dev avg loss 0.235804, throughput 9.39932K wps
[Epoch 161 Batch 30/162] avg loss 0.00087648, throughput 9.61124K wps
[Epoch 161 Batch 60/162] avg loss 0.000777377, throughput 9.44051K wps
[Epoch 161 Batch 90/162] avg loss 0.000800032, throughput 9.29807K wps
[Epoch 161 Batch 120/162] avg loss 0.000801509, throughput 9.3849K wps
[Epoch 161 Batch 150/162] avg loss 0.000840791, throughput 9.41642K wps
Begin Testing...
[Epoch 161] train avg loss 0.000817239, dev acc 0.9144, dev avg loss 0.235484, throughput 9.44081K wps
Observed Improvement.
Begin Testing...
[Epoch 162 Batch 30/162] avg loss 0.000776994, throughput 9.503K wps
[Epoch 162 Batch 60/162] avg loss 0.00074728, throughput 9.37362K wps
[Epoch 162 Batch 90/162] avg loss 0.00075795, throughput 9.45251K wps
[Epoch 162 Batch 120/162] avg loss 0.000736203, throughput 9.28863K wps
[Epoch 162 Batch 150/162] avg loss 0.000900352, throughput 9.52413K wps
Begin Testing...
[Epoch 162] train avg loss 0.000785256, dev acc 0.9122, dev avg loss 0.237575, throughput 9.44251K wps
[Epoch 163 Batch 30/162] avg loss 0.000781497, throughput 9.7172K wps
[Epoch 163 Batch 60/162] avg loss 0.000681183, throughput 9.45398K wps
[Epoch 163 Batch 90/162] avg loss 0.000800926, throughput 9.39823K wps
[Epoch 163 Batch 120/162] avg loss 0.000749858, throughput 9.23606K wps
[Epoch 163 Batch 150/162] avg loss 0.000820587, throughput 9.37697K wps
Begin Testing...
[Epoch 163] train avg loss 0.000760974, dev acc 0.9144, dev avg loss 0.237828, throughput 9.43499K wps
Observed Improvement.
Begin Testing...
[Epoch 164 Batch 30/162] avg loss 0.000834988, throughput 9.65232K wps
[Epoch 164 Batch 60/162] avg loss 0.000749508, throughput 9.45462K wps
[Epoch 164 Batch 90/162] avg loss 0.000792896, throughput 9.50808K wps
[Epoch 164 Batch 120/162] avg loss 0.000769581, throughput 9.37155K wps
[Epoch 164 Batch 150/162] avg loss 0.000786567, throughput 9.37511K wps
Begin Testing...
[Epoch 164] train avg loss 0.00079725, dev acc 0.9133, dev avg loss 0.237895, throughput 9.46094K wps
[Epoch 165 Batch 30/162] avg loss 0.000767824, throughput 9.51165K wps
[Epoch 165 Batch 60/162] avg loss 0.000716005, throughput 9.37219K wps
[Epoch 165 Batch 90/162] avg loss 0.000955178, throughput 9.34643K wps
[Epoch 165 Batch 120/162] avg loss 0.000787005, throughput 9.47038K wps
[Epoch 165 Batch 150/162] avg loss 0.000818594, throughput 9.27629K wps
Begin Testing...
[Epoch 165] train avg loss 0.00079942, dev acc 0.9144, dev avg loss 0.236626, throughput 9.39118K wps
Observed Improvement.
Begin Testing...
[Epoch 166 Batch 30/162] avg loss 0.000829008, throughput 9.44915K wps
[Epoch 166 Batch 60/162] avg loss 0.000756029, throughput 9.36664K wps
[Epoch 166 Batch 90/162] avg loss 0.000797729, throughput 9.54714K wps
[Epoch 166 Batch 120/162] avg loss 0.000872829, throughput 9.54003K wps
[Epoch 166 Batch 150/162] avg loss 0.000797752, throughput 9.3665K wps
Begin Testing...
[Epoch 166] train avg loss 0.000817927, dev acc 0.9111, dev avg loss 0.238333, throughput 9.45563K wps
[Epoch 167 Batch 30/162] avg loss 0.000754487, throughput 9.59386K wps
[Epoch 167 Batch 60/162] avg loss 0.00073782, throughput 9.31198K wps
[Epoch 167 Batch 90/162] avg loss 0.00082441, throughput 9.47612K wps
[Epoch 167 Batch 120/162] avg loss 0.00083755, throughput 9.47738K wps
[Epoch 167 Batch 150/162] avg loss 0.000729551, throughput 9.38583K wps
Begin Testing...
[Epoch 167] train avg loss 0.000772936, dev acc 0.9144, dev avg loss 0.236246, throughput 9.43843K wps
Observed Improvement.
Begin Testing...
[Epoch 168 Batch 30/162] avg loss 0.00079958, throughput 9.70249K wps
[Epoch 168 Batch 60/162] avg loss 0.0006881, throughput 9.5871K wps
[Epoch 168 Batch 90/162] avg loss 0.000731263, throughput 9.42054K wps
[Epoch 168 Batch 120/162] avg loss 0.00068885, throughput 9.56498K wps
[Epoch 168 Batch 150/162] avg loss 0.000832568, throughput 9.23526K wps
Begin Testing...
[Epoch 168] train avg loss 0.00074904, dev acc 0.9156, dev avg loss 0.237294, throughput 9.4877K wps
Observed Improvement.
Begin Testing...
[Epoch 169 Batch 30/162] avg loss 0.000813106, throughput 9.64354K wps
[Epoch 169 Batch 60/162] avg loss 0.000764835, throughput 9.38687K wps
[Epoch 169 Batch 90/162] avg loss 0.000741596, throughput 9.5163K wps
[Epoch 169 Batch 120/162] avg loss 0.000674529, throughput 9.26677K wps
[Epoch 169 Batch 150/162] avg loss 0.000723537, throughput 9.31368K wps
Begin Testing...
[Epoch 169] train avg loss 0.000745515, dev acc 0.9156, dev avg loss 0.23677, throughput 9.43191K wps
Observed Improvement.
Begin Testing...
[Epoch 170 Batch 30/162] avg loss 0.000687942, throughput 9.58953K wps
[Epoch 170 Batch 60/162] avg loss 0.000720402, throughput 9.42476K wps
[Epoch 170 Batch 90/162] avg loss 0.000722651, throughput 9.32119K wps
[Epoch 170 Batch 120/162] avg loss 0.000852033, throughput 9.40147K wps
[Epoch 170 Batch 150/162] avg loss 0.000701915, throughput 9.42697K wps
Begin Testing...
[Epoch 170] train avg loss 0.000747658, dev acc 0.9111, dev avg loss 0.238668, throughput 9.42518K wps
[Epoch 171 Batch 30/162] avg loss 0.000681045, throughput 9.64017K wps
[Epoch 171 Batch 60/162] avg loss 0.000726446, throughput 9.4019K wps
[Epoch 171 Batch 90/162] avg loss 0.000699263, throughput 9.48402K wps
[Epoch 171 Batch 120/162] avg loss 0.000717328, throughput 9.43173K wps
[Epoch 171 Batch 150/162] avg loss 0.000704462, throughput 9.54714K wps
Begin Testing...
[Epoch 171] train avg loss 0.000714912, dev acc 0.9156, dev avg loss 0.23746, throughput 9.47291K wps
Observed Improvement.
Begin Testing...
[Epoch 172 Batch 30/162] avg loss 0.000764683, throughput 9.70724K wps
[Epoch 172 Batch 60/162] avg loss 0.000739441, throughput 9.4853K wps
[Epoch 172 Batch 90/162] avg loss 0.000764415, throughput 9.50287K wps
[Epoch 172 Batch 120/162] avg loss 0.000803184, throughput 9.5304K wps
[Epoch 172 Batch 150/162] avg loss 0.000746847, throughput 9.38065K wps
Begin Testing...
[Epoch 172] train avg loss 0.000762464, dev acc 0.9156, dev avg loss 0.23762, throughput 9.50558K wps
Observed Improvement.
Begin Testing...
[Epoch 173 Batch 30/162] avg loss 0.000828047, throughput 9.56269K wps
[Epoch 173 Batch 60/162] avg loss 0.000771306, throughput 9.52749K wps
[Epoch 173 Batch 90/162] avg loss 0.000776211, throughput 9.47328K wps
[Epoch 173 Batch 120/162] avg loss 0.0006452, throughput 9.37725K wps
[Epoch 173 Batch 150/162] avg loss 0.00072588, throughput 9.47474K wps
Begin Testing...
[Epoch 173] train avg loss 0.000757441, dev acc 0.9144, dev avg loss 0.237971, throughput 9.46995K wps
[Epoch 174 Batch 30/162] avg loss 0.000750212, throughput 9.6125K wps
[Epoch 174 Batch 60/162] avg loss 0.000711264, throughput 9.36051K wps
[Epoch 174 Batch 90/162] avg loss 0.00067877, throughput 9.44997K wps
[Epoch 174 Batch 120/162] avg loss 0.00068847, throughput 9.43314K wps
[Epoch 174 Batch 150/162] avg loss 0.000632529, throughput 9.40916K wps
Begin Testing...
[Epoch 174] train avg loss 0.000697922, dev acc 0.9156, dev avg loss 0.237142, throughput 9.44105K wps
Observed Improvement.
Begin Testing...
[Epoch 175 Batch 30/162] avg loss 0.000670086, throughput 9.48812K wps
[Epoch 175 Batch 60/162] avg loss 0.000708765, throughput 9.45834K wps
[Epoch 175 Batch 90/162] avg loss 0.0006821, throughput 9.38404K wps
[Epoch 175 Batch 120/162] avg loss 0.000664069, throughput 9.59666K wps
[Epoch 175 Batch 150/162] avg loss 0.000747414, throughput 9.50297K wps
Begin Testing...
[Epoch 175] train avg loss 0.000693981, dev acc 0.9156, dev avg loss 0.237402, throughput 9.47869K wps
Observed Improvement.
Begin Testing...
[Epoch 176 Batch 30/162] avg loss 0.000780802, throughput 9.55923K wps
[Epoch 176 Batch 60/162] avg loss 0.000669394, throughput 9.34671K wps
[Epoch 176 Batch 90/162] avg loss 0.000682581, throughput 9.53023K wps
[Epoch 176 Batch 120/162] avg loss 0.000650482, throughput 9.39608K wps
[Epoch 176 Batch 150/162] avg loss 0.000650116, throughput 9.3975K wps
Begin Testing...
[Epoch 176] train avg loss 0.00068729, dev acc 0.9144, dev avg loss 0.23812, throughput 9.43322K wps
[Epoch 177 Batch 30/162] avg loss 0.000716022, throughput 9.48429K wps
[Epoch 177 Batch 60/162] avg loss 0.000805225, throughput 9.49015K wps
[Epoch 177 Batch 90/162] avg loss 0.000680471, throughput 9.43213K wps
[Epoch 177 Batch 120/162] avg loss 0.000658924, throughput 9.4411K wps
[Epoch 177 Batch 150/162] avg loss 0.000689563, throughput 9.36331K wps
Begin Testing...
[Epoch 177] train avg loss 0.000705624, dev acc 0.9144, dev avg loss 0.238114, throughput 9.44612K wps
[Epoch 178 Batch 30/162] avg loss 0.000732357, throughput 9.61901K wps
[Epoch 178 Batch 60/162] avg loss 0.000680715, throughput 9.26228K wps
[Epoch 178 Batch 90/162] avg loss 0.000644196, throughput 9.60399K wps
[Epoch 178 Batch 120/162] avg loss 0.000659631, throughput 9.40373K wps
[Epoch 178 Batch 150/162] avg loss 0.000693343, throughput 9.52585K wps
Begin Testing...
[Epoch 178] train avg loss 0.000683294, dev acc 0.9144, dev avg loss 0.238199, throughput 9.47297K wps
[Epoch 179 Batch 30/162] avg loss 0.00068776, throughput 9.845K wps
[Epoch 179 Batch 60/162] avg loss 0.000642812, throughput 9.4262K wps
[Epoch 179 Batch 90/162] avg loss 0.000622357, throughput 9.29547K wps
[Epoch 179 Batch 120/162] avg loss 0.000713515, throughput 9.59705K wps
[Epoch 179 Batch 150/162] avg loss 0.000753215, throughput 9.50812K wps
Begin Testing...
[Epoch 179] train avg loss 0.000689928, dev acc 0.9122, dev avg loss 0.238651, throughput 9.53214K wps
[Epoch 180 Batch 30/162] avg loss 0.000640087, throughput 9.67519K wps
[Epoch 180 Batch 60/162] avg loss 0.000736041, throughput 9.24171K wps
[Epoch 180 Batch 90/162] avg loss 0.000692541, throughput 9.55277K wps
[Epoch 180 Batch 120/162] avg loss 0.000630193, throughput 9.51878K wps
[Epoch 180 Batch 150/162] avg loss 0.000601577, throughput 9.31569K wps
Begin Testing...
[Epoch 180] train avg loss 0.000663712, dev acc 0.9133, dev avg loss 0.2394, throughput 9.466K wps
[Epoch 181 Batch 30/162] avg loss 0.000642536, throughput 9.46184K wps
[Epoch 181 Batch 60/162] avg loss 0.000650155, throughput 9.41155K wps
[Epoch 181 Batch 90/162] avg loss 0.00070878, throughput 9.50996K wps
[Epoch 181 Batch 120/162] avg loss 0.000738268, throughput 9.34474K wps
[Epoch 181 Batch 150/162] avg loss 0.000645051, throughput 9.37267K wps
Begin Testing...
[Epoch 181] train avg loss 0.000681017, dev acc 0.9144, dev avg loss 0.238441, throughput 9.41465K wps
[Epoch 182 Batch 30/162] avg loss 0.000618705, throughput 9.55039K wps
[Epoch 182 Batch 60/162] avg loss 0.000735763, throughput 9.4283K wps
[Epoch 182 Batch 90/162] avg loss 0.00068169, throughput 9.50858K wps
[Epoch 182 Batch 120/162] avg loss 0.000628542, throughput 9.51514K wps
[Epoch 182 Batch 150/162] avg loss 0.000637917, throughput 9.30085K wps
Begin Testing...
[Epoch 182] train avg loss 0.00067241, dev acc 0.9100, dev avg loss 0.241031, throughput 9.46862K wps
[Epoch 183 Batch 30/162] avg loss 0.00072351, throughput 9.60913K wps
[Epoch 183 Batch 60/162] avg loss 0.000634973, throughput 9.30007K wps
[Epoch 183 Batch 90/162] avg loss 0.000701741, throughput 9.42942K wps
[Epoch 183 Batch 120/162] avg loss 0.000638283, throughput 9.57968K wps
[Epoch 183 Batch 150/162] avg loss 0.000636179, throughput 9.41505K wps
Begin Testing...
[Epoch 183] train avg loss 0.000672907, dev acc 0.9133, dev avg loss 0.240599, throughput 9.46029K wps
[Epoch 184 Batch 30/162] avg loss 0.000643019, throughput 9.58981K wps
[Epoch 184 Batch 60/162] avg loss 0.000620378, throughput 9.41458K wps
[Epoch 184 Batch 90/162] avg loss 0.0006748, throughput 9.43454K wps
[Epoch 184 Batch 120/162] avg loss 0.000683673, throughput 9.40034K wps
[Epoch 184 Batch 150/162] avg loss 0.000664289, throughput 9.43K wps
Begin Testing...
[Epoch 184] train avg loss 0.000655193, dev acc 0.9133, dev avg loss 0.2395, throughput 9.47462K wps
[Epoch 185 Batch 30/162] avg loss 0.00062896, throughput 9.56471K wps
[Epoch 185 Batch 60/162] avg loss 0.000644263, throughput 9.36118K wps
[Epoch 185 Batch 90/162] avg loss 0.000656446, throughput 9.2888K wps
[Epoch 185 Batch 120/162] avg loss 0.000692776, throughput 9.35534K wps
[Epoch 185 Batch 150/162] avg loss 0.000670675, throughput 9.5091K wps
Begin Testing...
[Epoch 185] train avg loss 0.000662057, dev acc 0.9144, dev avg loss 0.239206, throughput 9.42691K wps
[Epoch 186 Batch 30/162] avg loss 0.00064991, throughput 9.60526K wps
[Epoch 186 Batch 60/162] avg loss 0.000572278, throughput 9.36914K wps
[Epoch 186 Batch 90/162] avg loss 0.000685476, throughput 9.32833K wps
[Epoch 186 Batch 120/162] avg loss 0.000598402, throughput 9.55251K wps
[Epoch 186 Batch 150/162] avg loss 0.000617555, throughput 9.3276K wps
Begin Testing...
[Epoch 186] train avg loss 0.000625329, dev acc 0.9144, dev avg loss 0.239312, throughput 9.43554K wps
[Epoch 187 Batch 30/162] avg loss 0.000671052, throughput 9.64039K wps
[Epoch 187 Batch 60/162] avg loss 0.000589315, throughput 9.34484K wps
[Epoch 187 Batch 90/162] avg loss 0.000636419, throughput 9.3342K wps
[Epoch 187 Batch 120/162] avg loss 0.000618316, throughput 9.42849K wps
[Epoch 187 Batch 150/162] avg loss 0.000720053, throughput 9.39445K wps
Begin Testing...
[Epoch 187] train avg loss 0.000649439, dev acc 0.9144, dev avg loss 0.240043, throughput 9.43347K wps
[Epoch 188 Batch 30/162] avg loss 0.000603069, throughput 9.65358K wps
[Epoch 188 Batch 60/162] avg loss 0.000612483, throughput 9.37045K wps
[Epoch 188 Batch 90/162] avg loss 0.000752463, throughput 9.45928K wps
[Epoch 188 Batch 120/162] avg loss 0.000605337, throughput 9.55011K wps
[Epoch 188 Batch 150/162] avg loss 0.000602911, throughput 9.28674K wps
Begin Testing...
[Epoch 188] train avg loss 0.000639905, dev acc 0.9144, dev avg loss 0.239383, throughput 9.44578K wps
[Epoch 189 Batch 30/162] avg loss 0.000599811, throughput 9.5491K wps
[Epoch 189 Batch 60/162] avg loss 0.000665367, throughput 9.3089K wps
[Epoch 189 Batch 90/162] avg loss 0.000544917, throughput 9.38214K wps
[Epoch 189 Batch 120/162] avg loss 0.00061743, throughput 9.42916K wps
[Epoch 189 Batch 150/162] avg loss 0.000568779, throughput 9.36303K wps
Begin Testing...
[Epoch 189] train avg loss 0.000605034, dev acc 0.9144, dev avg loss 0.240067, throughput 9.41094K wps
[Epoch 190 Batch 30/162] avg loss 0.000646258, throughput 9.49799K wps
[Epoch 190 Batch 60/162] avg loss 0.000590138, throughput 9.55964K wps
[Epoch 190 Batch 90/162] avg loss 0.000550726, throughput 9.39024K wps
[Epoch 190 Batch 120/162] avg loss 0.000667505, throughput 9.30618K wps
[Epoch 190 Batch 150/162] avg loss 0.00053516, throughput 9.46645K wps
Begin Testing...
[Epoch 190] train avg loss 0.000597574, dev acc 0.9144, dev avg loss 0.240258, throughput 9.45056K wps
[Epoch 191 Batch 30/162] avg loss 0.000610235, throughput 9.72878K wps
[Epoch 191 Batch 60/162] avg loss 0.000599775, throughput 9.56457K wps
[Epoch 191 Batch 90/162] avg loss 0.000563973, throughput 9.61996K wps
[Epoch 191 Batch 120/162] avg loss 0.000636488, throughput 9.41013K wps
[Epoch 191 Batch 150/162] avg loss 0.000579071, throughput 9.52341K wps
Begin Testing...
[Epoch 191] train avg loss 0.00059574, dev acc 0.9156, dev avg loss 0.240749, throughput 9.54506K wps
Observed Improvement.
Begin Testing...
[Epoch 192 Batch 30/162] avg loss 0.000631283, throughput 9.66753K wps
[Epoch 192 Batch 60/162] avg loss 0.000575503, throughput 9.38822K wps
[Epoch 192 Batch 90/162] avg loss 0.000676294, throughput 9.6138K wps
[Epoch 192 Batch 120/162] avg loss 0.000655896, throughput 9.55832K wps
[Epoch 192 Batch 150/162] avg loss 0.000602517, throughput 9.37693K wps
Begin Testing...
[Epoch 192] train avg loss 0.00062065, dev acc 0.9144, dev avg loss 0.239918, throughput 9.50411K wps
[Epoch 193 Batch 30/162] avg loss 0.000665896, throughput 9.75444K wps
[Epoch 193 Batch 60/162] avg loss 0.000587303, throughput 9.30401K wps
[Epoch 193 Batch 90/162] avg loss 0.000544072, throughput 9.52294K wps
[Epoch 193 Batch 120/162] avg loss 0.000681864, throughput 9.39357K wps
[Epoch 193 Batch 150/162] avg loss 0.000640455, throughput 9.51928K wps
Begin Testing...
[Epoch 193] train avg loss 0.000633099, dev acc 0.9167, dev avg loss 0.238982, throughput 9.4913K wps
Observed Improvement.
Begin Testing...
[Epoch 194 Batch 30/162] avg loss 0.000680255, throughput 9.70123K wps
[Epoch 194 Batch 60/162] avg loss 0.000520306, throughput 9.32461K wps
[Epoch 194 Batch 90/162] avg loss 0.000577253, throughput 9.37473K wps
[Epoch 194 Batch 120/162] avg loss 0.00066961, throughput 9.28598K wps
[Epoch 194 Batch 150/162] avg loss 0.00055919, throughput 9.49426K wps
Begin Testing...
[Epoch 194] train avg loss 0.000602644, dev acc 0.9133, dev avg loss 0.240838, throughput 9.42909K wps
[Epoch 195 Batch 30/162] avg loss 0.000520735, throughput 9.64109K wps
[Epoch 195 Batch 60/162] avg loss 0.000567012, throughput 9.56116K wps
[Epoch 195 Batch 90/162] avg loss 0.000584183, throughput 9.35941K wps
[Epoch 195 Batch 120/162] avg loss 0.000662558, throughput 9.43725K wps
[Epoch 195 Batch 150/162] avg loss 0.000596551, throughput 9.38135K wps
Begin Testing...
[Epoch 195] train avg loss 0.000587296, dev acc 0.9144, dev avg loss 0.240449, throughput 9.47305K wps
[Epoch 196 Batch 30/162] avg loss 0.000679323, throughput 9.45091K wps
[Epoch 196 Batch 60/162] avg loss 0.000539211, throughput 9.50913K wps
[Epoch 196 Batch 90/162] avg loss 0.000590472, throughput 9.39067K wps
[Epoch 196 Batch 120/162] avg loss 0.000548879, throughput 9.44091K wps
[Epoch 196 Batch 150/162] avg loss 0.000577221, throughput 9.46251K wps
Begin Testing...
[Epoch 196] train avg loss 0.000584299, dev acc 0.9156, dev avg loss 0.241365, throughput 9.45446K wps
[Epoch 197 Batch 30/162] avg loss 0.000554944, throughput 9.58904K wps
[Epoch 197 Batch 60/162] avg loss 0.000590533, throughput 9.36049K wps
[Epoch 197 Batch 90/162] avg loss 0.00055171, throughput 9.36764K wps
[Epoch 197 Batch 120/162] avg loss 0.000560373, throughput 9.46759K wps
[Epoch 197 Batch 150/162] avg loss 0.000587636, throughput 9.40665K wps
Begin Testing...
[Epoch 197] train avg loss 0.000568869, dev acc 0.9156, dev avg loss 0.24096, throughput 9.43809K wps
[Epoch 198 Batch 30/162] avg loss 0.000609953, throughput 9.65466K wps
[Epoch 198 Batch 60/162] avg loss 0.000663976, throughput 9.48664K wps
[Epoch 198 Batch 90/162] avg loss 0.000546994, throughput 9.35625K wps
[Epoch 198 Batch 120/162] avg loss 0.000533446, throughput 9.60864K wps
[Epoch 198 Batch 150/162] avg loss 0.000598003, throughput 9.48772K wps
Begin Testing...
[Epoch 198] train avg loss 0.000591589, dev acc 0.9156, dev avg loss 0.240835, throughput 9.49089K wps
[Epoch 199 Batch 30/162] avg loss 0.000524167, throughput 9.5962K wps
[Epoch 199 Batch 60/162] avg loss 0.000583836, throughput 9.50154K wps
[Epoch 199 Batch 90/162] avg loss 0.000660868, throughput 9.50269K wps
[Epoch 199 Batch 120/162] avg loss 0.000653351, throughput 9.51967K wps
[Epoch 199 Batch 150/162] avg loss 0.000548174, throughput 9.38569K wps
Begin Testing...
[Epoch 199] train avg loss 0.000588433, dev acc 0.9144, dev avg loss 0.240379, throughput 9.50809K wps
Test loss 0.209982, test acc 0.9160
Total time cost 445.33s
[Epoch 0 Batch 30/162] avg loss 0.0139577, throughput 7.32806K wps
[Epoch 0 Batch 60/162] avg loss 0.0137784, throughput 9.49605K wps
[Epoch 0 Batch 90/162] avg loss 0.013721, throughput 9.40992K wps
[Epoch 0 Batch 120/162] avg loss 0.0136148, throughput 9.43605K wps
[Epoch 0 Batch 150/162] avg loss 0.0134209, throughput 9.46936K wps
Begin Testing...
[Epoch 0] train avg loss 0.0136806, dev acc 0.7678, dev avg loss 0.660014, throughput 8.96481K wps
Observed Improvement.
Begin Testing...
[Epoch 1 Batch 30/162] avg loss 0.0133773, throughput 9.57047K wps
[Epoch 1 Batch 60/162] avg loss 0.0132626, throughput 9.32376K wps
[Epoch 1 Batch 90/162] avg loss 0.0131511, throughput 9.38731K wps
[Epoch 1 Batch 120/162] avg loss 0.0129115, throughput 9.5167K wps
[Epoch 1 Batch 150/162] avg loss 0.0128618, throughput 9.38969K wps
Begin Testing...
[Epoch 1] train avg loss 0.0131133, dev acc 0.8356, dev avg loss 0.627847, throughput 9.42531K wps
Observed Improvement.
Begin Testing...
[Epoch 2 Batch 30/162] avg loss 0.0125502, throughput 9.49925K wps
[Epoch 2 Batch 60/162] avg loss 0.0124798, throughput 9.44233K wps
[Epoch 2 Batch 90/162] avg loss 0.0123957, throughput 9.58979K wps
[Epoch 2 Batch 120/162] avg loss 0.0122335, throughput 9.33945K wps
[Epoch 2 Batch 150/162] avg loss 0.0121011, throughput 9.41543K wps
Begin Testing...
[Epoch 2] train avg loss 0.0123434, dev acc 0.8544, dev avg loss 0.592322, throughput 9.45586K wps
Observed Improvement.
Begin Testing...
[Epoch 3 Batch 30/162] avg loss 0.0120874, throughput 9.46417K wps
[Epoch 3 Batch 60/162] avg loss 0.0118032, throughput 9.37873K wps
[Epoch 3 Batch 90/162] avg loss 0.0117256, throughput 9.32899K wps
[Epoch 3 Batch 120/162] avg loss 0.0114886, throughput 9.23493K wps
[Epoch 3 Batch 150/162] avg loss 0.0114642, throughput 9.27576K wps
Begin Testing...
[Epoch 3] train avg loss 0.0116819, dev acc 0.8644, dev avg loss 0.551927, throughput 9.33025K wps
Observed Improvement.
Begin Testing...
[Epoch 4 Batch 30/162] avg loss 0.0111686, throughput 9.48278K wps
[Epoch 4 Batch 60/162] avg loss 0.0111451, throughput 9.50001K wps
[Epoch 4 Batch 90/162] avg loss 0.0107765, throughput 9.28584K wps
[Epoch 4 Batch 120/162] avg loss 0.0105843, throughput 9.58707K wps
[Epoch 4 Batch 150/162] avg loss 0.0106473, throughput 9.54761K wps
Begin Testing...
[Epoch 4] train avg loss 0.0108324, dev acc 0.8678, dev avg loss 0.509649, throughput 9.48645K wps
Observed Improvement.
Begin Testing...
[Epoch 5 Batch 30/162] avg loss 0.0103033, throughput 9.48878K wps
[Epoch 5 Batch 60/162] avg loss 0.0103413, throughput 9.37991K wps
[Epoch 5 Batch 90/162] avg loss 0.0100998, throughput 9.36053K wps
[Epoch 5 Batch 120/162] avg loss 0.00992682, throughput 9.28552K wps
[Epoch 5 Batch 150/162] avg loss 0.0097555, throughput 9.51971K wps
Begin Testing...
[Epoch 5] train avg loss 0.0100511, dev acc 0.8689, dev avg loss 0.467881, throughput 9.39704K wps
Observed Improvement.
Begin Testing...
[Epoch 6 Batch 30/162] avg loss 0.00962422, throughput 9.60577K wps
[Epoch 6 Batch 60/162] avg loss 0.00952823, throughput 9.41056K wps
[Epoch 6 Batch 90/162] avg loss 0.00913258, throughput 9.3573K wps
[Epoch 6 Batch 120/162] avg loss 0.00908532, throughput 9.31946K wps
[Epoch 6 Batch 150/162] avg loss 0.00884828, throughput 9.3235K wps
Begin Testing...
[Epoch 6] train avg loss 0.00922888, dev acc 0.8733, dev avg loss 0.430268, throughput 9.38119K wps
Observed Improvement.
Begin Testing...
[Epoch 7 Batch 30/162] avg loss 0.00876181, throughput 9.52579K wps
[Epoch 7 Batch 60/162] avg loss 0.00858846, throughput 9.40672K wps
[Epoch 7 Batch 90/162] avg loss 0.00856946, throughput 9.48118K wps
[Epoch 7 Batch 120/162] avg loss 0.0084055, throughput 9.37768K wps
[Epoch 7 Batch 150/162] avg loss 0.00862719, throughput 9.34004K wps
Begin Testing...
[Epoch 7] train avg loss 0.00857501, dev acc 0.8800, dev avg loss 0.399365, throughput 9.43413K wps
Observed Improvement.
Begin Testing...
[Epoch 8 Batch 30/162] avg loss 0.00818365, throughput 9.48046K wps
[Epoch 8 Batch 60/162] avg loss 0.0084371, throughput 9.53648K wps
[Epoch 8 Batch 90/162] avg loss 0.00791457, throughput 9.50342K wps
[Epoch 8 Batch 120/162] avg loss 0.0080469, throughput 9.53423K wps
[Epoch 8 Batch 150/162] avg loss 0.00772757, throughput 9.29782K wps
Begin Testing...
[Epoch 8] train avg loss 0.00804565, dev acc 0.8800, dev avg loss 0.37406, throughput 9.4605K wps
Observed Improvement.
Begin Testing...
[Epoch 9 Batch 30/162] avg loss 0.00783388, throughput 9.56713K wps
[Epoch 9 Batch 60/162] avg loss 0.00750875, throughput 9.35717K wps
[Epoch 9 Batch 90/162] avg loss 0.00736492, throughput 9.44289K wps
[Epoch 9 Batch 120/162] avg loss 0.00754926, throughput 9.23657K wps
[Epoch 9 Batch 150/162] avg loss 0.00744379, throughput 9.4436K wps
Begin Testing...
[Epoch 9] train avg loss 0.00755149, dev acc 0.8800, dev avg loss 0.354962, throughput 9.40242K wps
Observed Improvement.
Begin Testing...
[Epoch 10 Batch 30/162] avg loss 0.00755768, throughput 9.68517K wps
[Epoch 10 Batch 60/162] avg loss 0.00720936, throughput 9.39928K wps
[Epoch 10 Batch 90/162] avg loss 0.00738943, throughput 9.29113K wps
[Epoch 10 Batch 120/162] avg loss 0.00691371, throughput 9.5131K wps
[Epoch 10 Batch 150/162] avg loss 0.00688952, throughput 9.49598K wps
Begin Testing...
[Epoch 10] train avg loss 0.00715782, dev acc 0.8789, dev avg loss 0.339258, throughput 9.48529K wps
[Epoch 11 Batch 30/162] avg loss 0.00726859, throughput 9.45778K wps
[Epoch 11 Batch 60/162] avg loss 0.00685523, throughput 9.50418K wps
[Epoch 11 Batch 90/162] avg loss 0.00709674, throughput 9.34463K wps
[Epoch 11 Batch 120/162] avg loss 0.00683972, throughput 9.42154K wps
[Epoch 11 Batch 150/162] avg loss 0.00667701, throughput 9.38437K wps
Begin Testing...
[Epoch 11] train avg loss 0.00694511, dev acc 0.8856, dev avg loss 0.325453, throughput 9.42915K wps
Observed Improvement.
Begin Testing...
[Epoch 12 Batch 30/162] avg loss 0.00658511, throughput 9.53781K wps
[Epoch 12 Batch 60/162] avg loss 0.0066232, throughput 9.45031K wps
[Epoch 12 Batch 90/162] avg loss 0.00636032, throughput 9.52814K wps
[Epoch 12 Batch 120/162] avg loss 0.00656824, throughput 9.54224K wps
[Epoch 12 Batch 150/162] avg loss 0.00687836, throughput 9.51515K wps
Begin Testing...
[Epoch 12] train avg loss 0.0066063, dev acc 0.8833, dev avg loss 0.318474, throughput 9.51136K wps
[Epoch 13 Batch 30/162] avg loss 0.00684829, throughput 9.63349K wps
[Epoch 13 Batch 60/162] avg loss 0.00619478, throughput 9.41457K wps
[Epoch 13 Batch 90/162] avg loss 0.00646056, throughput 9.24207K wps
[Epoch 13 Batch 120/162] avg loss 0.00596369, throughput 9.39488K wps
[Epoch 13 Batch 150/162] avg loss 0.00611978, throughput 9.27781K wps
Begin Testing...
[Epoch 13] train avg loss 0.00632501, dev acc 0.8900, dev avg loss 0.306087, throughput 9.37186K wps
Observed Improvement.
Begin Testing...
[Epoch 14 Batch 30/162] avg loss 0.00656697, throughput 9.60512K wps
[Epoch 14 Batch 60/162] avg loss 0.00582908, throughput 9.43815K wps
[Epoch 14 Batch 90/162] avg loss 0.0060612, throughput 9.27608K wps
[Epoch 14 Batch 120/162] avg loss 0.00637264, throughput 9.43701K wps
[Epoch 14 Batch 150/162] avg loss 0.00628012, throughput 9.34968K wps
Begin Testing...
[Epoch 14] train avg loss 0.00619083, dev acc 0.8911, dev avg loss 0.299656, throughput 9.40719K wps
Observed Improvement.
Begin Testing...
[Epoch 15 Batch 30/162] avg loss 0.00612801, throughput 9.38339K wps
[Epoch 15 Batch 60/162] avg loss 0.00602083, throughput 9.57949K wps
[Epoch 15 Batch 90/162] avg loss 0.00592529, throughput 9.24781K wps
[Epoch 15 Batch 120/162] avg loss 0.00599034, throughput 9.34851K wps
[Epoch 15 Batch 150/162] avg loss 0.00601609, throughput 9.28601K wps
Begin Testing...
[Epoch 15] train avg loss 0.00602134, dev acc 0.8911, dev avg loss 0.293683, throughput 9.36178K wps
Observed Improvement.
Begin Testing...
[Epoch 16 Batch 30/162] avg loss 0.00586356, throughput 9.75246K wps
[Epoch 16 Batch 60/162] avg loss 0.00596519, throughput 9.21761K wps
[Epoch 16 Batch 90/162] avg loss 0.00600566, throughput 9.31146K wps
[Epoch 16 Batch 120/162] avg loss 0.00570315, throughput 9.29259K wps
[Epoch 16 Batch 150/162] avg loss 0.00582301, throughput 9.49022K wps
Begin Testing...
[Epoch 16] train avg loss 0.00586086, dev acc 0.8922, dev avg loss 0.28805, throughput 9.42561K wps
Observed Improvement.
Begin Testing...
[Epoch 17 Batch 30/162] avg loss 0.00598131, throughput 9.75096K wps
[Epoch 17 Batch 60/162] avg loss 0.00602191, throughput 9.34629K wps
[Epoch 17 Batch 90/162] avg loss 0.00603747, throughput 9.43104K wps
[Epoch 17 Batch 120/162] avg loss 0.00561201, throughput 9.49718K wps
[Epoch 17 Batch 150/162] avg loss 0.00552843, throughput 9.65088K wps
Begin Testing...
[Epoch 17] train avg loss 0.00581316, dev acc 0.8944, dev avg loss 0.28126, throughput 9.53793K wps
Observed Improvement.
Begin Testing...
[Epoch 18 Batch 30/162] avg loss 0.00588131, throughput 9.60645K wps
[Epoch 18 Batch 60/162] avg loss 0.00527257, throughput 9.30314K wps
[Epoch 18 Batch 90/162] avg loss 0.00552946, throughput 9.60997K wps
[Epoch 18 Batch 120/162] avg loss 0.00574052, throughput 9.27769K wps
[Epoch 18 Batch 150/162] avg loss 0.00585075, throughput 9.39453K wps
Begin Testing...
[Epoch 18] train avg loss 0.0056514, dev acc 0.8989, dev avg loss 0.277289, throughput 9.43653K wps
Observed Improvement.
Begin Testing...
[Epoch 19 Batch 30/162] avg loss 0.00527944, throughput 9.69935K wps
[Epoch 19 Batch 60/162] avg loss 0.00558923, throughput 9.4842K wps
[Epoch 19 Batch 90/162] avg loss 0.00552476, throughput 9.51675K wps
[Epoch 19 Batch 120/162] avg loss 0.00555223, throughput 9.44938K wps
[Epoch 19 Batch 150/162] avg loss 0.00553408, throughput 9.37387K wps
Begin Testing...
[Epoch 19] train avg loss 0.00552031, dev acc 0.8989, dev avg loss 0.273317, throughput 9.50088K wps
Observed Improvement.
Begin Testing...
[Epoch 20 Batch 30/162] avg loss 0.00565588, throughput 9.80042K wps
[Epoch 20 Batch 60/162] avg loss 0.00561268, throughput 9.46343K wps
[Epoch 20 Batch 90/162] avg loss 0.00540894, throughput 9.56078K wps
[Epoch 20 Batch 120/162] avg loss 0.00491343, throughput 9.33165K wps
[Epoch 20 Batch 150/162] avg loss 0.00573256, throughput 9.31314K wps
Begin Testing...
[Epoch 20] train avg loss 0.00544168, dev acc 0.8989, dev avg loss 0.268727, throughput 9.49842K wps
Observed Improvement.
Begin Testing...
[Epoch 21 Batch 30/162] avg loss 0.00522278, throughput 9.42495K wps
[Epoch 21 Batch 60/162] avg loss 0.00556865, throughput 9.26271K wps
[Epoch 21 Batch 90/162] avg loss 0.0053489, throughput 9.41956K wps
[Epoch 21 Batch 120/162] avg loss 0.00506265, throughput 9.24721K wps
[Epoch 21 Batch 150/162] avg loss 0.00526247, throughput 9.4924K wps
Begin Testing...
[Epoch 21] train avg loss 0.00527572, dev acc 0.9011, dev avg loss 0.265287, throughput 9.36924K wps
Observed Improvement.
Begin Testing...
[Epoch 22 Batch 30/162] avg loss 0.00514974, throughput 9.57517K wps
[Epoch 22 Batch 60/162] avg loss 0.00526714, throughput 9.34412K wps
[Epoch 22 Batch 90/162] avg loss 0.00502675, throughput 9.38803K wps
[Epoch 22 Batch 120/162] avg loss 0.00562323, throughput 9.35696K wps
[Epoch 22 Batch 150/162] avg loss 0.00498877, throughput 9.45085K wps
Begin Testing...
[Epoch 22] train avg loss 0.00518539, dev acc 0.9000, dev avg loss 0.262579, throughput 9.40829K wps
[Epoch 23 Batch 30/162] avg loss 0.00528424, throughput 9.81691K wps
[Epoch 23 Batch 60/162] avg loss 0.00513692, throughput 9.41316K wps
[Epoch 23 Batch 90/162] avg loss 0.00506792, throughput 9.66352K wps
[Epoch 23 Batch 120/162] avg loss 0.00521759, throughput 9.44756K wps
[Epoch 23 Batch 150/162] avg loss 0.00480998, throughput 9.48172K wps
Begin Testing...
[Epoch 23] train avg loss 0.00507666, dev acc 0.8978, dev avg loss 0.260427, throughput 9.56841K wps
[Epoch 24 Batch 30/162] avg loss 0.00512343, throughput 9.52201K wps
[Epoch 24 Batch 60/162] avg loss 0.00499039, throughput 9.59068K wps
[Epoch 24 Batch 90/162] avg loss 0.00525063, throughput 9.35033K wps
[Epoch 24 Batch 120/162] avg loss 0.00492358, throughput 9.49007K wps
[Epoch 24 Batch 150/162] avg loss 0.00490575, throughput 9.3638K wps
Begin Testing...
[Epoch 24] train avg loss 0.00498707, dev acc 0.8978, dev avg loss 0.257251, throughput 9.46279K wps
[Epoch 25 Batch 30/162] avg loss 0.00473282, throughput 9.69902K wps
[Epoch 25 Batch 60/162] avg loss 0.00479897, throughput 9.47792K wps
[Epoch 25 Batch 90/162] avg loss 0.00473241, throughput 9.53356K wps
[Epoch 25 Batch 120/162] avg loss 0.00502588, throughput 9.39198K wps
[Epoch 25 Batch 150/162] avg loss 0.0050131, throughput 9.45579K wps
Begin Testing...
[Epoch 25] train avg loss 0.00482421, dev acc 0.9011, dev avg loss 0.253929, throughput 9.48648K wps
Observed Improvement.
Begin Testing...
[Epoch 26 Batch 30/162] avg loss 0.00458386, throughput 9.54865K wps
[Epoch 26 Batch 60/162] avg loss 0.00476199, throughput 9.437K wps
[Epoch 26 Batch 90/162] avg loss 0.00462205, throughput 9.4687K wps
[Epoch 26 Batch 120/162] avg loss 0.00485405, throughput 9.33453K wps
[Epoch 26 Batch 150/162] avg loss 0.00506463, throughput 9.30712K wps
Begin Testing...
[Epoch 26] train avg loss 0.00475934, dev acc 0.9011, dev avg loss 0.250786, throughput 9.4315K wps
Observed Improvement.
Begin Testing...
[Epoch 27 Batch 30/162] avg loss 0.00495455, throughput 9.66012K wps
[Epoch 27 Batch 60/162] avg loss 0.00459179, throughput 9.12703K wps
[Epoch 27 Batch 90/162] avg loss 0.00442528, throughput 9.34731K wps
[Epoch 27 Batch 120/162] avg loss 0.00470762, throughput 9.51354K wps
[Epoch 27 Batch 150/162] avg loss 0.00491699, throughput 9.29035K wps
Begin Testing...
[Epoch 27] train avg loss 0.00472948, dev acc 0.9011, dev avg loss 0.24966, throughput 9.38206K wps
Observed Improvement.
Begin Testing...
[Epoch 28 Batch 30/162] avg loss 0.00450044, throughput 9.84596K wps
[Epoch 28 Batch 60/162] avg loss 0.00435868, throughput 9.35097K wps
[Epoch 28 Batch 90/162] avg loss 0.00499091, throughput 9.55115K wps
[Epoch 28 Batch 120/162] avg loss 0.0045662, throughput 9.40835K wps
[Epoch 28 Batch 150/162] avg loss 0.0046634, throughput 9.46574K wps
Begin Testing...
[Epoch 28] train avg loss 0.00461064, dev acc 0.9033, dev avg loss 0.246396, throughput 9.50198K wps
Observed Improvement.
Begin Testing...
[Epoch 29 Batch 30/162] avg loss 0.00498104, throughput 9.51413K wps
[Epoch 29 Batch 60/162] avg loss 0.0045087, throughput 9.34457K wps
[Epoch 29 Batch 90/162] avg loss 0.00428908, throughput 9.37145K wps
[Epoch 29 Batch 120/162] avg loss 0.00451376, throughput 9.25816K wps
[Epoch 29 Batch 150/162] avg loss 0.00447216, throughput 9.25276K wps
Begin Testing...
[Epoch 29] train avg loss 0.00453093, dev acc 0.9033, dev avg loss 0.244233, throughput 9.35076K wps
Observed Improvement.
Begin Testing...
[Epoch 30 Batch 30/162] avg loss 0.00461998, throughput 9.57098K wps
[Epoch 30 Batch 60/162] avg loss 0.0046053, throughput 9.24077K wps
[Epoch 30 Batch 90/162] avg loss 0.0047366, throughput 9.34549K wps
[Epoch 30 Batch 120/162] avg loss 0.00435281, throughput 9.36108K wps
[Epoch 30 Batch 150/162] avg loss 0.00408087, throughput 9.35771K wps
Begin Testing...
[Epoch 30] train avg loss 0.0044957, dev acc 0.9056, dev avg loss 0.242283, throughput 9.39002K wps
Observed Improvement.
Begin Testing...
[Epoch 31 Batch 30/162] avg loss 0.00447294, throughput 9.59893K wps
[Epoch 31 Batch 60/162] avg loss 0.00420574, throughput 9.40163K wps
[Epoch 31 Batch 90/162] avg loss 0.0044491, throughput 9.42201K wps
[Epoch 31 Batch 120/162] avg loss 0.00429182, throughput 9.50393K wps
[Epoch 31 Batch 150/162] avg loss 0.00466375, throughput 9.54949K wps
Begin Testing...
[Epoch 31] train avg loss 0.00442617, dev acc 0.9067, dev avg loss 0.240354, throughput 9.48952K wps
Observed Improvement.
Begin Testing...
[Epoch 32 Batch 30/162] avg loss 0.00446713, throughput 9.69587K wps
[Epoch 32 Batch 60/162] avg loss 0.00420414, throughput 9.40507K wps
[Epoch 32 Batch 90/162] avg loss 0.00466823, throughput 9.6539K wps
[Epoch 32 Batch 120/162] avg loss 0.00438461, throughput 9.33678K wps
[Epoch 32 Batch 150/162] avg loss 0.00393729, throughput 9.45048K wps
Begin Testing...
[Epoch 32] train avg loss 0.00434896, dev acc 0.9089, dev avg loss 0.238621, throughput 9.51696K wps
Observed Improvement.
Begin Testing...
[Epoch 33 Batch 30/162] avg loss 0.00418788, throughput 9.65906K wps
[Epoch 33 Batch 60/162] avg loss 0.0040762, throughput 9.44397K wps
[Epoch 33 Batch 90/162] avg loss 0.00440066, throughput 9.45069K wps
[Epoch 33 Batch 120/162] avg loss 0.00403927, throughput 9.51159K wps
[Epoch 33 Batch 150/162] avg loss 0.0043793, throughput 9.37179K wps
Begin Testing...
[Epoch 33] train avg loss 0.00422252, dev acc 0.9044, dev avg loss 0.239687, throughput 9.46715K wps
[Epoch 34 Batch 30/162] avg loss 0.00420516, throughput 9.6888K wps
[Epoch 34 Batch 60/162] avg loss 0.0042716, throughput 9.55776K wps
[Epoch 34 Batch 90/162] avg loss 0.00414765, throughput 9.33513K wps
[Epoch 34 Batch 120/162] avg loss 0.00415101, throughput 9.54013K wps
[Epoch 34 Batch 150/162] avg loss 0.00420036, throughput 9.50655K wps
Begin Testing...
[Epoch 34] train avg loss 0.00421206, dev acc 0.9100, dev avg loss 0.235111, throughput 9.5178K wps
Observed Improvement.
Begin Testing...
[Epoch 35 Batch 30/162] avg loss 0.00442798, throughput 9.59353K wps
[Epoch 35 Batch 60/162] avg loss 0.00388843, throughput 9.39334K wps
[Epoch 35 Batch 90/162] avg loss 0.00424558, throughput 9.52359K wps
[Epoch 35 Batch 120/162] avg loss 0.00390578, throughput 9.39936K wps
[Epoch 35 Batch 150/162] avg loss 0.0041836, throughput 9.46473K wps
Begin Testing...
[Epoch 35] train avg loss 0.00414385, dev acc 0.9111, dev avg loss 0.234628, throughput 9.48279K wps
Observed Improvement.
Begin Testing...
[Epoch 36 Batch 30/162] avg loss 0.00352437, throughput 9.55552K wps
[Epoch 36 Batch 60/162] avg loss 0.00405335, throughput 9.37121K wps
[Epoch 36 Batch 90/162] avg loss 0.00367173, throughput 9.35721K wps
[Epoch 36 Batch 120/162] avg loss 0.00399542, throughput 9.38733K wps
[Epoch 36 Batch 150/162] avg loss 0.00459285, throughput 9.62239K wps
Begin Testing...
[Epoch 36] train avg loss 0.00398581, dev acc 0.9089, dev avg loss 0.233608, throughput 9.46064K wps
[Epoch 37 Batch 30/162] avg loss 0.00373987, throughput 9.49691K wps
[Epoch 37 Batch 60/162] avg loss 0.00415808, throughput 9.38928K wps
[Epoch 37 Batch 90/162] avg loss 0.00406089, throughput 9.51604K wps
[Epoch 37 Batch 120/162] avg loss 0.00374069, throughput 9.47642K wps
[Epoch 37 Batch 150/162] avg loss 0.00396706, throughput 9.4073K wps
Begin Testing...
[Epoch 37] train avg loss 0.00395538, dev acc 0.9089, dev avg loss 0.23242, throughput 9.46242K wps
[Epoch 38 Batch 30/162] avg loss 0.00404117, throughput 9.5792K wps
[Epoch 38 Batch 60/162] avg loss 0.00381725, throughput 9.48858K wps
[Epoch 38 Batch 90/162] avg loss 0.00402549, throughput 9.30221K wps
[Epoch 38 Batch 120/162] avg loss 0.00387244, throughput 9.34296K wps
[Epoch 38 Batch 150/162] avg loss 0.00402056, throughput 9.43134K wps
Begin Testing...
[Epoch 38] train avg loss 0.00395622, dev acc 0.9133, dev avg loss 0.229041, throughput 9.4143K wps
Observed Improvement.
Begin Testing...
[Epoch 39 Batch 30/162] avg loss 0.00377506, throughput 9.71332K wps
[Epoch 39 Batch 60/162] avg loss 0.00368479, throughput 9.61253K wps
[Epoch 39 Batch 90/162] avg loss 0.00398747, throughput 9.34741K wps
[Epoch 39 Batch 120/162] avg loss 0.00386424, throughput 9.44706K wps
[Epoch 39 Batch 150/162] avg loss 0.00399345, throughput 9.43075K wps
Begin Testing...
[Epoch 39] train avg loss 0.00389146, dev acc 0.9100, dev avg loss 0.230367, throughput 9.50468K wps
[Epoch 40 Batch 30/162] avg loss 0.00399575, throughput 9.76397K wps
[Epoch 40 Batch 60/162] avg loss 0.00401105, throughput 9.50243K wps
[Epoch 40 Batch 90/162] avg loss 0.00378792, throughput 9.5264K wps
[Epoch 40 Batch 120/162] avg loss 0.0034886, throughput 9.32741K wps
[Epoch 40 Batch 150/162] avg loss 0.0034635, throughput 9.36866K wps
Begin Testing...
[Epoch 40] train avg loss 0.00376748, dev acc 0.9156, dev avg loss 0.226979, throughput 9.49957K wps
Observed Improvement.
Begin Testing...
[Epoch 41 Batch 30/162] avg loss 0.00367757, throughput 9.56744K wps
[Epoch 41 Batch 60/162] avg loss 0.00355687, throughput 9.44885K wps
[Epoch 41 Batch 90/162] avg loss 0.0035562, throughput 9.31885K wps
[Epoch 41 Batch 120/162] avg loss 0.00424681, throughput 9.39943K wps
[Epoch 41 Batch 150/162] avg loss 0.00365142, throughput 9.35604K wps
Begin Testing...
[Epoch 41] train avg loss 0.00378031, dev acc 0.9111, dev avg loss 0.226678, throughput 9.42645K wps
[Epoch 42 Batch 30/162] avg loss 0.00359866, throughput 9.76635K wps
[Epoch 42 Batch 60/162] avg loss 0.00374081, throughput 9.50995K wps
[Epoch 42 Batch 90/162] avg loss 0.00358194, throughput 9.29239K wps
[Epoch 42 Batch 120/162] avg loss 0.00378545, throughput 9.33937K wps
[Epoch 42 Batch 150/162] avg loss 0.00388271, throughput 9.24699K wps
Begin Testing...
[Epoch 42] train avg loss 0.00369635, dev acc 0.9144, dev avg loss 0.224951, throughput 9.42789K wps
[Epoch 43 Batch 30/162] avg loss 0.00375167, throughput 9.53517K wps
[Epoch 43 Batch 60/162] avg loss 0.00364869, throughput 9.34265K wps
[Epoch 43 Batch 90/162] avg loss 0.00330231, throughput 9.41817K wps
[Epoch 43 Batch 120/162] avg loss 0.00364492, throughput 9.31893K wps
[Epoch 43 Batch 150/162] avg loss 0.00352698, throughput 9.4981K wps
Begin Testing...
[Epoch 43] train avg loss 0.00361189, dev acc 0.9156, dev avg loss 0.224043, throughput 9.4339K wps
Observed Improvement.
Begin Testing...
[Epoch 44 Batch 30/162] avg loss 0.00342193, throughput 9.54262K wps
[Epoch 44 Batch 60/162] avg loss 0.00369784, throughput 9.44569K wps
[Epoch 44 Batch 90/162] avg loss 0.00352535, throughput 9.52391K wps
[Epoch 44 Batch 120/162] avg loss 0.0032371, throughput 9.29821K wps
[Epoch 44 Batch 150/162] avg loss 0.00349533, throughput 9.31206K wps
Begin Testing...
[Epoch 44] train avg loss 0.0035113, dev acc 0.9156, dev avg loss 0.223126, throughput 9.43164K wps
Observed Improvement.
Begin Testing...
[Epoch 45 Batch 30/162] avg loss 0.00345407, throughput 9.47783K wps
[Epoch 45 Batch 60/162] avg loss 0.00351533, throughput 9.36201K wps
[Epoch 45 Batch 90/162] avg loss 0.00339695, throughput 9.43404K wps
[Epoch 45 Batch 120/162] avg loss 0.00370953, throughput 9.34184K wps
[Epoch 45 Batch 150/162] avg loss 0.00330461, throughput 9.3358K wps
Begin Testing...
[Epoch 45] train avg loss 0.00348241, dev acc 0.9156, dev avg loss 0.221139, throughput 9.39452K wps
Observed Improvement.
Begin Testing...
[Epoch 46 Batch 30/162] avg loss 0.00369471, throughput 9.65403K wps
[Epoch 46 Batch 60/162] avg loss 0.00313452, throughput 9.33946K wps
[Epoch 46 Batch 90/162] avg loss 0.00377554, throughput 9.48598K wps
[Epoch 46 Batch 120/162] avg loss 0.00333299, throughput 9.3951K wps
[Epoch 46 Batch 150/162] avg loss 0.00329425, throughput 9.37143K wps
Begin Testing...
[Epoch 46] train avg loss 0.00343579, dev acc 0.9167, dev avg loss 0.221783, throughput 9.44185K wps
Observed Improvement.
Begin Testing...
[Epoch 47 Batch 30/162] avg loss 0.00321181, throughput 9.66545K wps
[Epoch 47 Batch 60/162] avg loss 0.00339087, throughput 9.43075K wps
[Epoch 47 Batch 90/162] avg loss 0.00361375, throughput 9.31213K wps
[Epoch 47 Batch 120/162] avg loss 0.00352803, throughput 9.44965K wps
[Epoch 47 Batch 150/162] avg loss 0.00356952, throughput 9.54141K wps
Begin Testing...
[Epoch 47] train avg loss 0.00342298, dev acc 0.9156, dev avg loss 0.219646, throughput 9.46548K wps
[Epoch 48 Batch 30/162] avg loss 0.00356556, throughput 9.43776K wps
[Epoch 48 Batch 60/162] avg loss 0.00339116, throughput 9.42725K wps
[Epoch 48 Batch 90/162] avg loss 0.00317187, throughput 9.30055K wps
[Epoch 48 Batch 120/162] avg loss 0.0034982, throughput 9.47272K wps
[Epoch 48 Batch 150/162] avg loss 0.00339109, throughput 9.50377K wps
Begin Testing...
[Epoch 48] train avg loss 0.00338265, dev acc 0.9156, dev avg loss 0.218502, throughput 9.44167K wps
[Epoch 49 Batch 30/162] avg loss 0.00330427, throughput 9.61075K wps
[Epoch 49 Batch 60/162] avg loss 0.00297916, throughput 9.48134K wps
[Epoch 49 Batch 90/162] avg loss 0.00349719, throughput 9.2584K wps
[Epoch 49 Batch 120/162] avg loss 0.00319061, throughput 9.51498K wps
[Epoch 49 Batch 150/162] avg loss 0.00334765, throughput 9.36327K wps
Begin Testing...
[Epoch 49] train avg loss 0.00328629, dev acc 0.9156, dev avg loss 0.217767, throughput 9.45674K wps
[Epoch 50 Batch 30/162] avg loss 0.00339999, throughput 9.72961K wps
[Epoch 50 Batch 60/162] avg loss 0.00314513, throughput 9.46687K wps
[Epoch 50 Batch 90/162] avg loss 0.00316926, throughput 9.5752K wps
[Epoch 50 Batch 120/162] avg loss 0.00331684, throughput 9.36425K wps
[Epoch 50 Batch 150/162] avg loss 0.00322733, throughput 9.39523K wps
Begin Testing...
[Epoch 50] train avg loss 0.00322195, dev acc 0.9211, dev avg loss 0.219569, throughput 9.48555K wps
Observed Improvement.
Begin Testing...
[Epoch 51 Batch 30/162] avg loss 0.00304803, throughput 9.73883K wps
[Epoch 51 Batch 60/162] avg loss 0.00321169, throughput 9.46602K wps
[Epoch 51 Batch 90/162] avg loss 0.00317829, throughput 9.42211K wps
[Epoch 51 Batch 120/162] avg loss 0.00316004, throughput 9.25269K wps
[Epoch 51 Batch 150/162] avg loss 0.00338718, throughput 9.58267K wps
Begin Testing...
[Epoch 51] train avg loss 0.0031882, dev acc 0.9144, dev avg loss 0.216296, throughput 9.49293K wps
[Epoch 52 Batch 30/162] avg loss 0.00333176, throughput 9.67479K wps
[Epoch 52 Batch 60/162] avg loss 0.00318557, throughput 9.41824K wps
[Epoch 52 Batch 90/162] avg loss 0.00339549, throughput 9.49176K wps
[Epoch 52 Batch 120/162] avg loss 0.00312968, throughput 9.55219K wps
[Epoch 52 Batch 150/162] avg loss 0.0032367, throughput 9.46778K wps
Begin Testing...
[Epoch 52] train avg loss 0.00325888, dev acc 0.9167, dev avg loss 0.216058, throughput 9.51604K wps
[Epoch 53 Batch 30/162] avg loss 0.00296955, throughput 9.61285K wps
[Epoch 53 Batch 60/162] avg loss 0.00324922, throughput 9.51853K wps
[Epoch 53 Batch 90/162] avg loss 0.00332317, throughput 9.51737K wps
[Epoch 53 Batch 120/162] avg loss 0.00331453, throughput 9.45092K wps
[Epoch 53 Batch 150/162] avg loss 0.00283979, throughput 9.47417K wps
Begin Testing...
[Epoch 53] train avg loss 0.00314601, dev acc 0.9156, dev avg loss 0.215343, throughput 9.51561K wps
[Epoch 54 Batch 30/162] avg loss 0.00283139, throughput 9.58011K wps
[Epoch 54 Batch 60/162] avg loss 0.00296102, throughput 9.52525K wps
[Epoch 54 Batch 90/162] avg loss 0.00295842, throughput 9.37901K wps
[Epoch 54 Batch 120/162] avg loss 0.00333322, throughput 9.54805K wps
[Epoch 54 Batch 150/162] avg loss 0.0032159, throughput 9.58588K wps
Begin Testing...
[Epoch 54] train avg loss 0.00306505, dev acc 0.9189, dev avg loss 0.21512, throughput 9.50151K wps
[Epoch 55 Batch 30/162] avg loss 0.0028996, throughput 9.56908K wps
[Epoch 55 Batch 60/162] avg loss 0.00316301, throughput 9.36801K wps
[Epoch 55 Batch 90/162] avg loss 0.00310384, throughput 9.24167K wps
[Epoch 55 Batch 120/162] avg loss 0.00293129, throughput 9.29804K wps
[Epoch 55 Batch 150/162] avg loss 0.00289918, throughput 9.48217K wps
Begin Testing...
[Epoch 55] train avg loss 0.00300926, dev acc 0.9222, dev avg loss 0.217029, throughput 9.40464K wps
Observed Improvement.
Begin Testing...
[Epoch 56 Batch 30/162] avg loss 0.00307283, throughput 9.5382K wps
[Epoch 56 Batch 60/162] avg loss 0.00292158, throughput 9.37002K wps
[Epoch 56 Batch 90/162] avg loss 0.00305069, throughput 9.35315K wps
[Epoch 56 Batch 120/162] avg loss 0.00305629, throughput 9.28387K wps
[Epoch 56 Batch 150/162] avg loss 0.00287286, throughput 9.4076K wps
Begin Testing...
[Epoch 56] train avg loss 0.00297856, dev acc 0.9167, dev avg loss 0.213407, throughput 9.39759K wps
[Epoch 57 Batch 30/162] avg loss 0.00266812, throughput 9.6518K wps
[Epoch 57 Batch 60/162] avg loss 0.00264816, throughput 9.27253K wps
[Epoch 57 Batch 90/162] avg loss 0.00293006, throughput 9.48812K wps
[Epoch 57 Batch 120/162] avg loss 0.00299864, throughput 9.30066K wps
[Epoch 57 Batch 150/162] avg loss 0.00305861, throughput 9.34774K wps
Begin Testing...
[Epoch 57] train avg loss 0.00287291, dev acc 0.9144, dev avg loss 0.212358, throughput 9.39697K wps
[Epoch 58 Batch 30/162] avg loss 0.00282065, throughput 9.45591K wps
[Epoch 58 Batch 60/162] avg loss 0.00283707, throughput 9.53757K wps
[Epoch 58 Batch 90/162] avg loss 0.00319937, throughput 9.28189K wps
[Epoch 58 Batch 120/162] avg loss 0.0029303, throughput 9.51511K wps
[Epoch 58 Batch 150/162] avg loss 0.00276198, throughput 9.58722K wps
Begin Testing...
[Epoch 58] train avg loss 0.00290776, dev acc 0.9189, dev avg loss 0.213334, throughput 9.47794K wps
[Epoch 59 Batch 30/162] avg loss 0.00297311, throughput 9.54046K wps
[Epoch 59 Batch 60/162] avg loss 0.00279115, throughput 9.33268K wps
[Epoch 59 Batch 90/162] avg loss 0.00277007, throughput 9.39301K wps
[Epoch 59 Batch 120/162] avg loss 0.0028562, throughput 9.4156K wps
[Epoch 59 Batch 150/162] avg loss 0.00262711, throughput 9.39935K wps
Begin Testing...
[Epoch 59] train avg loss 0.00282039, dev acc 0.9189, dev avg loss 0.2114, throughput 9.41974K wps
[Epoch 60 Batch 30/162] avg loss 0.00279512, throughput 9.67375K wps
[Epoch 60 Batch 60/162] avg loss 0.00273873, throughput 9.42438K wps
[Epoch 60 Batch 90/162] avg loss 0.0030028, throughput 9.44557K wps
[Epoch 60 Batch 120/162] avg loss 0.00288723, throughput 9.331K wps
[Epoch 60 Batch 150/162] avg loss 0.00271712, throughput 9.32526K wps
Begin Testing...
[Epoch 60] train avg loss 0.00281298, dev acc 0.9200, dev avg loss 0.211309, throughput 9.42634K wps
[Epoch 61 Batch 30/162] avg loss 0.00294816, throughput 9.51429K wps
[Epoch 61 Batch 60/162] avg loss 0.00284839, throughput 9.40179K wps
[Epoch 61 Batch 90/162] avg loss 0.00299768, throughput 9.36694K wps
[Epoch 61 Batch 120/162] avg loss 0.00261078, throughput 9.60789K wps
[Epoch 61 Batch 150/162] avg loss 0.00270688, throughput 9.38015K wps
Begin Testing...
[Epoch 61] train avg loss 0.00279527, dev acc 0.9200, dev avg loss 0.21087, throughput 9.45402K wps
[Epoch 62 Batch 30/162] avg loss 0.00263236, throughput 9.58715K wps
[Epoch 62 Batch 60/162] avg loss 0.00296344, throughput 9.40692K wps
[Epoch 62 Batch 90/162] avg loss 0.00259809, throughput 9.33439K wps
[Epoch 62 Batch 120/162] avg loss 0.00295865, throughput 9.4177K wps
[Epoch 62 Batch 150/162] avg loss 0.00252636, throughput 9.42066K wps
Begin Testing...
[Epoch 62] train avg loss 0.00272715, dev acc 0.9244, dev avg loss 0.21214, throughput 9.42957K wps
Observed Improvement.
Begin Testing...
[Epoch 63 Batch 30/162] avg loss 0.00259946, throughput 9.62408K wps
[Epoch 63 Batch 60/162] avg loss 0.00281981, throughput 9.31697K wps
[Epoch 63 Batch 90/162] avg loss 0.00243692, throughput 9.41388K wps
[Epoch 63 Batch 120/162] avg loss 0.00285183, throughput 9.32123K wps
[Epoch 63 Batch 150/162] avg loss 0.00277472, throughput 9.48243K wps
Begin Testing...
[Epoch 63] train avg loss 0.00269007, dev acc 0.9200, dev avg loss 0.209419, throughput 9.45066K wps
[Epoch 64 Batch 30/162] avg loss 0.0026411, throughput 9.55086K wps
[Epoch 64 Batch 60/162] avg loss 0.00278089, throughput 9.31499K wps
[Epoch 64 Batch 90/162] avg loss 0.00272726, throughput 9.57048K wps
[Epoch 64 Batch 120/162] avg loss 0.00259458, throughput 9.54364K wps
[Epoch 64 Batch 150/162] avg loss 0.00259671, throughput 9.55141K wps
Begin Testing...
[Epoch 64] train avg loss 0.00264944, dev acc 0.9178, dev avg loss 0.208731, throughput 9.49947K wps
[Epoch 65 Batch 30/162] avg loss 0.00268928, throughput 9.56334K wps
[Epoch 65 Batch 60/162] avg loss 0.0024081, throughput 9.2991K wps
[Epoch 65 Batch 90/162] avg loss 0.00253225, throughput 9.40404K wps
[Epoch 65 Batch 120/162] avg loss 0.00286534, throughput 9.31891K wps
[Epoch 65 Batch 150/162] avg loss 0.00263743, throughput 9.39043K wps
Begin Testing...
[Epoch 65] train avg loss 0.00262263, dev acc 0.9222, dev avg loss 0.217683, throughput 9.38169K wps
[Epoch 66 Batch 30/162] avg loss 0.00273375, throughput 9.54129K wps
[Epoch 66 Batch 60/162] avg loss 0.00250165, throughput 9.28194K wps
[Epoch 66 Batch 90/162] avg loss 0.00227271, throughput 9.24945K wps
[Epoch 66 Batch 120/162] avg loss 0.00250611, throughput 9.58127K wps
[Epoch 66 Batch 150/162] avg loss 0.00263276, throughput 9.32298K wps
Begin Testing...
[Epoch 66] train avg loss 0.00252711, dev acc 0.9211, dev avg loss 0.209025, throughput 9.41482K wps
[Epoch 67 Batch 30/162] avg loss 0.0023783, throughput 9.44398K wps
[Epoch 67 Batch 60/162] avg loss 0.00264083, throughput 9.41664K wps
[Epoch 67 Batch 90/162] avg loss 0.00270168, throughput 9.43456K wps
[Epoch 67 Batch 120/162] avg loss 0.00255234, throughput 9.53434K wps
[Epoch 67 Batch 150/162] avg loss 0.00248975, throughput 9.40287K wps
Begin Testing...
[Epoch 67] train avg loss 0.00257764, dev acc 0.9222, dev avg loss 0.20864, throughput 9.44695K wps
[Epoch 68 Batch 30/162] avg loss 0.00248602, throughput 9.60105K wps
[Epoch 68 Batch 60/162] avg loss 0.00241206, throughput 9.37498K wps
[Epoch 68 Batch 90/162] avg loss 0.00261767, throughput 9.53763K wps
[Epoch 68 Batch 120/162] avg loss 0.00252987, throughput 9.59952K wps
[Epoch 68 Batch 150/162] avg loss 0.00263196, throughput 9.44156K wps
Begin Testing...
[Epoch 68] train avg loss 0.00252222, dev acc 0.9222, dev avg loss 0.209025, throughput 9.49215K wps
[Epoch 69 Batch 30/162] avg loss 0.00257744, throughput 9.51185K wps
[Epoch 69 Batch 60/162] avg loss 0.00240963, throughput 9.35546K wps
[Epoch 69 Batch 90/162] avg loss 0.00225741, throughput 9.273K wps
[Epoch 69 Batch 120/162] avg loss 0.00248136, throughput 9.45201K wps
[Epoch 69 Batch 150/162] avg loss 0.00263835, throughput 9.29093K wps
Begin Testing...
[Epoch 69] train avg loss 0.00246656, dev acc 0.9200, dev avg loss 0.207501, throughput 9.39072K wps
[Epoch 70 Batch 30/162] avg loss 0.00257716, throughput 9.79108K wps
[Epoch 70 Batch 60/162] avg loss 0.00238622, throughput 9.54606K wps
[Epoch 70 Batch 90/162] avg loss 0.00241528, throughput 9.37009K wps
[Epoch 70 Batch 120/162] avg loss 0.00237742, throughput 9.52574K wps
[Epoch 70 Batch 150/162] avg loss 0.00235501, throughput 9.40579K wps
Begin Testing...
[Epoch 70] train avg loss 0.00241617, dev acc 0.9244, dev avg loss 0.208104, throughput 9.50389K wps
Observed Improvement.
Begin Testing...
[Epoch 71 Batch 30/162] avg loss 0.00256968, throughput 9.69304K wps
[Epoch 71 Batch 60/162] avg loss 0.002199, throughput 9.22077K wps
[Epoch 71 Batch 90/162] avg loss 0.0025853, throughput 9.34078K wps
[Epoch 71 Batch 120/162] avg loss 0.00241332, throughput 9.39305K wps
[Epoch 71 Batch 150/162] avg loss 0.00231831, throughput 9.41703K wps
Begin Testing...
[Epoch 71] train avg loss 0.00242621, dev acc 0.9244, dev avg loss 0.209054, throughput 9.4221K wps
Observed Improvement.
Begin Testing...
[Epoch 72 Batch 30/162] avg loss 0.00218746, throughput 9.83961K wps
[Epoch 72 Batch 60/162] avg loss 0.00236692, throughput 9.62205K wps
[Epoch 72 Batch 90/162] avg loss 0.00235433, throughput 9.55524K wps
[Epoch 72 Batch 120/162] avg loss 0.00251754, throughput 9.46943K wps
[Epoch 72 Batch 150/162] avg loss 0.00251537, throughput 9.49556K wps
Begin Testing...
[Epoch 72] train avg loss 0.00240595, dev acc 0.9256, dev avg loss 0.207247, throughput 9.59049K wps
Observed Improvement.
Begin Testing...
[Epoch 73 Batch 30/162] avg loss 0.00229047, throughput 9.84596K wps
[Epoch 73 Batch 60/162] avg loss 0.00212348, throughput 9.59003K wps
[Epoch 73 Batch 90/162] avg loss 0.00252019, throughput 9.3789K wps
[Epoch 73 Batch 120/162] avg loss 0.00243575, throughput 9.42044K wps
[Epoch 73 Batch 150/162] avg loss 0.00224833, throughput 9.35089K wps
Begin Testing...
[Epoch 73] train avg loss 0.00232296, dev acc 0.9244, dev avg loss 0.207334, throughput 9.51405K wps
[Epoch 74 Batch 30/162] avg loss 0.00208926, throughput 9.84272K wps
[Epoch 74 Batch 60/162] avg loss 0.00219255, throughput 9.33242K wps
[Epoch 74 Batch 90/162] avg loss 0.00223768, throughput 9.41785K wps
[Epoch 74 Batch 120/162] avg loss 0.00210484, throughput 9.50019K wps
[Epoch 74 Batch 150/162] avg loss 0.00256881, throughput 9.46703K wps
Begin Testing...
[Epoch 74] train avg loss 0.00225917, dev acc 0.9256, dev avg loss 0.207519, throughput 9.50478K wps
Observed Improvement.
Begin Testing...
[Epoch 75 Batch 30/162] avg loss 0.00227324, throughput 9.51474K wps
[Epoch 75 Batch 60/162] avg loss 0.00210486, throughput 9.31085K wps
[Epoch 75 Batch 90/162] avg loss 0.00203041, throughput 9.49047K wps
[Epoch 75 Batch 120/162] avg loss 0.00232083, throughput 9.29421K wps
[Epoch 75 Batch 150/162] avg loss 0.00242434, throughput 9.48995K wps
Begin Testing...
[Epoch 75] train avg loss 0.00221609, dev acc 0.9244, dev avg loss 0.205696, throughput 9.39859K wps
[Epoch 76 Batch 30/162] avg loss 0.002584, throughput 9.63133K wps
[Epoch 76 Batch 60/162] avg loss 0.00223434, throughput 9.55934K wps
[Epoch 76 Batch 90/162] avg loss 0.00200503, throughput 9.37998K wps
[Epoch 76 Batch 120/162] avg loss 0.0022061, throughput 9.43218K wps
[Epoch 76 Batch 150/162] avg loss 0.00220366, throughput 9.22204K wps
Begin Testing...
[Epoch 76] train avg loss 0.00224055, dev acc 0.9278, dev avg loss 0.210187, throughput 9.43402K wps
Observed Improvement.
Begin Testing...
[Epoch 77 Batch 30/162] avg loss 0.00215917, throughput 9.52968K wps
[Epoch 77 Batch 60/162] avg loss 0.00245064, throughput 9.47393K wps
[Epoch 77 Batch 90/162] avg loss 0.00216222, throughput 9.27349K wps
[Epoch 77 Batch 120/162] avg loss 0.0020695, throughput 9.51123K wps
[Epoch 77 Batch 150/162] avg loss 0.00225657, throughput 9.32636K wps
Begin Testing...
[Epoch 77] train avg loss 0.00221045, dev acc 0.9256, dev avg loss 0.208501, throughput 9.43248K wps
[Epoch 78 Batch 30/162] avg loss 0.00218239, throughput 9.65287K wps
[Epoch 78 Batch 60/162] avg loss 0.00211587, throughput 9.45328K wps
[Epoch 78 Batch 90/162] avg loss 0.00207139, throughput 9.3041K wps
[Epoch 78 Batch 120/162] avg loss 0.00223113, throughput 9.47803K wps
[Epoch 78 Batch 150/162] avg loss 0.00228386, throughput 9.49644K wps
Begin Testing...
[Epoch 78] train avg loss 0.00215385, dev acc 0.9267, dev avg loss 0.206335, throughput 9.45587K wps
[Epoch 79 Batch 30/162] avg loss 0.00200604, throughput 9.69346K wps
[Epoch 79 Batch 60/162] avg loss 0.00207422, throughput 9.55447K wps
[Epoch 79 Batch 90/162] avg loss 0.00220029, throughput 9.23361K wps
[Epoch 79 Batch 120/162] avg loss 0.00220582, throughput 9.40673K wps
[Epoch 79 Batch 150/162] avg loss 0.0020605, throughput 9.30952K wps
Begin Testing...
[Epoch 79] train avg loss 0.00210708, dev acc 0.9267, dev avg loss 0.208375, throughput 9.45028K wps
[Epoch 80 Batch 30/162] avg loss 0.00182689, throughput 9.60553K wps
[Epoch 80 Batch 60/162] avg loss 0.00223671, throughput 9.42054K wps
[Epoch 80 Batch 90/162] avg loss 0.00214493, throughput 9.36083K wps
[Epoch 80 Batch 120/162] avg loss 0.00232897, throughput 9.56799K wps
[Epoch 80 Batch 150/162] avg loss 0.00217551, throughput 9.391K wps
Begin Testing...
[Epoch 80] train avg loss 0.00216827, dev acc 0.9267, dev avg loss 0.207031, throughput 9.47066K wps
[Epoch 81 Batch 30/162] avg loss 0.00224958, throughput 9.67819K wps
[Epoch 81 Batch 60/162] avg loss 0.00209998, throughput 9.50622K wps
[Epoch 81 Batch 90/162] avg loss 0.00213501, throughput 9.56551K wps
[Epoch 81 Batch 120/162] avg loss 0.00190779, throughput 9.5289K wps
[Epoch 81 Batch 150/162] avg loss 0.00217183, throughput 9.35495K wps
Begin Testing...
[Epoch 81] train avg loss 0.00210358, dev acc 0.9233, dev avg loss 0.20474, throughput 9.51279K wps
[Epoch 82 Batch 30/162] avg loss 0.00198056, throughput 9.70332K wps
[Epoch 82 Batch 60/162] avg loss 0.00216845, throughput 9.47577K wps
[Epoch 82 Batch 90/162] avg loss 0.00206459, throughput 9.35346K wps
[Epoch 82 Batch 120/162] avg loss 0.00208532, throughput 9.41614K wps
[Epoch 82 Batch 150/162] avg loss 0.00213291, throughput 9.45651K wps
Begin Testing...
[Epoch 82] train avg loss 0.00206858, dev acc 0.9211, dev avg loss 0.203971, throughput 9.46436K wps
[Epoch 83 Batch 30/162] avg loss 0.00191626, throughput 9.57696K wps
[Epoch 83 Batch 60/162] avg loss 0.00181627, throughput 9.46458K wps
[Epoch 83 Batch 90/162] avg loss 0.00198248, throughput 9.30247K wps
[Epoch 83 Batch 120/162] avg loss 0.00206587, throughput 9.47617K wps
[Epoch 83 Batch 150/162] avg loss 0.00216394, throughput 9.27024K wps
Begin Testing...
[Epoch 83] train avg loss 0.00198107, dev acc 0.9244, dev avg loss 0.204591, throughput 9.39822K wps
[Epoch 84 Batch 30/162] avg loss 0.00192526, throughput 9.62576K wps
[Epoch 84 Batch 60/162] avg loss 0.00205489, throughput 9.34862K wps
[Epoch 84 Batch 90/162] avg loss 0.00196269, throughput 9.65752K wps
[Epoch 84 Batch 120/162] avg loss 0.00215045, throughput 9.51851K wps
[Epoch 84 Batch 150/162] avg loss 0.00201985, throughput 9.38963K wps
Begin Testing...
[Epoch 84] train avg loss 0.00201562, dev acc 0.9233, dev avg loss 0.204243, throughput 9.49793K wps
[Epoch 85 Batch 30/162] avg loss 0.00208408, throughput 9.47403K wps
[Epoch 85 Batch 60/162] avg loss 0.00199362, throughput 9.40549K wps
[Epoch 85 Batch 90/162] avg loss 0.00181464, throughput 9.34391K wps
[Epoch 85 Batch 120/162] avg loss 0.00192487, throughput 9.47865K wps
[Epoch 85 Batch 150/162] avg loss 0.00203981, throughput 9.2984K wps
Begin Testing...
[Epoch 85] train avg loss 0.00196921, dev acc 0.9244, dev avg loss 0.204419, throughput 9.3935K wps
[Epoch 86 Batch 30/162] avg loss 0.00211766, throughput 9.48088K wps
[Epoch 86 Batch 60/162] avg loss 0.00211634, throughput 9.33728K wps
[Epoch 86 Batch 90/162] avg loss 0.00176884, throughput 9.4849K wps
[Epoch 86 Batch 120/162] avg loss 0.00195765, throughput 9.4098K wps
[Epoch 86 Batch 150/162] avg loss 0.00189764, throughput 9.43705K wps
Begin Testing...
[Epoch 86] train avg loss 0.00194579, dev acc 0.9267, dev avg loss 0.204502, throughput 9.44897K wps
[Epoch 87 Batch 30/162] avg loss 0.0020945, throughput 9.60847K wps
[Epoch 87 Batch 60/162] avg loss 0.00196261, throughput 9.56735K wps
[Epoch 87 Batch 90/162] avg loss 0.00185796, throughput 9.30321K wps
[Epoch 87 Batch 120/162] avg loss 0.00199678, throughput 9.55494K wps
[Epoch 87 Batch 150/162] avg loss 0.00180689, throughput 9.58483K wps
Begin Testing...
[Epoch 87] train avg loss 0.00193965, dev acc 0.9233, dev avg loss 0.203174, throughput 9.49587K wps
[Epoch 88 Batch 30/162] avg loss 0.00187115, throughput 9.47632K wps
[Epoch 88 Batch 60/162] avg loss 0.00200695, throughput 9.28715K wps
[Epoch 88 Batch 90/162] avg loss 0.0019297, throughput 9.57029K wps
[Epoch 88 Batch 120/162] avg loss 0.00186603, throughput 9.60643K wps
[Epoch 88 Batch 150/162] avg loss 0.00182841, throughput 9.49164K wps
Begin Testing...
[Epoch 88] train avg loss 0.00190691, dev acc 0.9278, dev avg loss 0.205605, throughput 9.47687K wps
Observed Improvement.
Begin Testing...
[Epoch 89 Batch 30/162] avg loss 0.00201157, throughput 9.58269K wps
[Epoch 89 Batch 60/162] avg loss 0.0018353, throughput 9.50575K wps
[Epoch 89 Batch 90/162] avg loss 0.00177407, throughput 9.25169K wps
[Epoch 89 Batch 120/162] avg loss 0.00221532, throughput 9.35484K wps
[Epoch 89 Batch 150/162] avg loss 0.0018236, throughput 9.51688K wps
Begin Testing...
[Epoch 89] train avg loss 0.00191971, dev acc 0.9278, dev avg loss 0.205617, throughput 9.432K wps
Observed Improvement.
Begin Testing...
[Epoch 90 Batch 30/162] avg loss 0.00173242, throughput 9.47193K wps
[Epoch 90 Batch 60/162] avg loss 0.00191319, throughput 9.50515K wps
[Epoch 90 Batch 90/162] avg loss 0.00198389, throughput 9.47178K wps
[Epoch 90 Batch 120/162] avg loss 0.00174323, throughput 9.41199K wps
[Epoch 90 Batch 150/162] avg loss 0.0017337, throughput 9.54413K wps
Begin Testing...
[Epoch 90] train avg loss 0.00181767, dev acc 0.9267, dev avg loss 0.203583, throughput 9.4627K wps
[Epoch 91 Batch 30/162] avg loss 0.00189083, throughput 9.6461K wps
[Epoch 91 Batch 60/162] avg loss 0.00181472, throughput 9.22837K wps
[Epoch 91 Batch 90/162] avg loss 0.00169943, throughput 9.52899K wps
[Epoch 91 Batch 120/162] avg loss 0.0017313, throughput 9.43203K wps
[Epoch 91 Batch 150/162] avg loss 0.00192294, throughput 9.42006K wps
Begin Testing...
[Epoch 91] train avg loss 0.0018223, dev acc 0.9244, dev avg loss 0.208061, throughput 9.43157K wps
[Epoch 92 Batch 30/162] avg loss 0.00180566, throughput 9.43464K wps
[Epoch 92 Batch 60/162] avg loss 0.00180804, throughput 9.38449K wps
[Epoch 92 Batch 90/162] avg loss 0.0018055, throughput 9.34958K wps
[Epoch 92 Batch 120/162] avg loss 0.00165529, throughput 9.53916K wps
[Epoch 92 Batch 150/162] avg loss 0.00190153, throughput 9.31061K wps
Begin Testing...
[Epoch 92] train avg loss 0.00181841, dev acc 0.9233, dev avg loss 0.204197, throughput 9.41949K wps
[Epoch 93 Batch 30/162] avg loss 0.00185685, throughput 9.49546K wps
[Epoch 93 Batch 60/162] avg loss 0.00179046, throughput 9.35015K wps
[Epoch 93 Batch 90/162] avg loss 0.00190219, throughput 9.53514K wps
[Epoch 93 Batch 120/162] avg loss 0.00181904, throughput 9.48194K wps
[Epoch 93 Batch 150/162] avg loss 0.0017963, throughput 9.4709K wps
Begin Testing...
[Epoch 93] train avg loss 0.00182173, dev acc 0.9278, dev avg loss 0.206679, throughput 9.44883K wps
Observed Improvement.
Begin Testing...
[Epoch 94 Batch 30/162] avg loss 0.00168518, throughput 9.50942K wps
[Epoch 94 Batch 60/162] avg loss 0.00186688, throughput 9.26592K wps
[Epoch 94 Batch 90/162] avg loss 0.00182856, throughput 9.36765K wps
[Epoch 94 Batch 120/162] avg loss 0.00156944, throughput 9.52968K wps
[Epoch 94 Batch 150/162] avg loss 0.0017619, throughput 9.39606K wps
Begin Testing...
[Epoch 94] train avg loss 0.00175731, dev acc 0.9256, dev avg loss 0.203263, throughput 9.4142K wps
[Epoch 95 Batch 30/162] avg loss 0.00171283, throughput 9.7524K wps
[Epoch 95 Batch 60/162] avg loss 0.00172209, throughput 9.24877K wps
[Epoch 95 Batch 90/162] avg loss 0.00165797, throughput 9.44415K wps
[Epoch 95 Batch 120/162] avg loss 0.00183828, throughput 9.31955K wps
[Epoch 95 Batch 150/162] avg loss 0.00164045, throughput 9.32758K wps
Begin Testing...
[Epoch 95] train avg loss 0.00172719, dev acc 0.9233, dev avg loss 0.202573, throughput 9.4151K wps
[Epoch 96 Batch 30/162] avg loss 0.00176907, throughput 9.6458K wps
[Epoch 96 Batch 60/162] avg loss 0.00171052, throughput 9.55114K wps
[Epoch 96 Batch 90/162] avg loss 0.00181498, throughput 9.34954K wps
[Epoch 96 Batch 120/162] avg loss 0.00167936, throughput 9.32675K wps
[Epoch 96 Batch 150/162] avg loss 0.00164998, throughput 9.37631K wps
Begin Testing...
[Epoch 96] train avg loss 0.00171289, dev acc 0.9267, dev avg loss 0.203153, throughput 9.45077K wps
[Epoch 97 Batch 30/162] avg loss 0.0017662, throughput 9.58925K wps
[Epoch 97 Batch 60/162] avg loss 0.00168992, throughput 9.34852K wps
[Epoch 97 Batch 90/162] avg loss 0.00190147, throughput 9.47117K wps
[Epoch 97 Batch 120/162] avg loss 0.00171246, throughput 9.20174K wps
[Epoch 97 Batch 150/162] avg loss 0.00146613, throughput 9.35364K wps
Begin Testing...
[Epoch 97] train avg loss 0.0016845, dev acc 0.9222, dev avg loss 0.202549, throughput 9.38771K wps
[Epoch 98 Batch 30/162] avg loss 0.00159472, throughput 9.58632K wps
[Epoch 98 Batch 60/162] avg loss 0.00175675, throughput 9.3405K wps
[Epoch 98 Batch 90/162] avg loss 0.00171115, throughput 9.40018K wps
[Epoch 98 Batch 120/162] avg loss 0.00159402, throughput 9.40784K wps
[Epoch 98 Batch 150/162] avg loss 0.00162926, throughput 9.44172K wps
Begin Testing...
[Epoch 98] train avg loss 0.00165201, dev acc 0.9244, dev avg loss 0.20302, throughput 9.44257K wps
[Epoch 99 Batch 30/162] avg loss 0.00173909, throughput 9.5522K wps
[Epoch 99 Batch 60/162] avg loss 0.00184814, throughput 9.42307K wps
[Epoch 99 Batch 90/162] avg loss 0.00162314, throughput 9.45897K wps
[Epoch 99 Batch 120/162] avg loss 0.00146513, throughput 9.56546K wps
[Epoch 99 Batch 150/162] avg loss 0.00161377, throughput 9.48221K wps
Begin Testing...
[Epoch 99] train avg loss 0.00166573, dev acc 0.9278, dev avg loss 0.203612, throughput 9.50336K wps
Observed Improvement.
Begin Testing...
[Epoch 100 Batch 30/162] avg loss 0.00169423, throughput 9.64733K wps
[Epoch 100 Batch 60/162] avg loss 0.00150395, throughput 9.36451K wps
[Epoch 100 Batch 90/162] avg loss 0.00164981, throughput 9.34563K wps
[Epoch 100 Batch 120/162] avg loss 0.00152704, throughput 9.28818K wps
[Epoch 100 Batch 150/162] avg loss 0.00173944, throughput 9.44414K wps
Begin Testing...
[Epoch 100] train avg loss 0.00163517, dev acc 0.9278, dev avg loss 0.203906, throughput 9.41164K wps
Observed Improvement.
Begin Testing...
[Epoch 101 Batch 30/162] avg loss 0.00147164, throughput 9.58227K wps
[Epoch 101 Batch 60/162] avg loss 0.00152576, throughput 9.39208K wps
[Epoch 101 Batch 90/162] avg loss 0.00173426, throughput 9.6238K wps
[Epoch 101 Batch 120/162] avg loss 0.00163067, throughput 9.39704K wps
[Epoch 101 Batch 150/162] avg loss 0.00165209, throughput 9.20497K wps
Begin Testing...
[Epoch 101] train avg loss 0.00158864, dev acc 0.9278, dev avg loss 0.203442, throughput 9.42551K wps
Observed Improvement.
Begin Testing...
[Epoch 102 Batch 30/162] avg loss 0.00156151, throughput 9.82321K wps
[Epoch 102 Batch 60/162] avg loss 0.00159652, throughput 9.18796K wps
[Epoch 102 Batch 90/162] avg loss 0.00144551, throughput 9.53362K wps
[Epoch 102 Batch 120/162] avg loss 0.00157773, throughput 9.35128K wps
[Epoch 102 Batch 150/162] avg loss 0.00166616, throughput 9.39012K wps
Begin Testing...
[Epoch 102] train avg loss 0.00154459, dev acc 0.9244, dev avg loss 0.202813, throughput 9.46497K wps
[Epoch 103 Batch 30/162] avg loss 0.00143935, throughput 9.6928K wps
[Epoch 103 Batch 60/162] avg loss 0.00150606, throughput 9.33141K wps
[Epoch 103 Batch 90/162] avg loss 0.00145512, throughput 9.391K wps
[Epoch 103 Batch 120/162] avg loss 0.00165503, throughput 9.31875K wps
[Epoch 103 Batch 150/162] avg loss 0.00162571, throughput 9.35115K wps
Begin Testing...
[Epoch 103] train avg loss 0.00153501, dev acc 0.9300, dev avg loss 0.204344, throughput 9.41236K wps
Observed Improvement.
Begin Testing...
[Epoch 104 Batch 30/162] avg loss 0.00145243, throughput 9.72365K wps
[Epoch 104 Batch 60/162] avg loss 0.00142141, throughput 9.53637K wps
[Epoch 104 Batch 90/162] avg loss 0.00148403, throughput 9.43498K wps
[Epoch 104 Batch 120/162] avg loss 0.00174326, throughput 9.56822K wps
[Epoch 104 Batch 150/162] avg loss 0.00134499, throughput 9.2609K wps
Begin Testing...
[Epoch 104] train avg loss 0.00148887, dev acc 0.9256, dev avg loss 0.202647, throughput 9.49624K wps
[Epoch 105 Batch 30/162] avg loss 0.00155669, throughput 9.75376K wps
[Epoch 105 Batch 60/162] avg loss 0.00144268, throughput 9.35198K wps
[Epoch 105 Batch 90/162] avg loss 0.00166025, throughput 9.32038K wps
[Epoch 105 Batch 120/162] avg loss 0.00145475, throughput 9.32812K wps
[Epoch 105 Batch 150/162] avg loss 0.00152302, throughput 9.461K wps
Begin Testing...
[Epoch 105] train avg loss 0.00155161, dev acc 0.9300, dev avg loss 0.204039, throughput 9.43606K wps
Observed Improvement.
Begin Testing...
[Epoch 106 Batch 30/162] avg loss 0.00154878, throughput 9.46675K wps
[Epoch 106 Batch 60/162] avg loss 0.00143228, throughput 9.55508K wps
[Epoch 106 Batch 90/162] avg loss 0.00145917, throughput 9.51473K wps
[Epoch 106 Batch 120/162] avg loss 0.00143901, throughput 9.27269K wps
[Epoch 106 Batch 150/162] avg loss 0.00152617, throughput 9.39699K wps
Begin Testing...
[Epoch 106] train avg loss 0.00147153, dev acc 0.9267, dev avg loss 0.203112, throughput 9.42909K wps
[Epoch 107 Batch 30/162] avg loss 0.00142399, throughput 9.60883K wps
[Epoch 107 Batch 60/162] avg loss 0.00150874, throughput 9.4849K wps
[Epoch 107 Batch 90/162] avg loss 0.00142399, throughput 9.5064K wps
[Epoch 107 Batch 120/162] avg loss 0.00152666, throughput 9.37109K wps
[Epoch 107 Batch 150/162] avg loss 0.00152154, throughput 9.4137K wps
Begin Testing...
[Epoch 107] train avg loss 0.0014748, dev acc 0.9244, dev avg loss 0.202566, throughput 9.48186K wps
[Epoch 108 Batch 30/162] avg loss 0.00136383, throughput 9.64858K wps
[Epoch 108 Batch 60/162] avg loss 0.00144503, throughput 9.5886K wps
[Epoch 108 Batch 90/162] avg loss 0.00149411, throughput 9.2689K wps
[Epoch 108 Batch 120/162] avg loss 0.00152364, throughput 9.44969K wps
[Epoch 108 Batch 150/162] avg loss 0.00141268, throughput 9.35517K wps
Begin Testing...
[Epoch 108] train avg loss 0.00145268, dev acc 0.9256, dev avg loss 0.203233, throughput 9.45171K wps
[Epoch 109 Batch 30/162] avg loss 0.00133058, throughput 9.58805K wps
[Epoch 109 Batch 60/162] avg loss 0.00139296, throughput 9.57445K wps
[Epoch 109 Batch 90/162] avg loss 0.00139451, throughput 9.35094K wps
[Epoch 109 Batch 120/162] avg loss 0.00143897, throughput 9.47598K wps
[Epoch 109 Batch 150/162] avg loss 0.00153214, throughput 9.46159K wps
Begin Testing...
[Epoch 109] train avg loss 0.00142413, dev acc 0.9300, dev avg loss 0.203549, throughput 9.47454K wps
Observed Improvement.
Begin Testing...
[Epoch 110 Batch 30/162] avg loss 0.00155467, throughput 9.60315K wps
[Epoch 110 Batch 60/162] avg loss 0.00132966, throughput 9.4391K wps
[Epoch 110 Batch 90/162] avg loss 0.00142604, throughput 9.41506K wps
[Epoch 110 Batch 120/162] avg loss 0.00132538, throughput 9.48308K wps
[Epoch 110 Batch 150/162] avg loss 0.00135844, throughput 9.55334K wps
Begin Testing...
[Epoch 110] train avg loss 0.00139693, dev acc 0.9300, dev avg loss 0.204749, throughput 9.4804K wps
Observed Improvement.
Begin Testing...
[Epoch 111 Batch 30/162] avg loss 0.00148706, throughput 9.51214K wps
[Epoch 111 Batch 60/162] avg loss 0.00128188, throughput 9.41677K wps
[Epoch 111 Batch 90/162] avg loss 0.00125431, throughput 9.44944K wps
[Epoch 111 Batch 120/162] avg loss 0.00154536, throughput 9.56035K wps
[Epoch 111 Batch 150/162] avg loss 0.00138976, throughput 9.34475K wps
Begin Testing...
[Epoch 111] train avg loss 0.00139583, dev acc 0.9267, dev avg loss 0.203193, throughput 9.44803K wps
[Epoch 112 Batch 30/162] avg loss 0.00130086, throughput 9.65281K wps
[Epoch 112 Batch 60/162] avg loss 0.00145844, throughput 9.36907K wps
[Epoch 112 Batch 90/162] avg loss 0.00128854, throughput 9.56143K wps
[Epoch 112 Batch 120/162] avg loss 0.00148728, throughput 9.33713K wps
[Epoch 112 Batch 150/162] avg loss 0.00149181, throughput 9.55043K wps
Begin Testing...
[Epoch 112] train avg loss 0.00140478, dev acc 0.9300, dev avg loss 0.205582, throughput 9.48891K wps
Observed Improvement.
Begin Testing...
[Epoch 113 Batch 30/162] avg loss 0.00143225, throughput 9.56638K wps
[Epoch 113 Batch 60/162] avg loss 0.0013402, throughput 9.39206K wps
[Epoch 113 Batch 90/162] avg loss 0.0013842, throughput 9.41948K wps
[Epoch 113 Batch 120/162] avg loss 0.00131164, throughput 9.37399K wps
[Epoch 113 Batch 150/162] avg loss 0.0011931, throughput 9.41034K wps
Begin Testing...
[Epoch 113] train avg loss 0.00135018, dev acc 0.9256, dev avg loss 0.203325, throughput 9.41288K wps
[Epoch 114 Batch 30/162] avg loss 0.00132566, throughput 9.60661K wps
[Epoch 114 Batch 60/162] avg loss 0.00136817, throughput 9.60526K wps
[Epoch 114 Batch 90/162] avg loss 0.00149376, throughput 9.37914K wps
[Epoch 114 Batch 120/162] avg loss 0.0013825, throughput 9.41737K wps
[Epoch 114 Batch 150/162] avg loss 0.00119939, throughput 9.40969K wps
Begin Testing...
[Epoch 114] train avg loss 0.00134506, dev acc 0.9289, dev avg loss 0.204735, throughput 9.48441K wps
[Epoch 115 Batch 30/162] avg loss 0.00141044, throughput 9.49937K wps
[Epoch 115 Batch 60/162] avg loss 0.00129606, throughput 9.39469K wps
[Epoch 115 Batch 90/162] avg loss 0.00130513, throughput 9.29736K wps
[Epoch 115 Batch 120/162] avg loss 0.00143158, throughput 9.3925K wps
[Epoch 115 Batch 150/162] avg loss 0.00137308, throughput 9.60505K wps
Begin Testing...
[Epoch 115] train avg loss 0.00135278, dev acc 0.9300, dev avg loss 0.204325, throughput 9.42855K wps
Observed Improvement.
Begin Testing...
[Epoch 116 Batch 30/162] avg loss 0.00133856, throughput 9.53564K wps
[Epoch 116 Batch 60/162] avg loss 0.00133457, throughput 9.40379K wps
[Epoch 116 Batch 90/162] avg loss 0.00120413, throughput 9.39078K wps
[Epoch 116 Batch 120/162] avg loss 0.00132599, throughput 9.41481K wps
[Epoch 116 Batch 150/162] avg loss 0.00132068, throughput 9.40372K wps
Begin Testing...
[Epoch 116] train avg loss 0.00130508, dev acc 0.9267, dev avg loss 0.203752, throughput 9.45161K wps
[Epoch 117 Batch 30/162] avg loss 0.00127572, throughput 9.75878K wps
[Epoch 117 Batch 60/162] avg loss 0.00118339, throughput 9.4722K wps
[Epoch 117 Batch 90/162] avg loss 0.00134391, throughput 9.57793K wps
[Epoch 117 Batch 120/162] avg loss 0.00139525, throughput 9.57794K wps
[Epoch 117 Batch 150/162] avg loss 0.0012647, throughput 9.57519K wps
Begin Testing...
[Epoch 117] train avg loss 0.00129721, dev acc 0.9289, dev avg loss 0.204241, throughput 9.5742K wps
[Epoch 118 Batch 30/162] avg loss 0.00124256, throughput 9.64826K wps
[Epoch 118 Batch 60/162] avg loss 0.00128714, throughput 9.45643K wps
[Epoch 118 Batch 90/162] avg loss 0.00126695, throughput 9.30811K wps
[Epoch 118 Batch 120/162] avg loss 0.00142096, throughput 9.57029K wps
[Epoch 118 Batch 150/162] avg loss 0.00149038, throughput 9.64895K wps
Begin Testing...
[Epoch 118] train avg loss 0.00132342, dev acc 0.9244, dev avg loss 0.203682, throughput 9.50004K wps
[Epoch 119 Batch 30/162] avg loss 0.00130218, throughput 9.6435K wps
[Epoch 119 Batch 60/162] avg loss 0.00135469, throughput 9.30984K wps
[Epoch 119 Batch 90/162] avg loss 0.00130258, throughput 9.5508K wps
[Epoch 119 Batch 120/162] avg loss 0.00124263, throughput 9.2635K wps
[Epoch 119 Batch 150/162] avg loss 0.00130907, throughput 9.5201K wps
Begin Testing...
[Epoch 119] train avg loss 0.00129746, dev acc 0.9267, dev avg loss 0.203613, throughput 9.44268K wps
[Epoch 120 Batch 30/162] avg loss 0.00114495, throughput 9.50343K wps
[Epoch 120 Batch 60/162] avg loss 0.001234, throughput 9.36252K wps
[Epoch 120 Batch 90/162] avg loss 0.0011495, throughput 9.34622K wps
[Epoch 120 Batch 120/162] avg loss 0.00132206, throughput 9.62225K wps
[Epoch 120 Batch 150/162] avg loss 0.00120413, throughput 9.28745K wps
Begin Testing...
[Epoch 120] train avg loss 0.00122625, dev acc 0.9289, dev avg loss 0.203438, throughput 9.41468K wps
[Epoch 121 Batch 30/162] avg loss 0.00133845, throughput 9.72424K wps
[Epoch 121 Batch 60/162] avg loss 0.0012029, throughput 9.49389K wps
[Epoch 121 Batch 90/162] avg loss 0.00133132, throughput 9.50511K wps
[Epoch 121 Batch 120/162] avg loss 0.0012475, throughput 9.64547K wps
[Epoch 121 Batch 150/162] avg loss 0.00129115, throughput 9.41604K wps
Begin Testing...
[Epoch 121] train avg loss 0.00126823, dev acc 0.9278, dev avg loss 0.204274, throughput 9.53006K wps
[Epoch 122 Batch 30/162] avg loss 0.00131787, throughput 9.66629K wps
[Epoch 122 Batch 60/162] avg loss 0.00104028, throughput 9.41489K wps
[Epoch 122 Batch 90/162] avg loss 0.00118583, throughput 9.4674K wps
[Epoch 122 Batch 120/162] avg loss 0.00115548, throughput 9.43033K wps
[Epoch 122 Batch 150/162] avg loss 0.00115729, throughput 9.49702K wps
Begin Testing...
[Epoch 122] train avg loss 0.0011811, dev acc 0.9267, dev avg loss 0.204121, throughput 9.50339K wps
[Epoch 123 Batch 30/162] avg loss 0.00114433, throughput 9.72558K wps
[Epoch 123 Batch 60/162] avg loss 0.00116582, throughput 9.24607K wps
[Epoch 123 Batch 90/162] avg loss 0.00128206, throughput 9.33462K wps
[Epoch 123 Batch 120/162] avg loss 0.00130178, throughput 9.3465K wps
[Epoch 123 Batch 150/162] avg loss 0.00119308, throughput 9.41474K wps
Begin Testing...
[Epoch 123] train avg loss 0.00119958, dev acc 0.9311, dev avg loss 0.205841, throughput 9.43077K wps
Observed Improvement.
Begin Testing...
[Epoch 124 Batch 30/162] avg loss 0.00122898, throughput 9.5053K wps
[Epoch 124 Batch 60/162] avg loss 0.00124661, throughput 9.58547K wps
[Epoch 124 Batch 90/162] avg loss 0.00106452, throughput 9.36988K wps
[Epoch 124 Batch 120/162] avg loss 0.00121448, throughput 9.427K wps
[Epoch 124 Batch 150/162] avg loss 0.00109225, throughput 9.48049K wps
Begin Testing...
[Epoch 124] train avg loss 0.00118512, dev acc 0.9300, dev avg loss 0.206674, throughput 9.45557K wps
[Epoch 125 Batch 30/162] avg loss 0.00129677, throughput 9.6843K wps
[Epoch 125 Batch 60/162] avg loss 0.0011061, throughput 9.38289K wps
[Epoch 125 Batch 90/162] avg loss 0.00108579, throughput 9.55422K wps
[Epoch 125 Batch 120/162] avg loss 0.00103728, throughput 9.3793K wps
[Epoch 125 Batch 150/162] avg loss 0.0013956, throughput 9.53306K wps
Begin Testing...
[Epoch 125] train avg loss 0.00118828, dev acc 0.9311, dev avg loss 0.205617, throughput 9.48362K wps
Observed Improvement.
Begin Testing...
[Epoch 126 Batch 30/162] avg loss 0.00116667, throughput 9.71283K wps
[Epoch 126 Batch 60/162] avg loss 0.00119379, throughput 9.39971K wps
[Epoch 126 Batch 90/162] avg loss 0.00103762, throughput 9.48152K wps
[Epoch 126 Batch 120/162] avg loss 0.00128346, throughput 9.31799K wps
[Epoch 126 Batch 150/162] avg loss 0.00127683, throughput 9.48183K wps
Begin Testing...
[Epoch 126] train avg loss 0.00119851, dev acc 0.9311, dev avg loss 0.205669, throughput 9.48158K wps
Observed Improvement.
Begin Testing...
[Epoch 127 Batch 30/162] avg loss 0.00113048, throughput 9.49985K wps
[Epoch 127 Batch 60/162] avg loss 0.00114098, throughput 9.41328K wps
[Epoch 127 Batch 90/162] avg loss 0.00112209, throughput 9.64142K wps
[Epoch 127 Batch 120/162] avg loss 0.00105071, throughput 9.58716K wps
[Epoch 127 Batch 150/162] avg loss 0.00127955, throughput 9.51895K wps
Begin Testing...
[Epoch 127] train avg loss 0.00114562, dev acc 0.9256, dev avg loss 0.20441, throughput 9.52356K wps
[Epoch 128 Batch 30/162] avg loss 0.00114873, throughput 9.80845K wps
[Epoch 128 Batch 60/162] avg loss 0.00116394, throughput 9.47118K wps
[Epoch 128 Batch 90/162] avg loss 0.00124525, throughput 9.61681K wps
[Epoch 128 Batch 120/162] avg loss 0.00109768, throughput 9.41459K wps
[Epoch 128 Batch 150/162] avg loss 0.00108655, throughput 9.38759K wps
Begin Testing...
[Epoch 128] train avg loss 0.00114934, dev acc 0.9278, dev avg loss 0.205567, throughput 9.51295K wps
[Epoch 129 Batch 30/162] avg loss 0.00125837, throughput 9.65171K wps
[Epoch 129 Batch 60/162] avg loss 0.00115189, throughput 9.21809K wps
[Epoch 129 Batch 90/162] avg loss 0.00117919, throughput 9.37832K wps
[Epoch 129 Batch 120/162] avg loss 0.00113498, throughput 9.37726K wps
[Epoch 129 Batch 150/162] avg loss 0.00119157, throughput 9.48916K wps
Begin Testing...
[Epoch 129] train avg loss 0.00117978, dev acc 0.9244, dev avg loss 0.204747, throughput 9.42372K wps
[Epoch 130 Batch 30/162] avg loss 0.00119519, throughput 9.60839K wps
[Epoch 130 Batch 60/162] avg loss 0.000988003, throughput 9.37438K wps
[Epoch 130 Batch 90/162] avg loss 0.00116063, throughput 9.36115K wps
[Epoch 130 Batch 120/162] avg loss 0.00117226, throughput 9.42503K wps
[Epoch 130 Batch 150/162] avg loss 0.00123668, throughput 9.46941K wps
Begin Testing...
[Epoch 130] train avg loss 0.00114516, dev acc 0.9256, dev avg loss 0.205161, throughput 9.4326K wps
[Epoch 131 Batch 30/162] avg loss 0.00104878, throughput 9.40524K wps
[Epoch 131 Batch 60/162] avg loss 0.00128413, throughput 9.46712K wps
[Epoch 131 Batch 90/162] avg loss 0.000946548, throughput 9.49282K wps
[Epoch 131 Batch 120/162] avg loss 0.00121314, throughput 9.43849K wps
[Epoch 131 Batch 150/162] avg loss 0.00119835, throughput 9.33158K wps
Begin Testing...
[Epoch 131] train avg loss 0.00113438, dev acc 0.9278, dev avg loss 0.204637, throughput 9.42794K wps
[Epoch 132 Batch 30/162] avg loss 0.0010235, throughput 9.58772K wps
[Epoch 132 Batch 60/162] avg loss 0.000985217, throughput 9.39667K wps
[Epoch 132 Batch 90/162] avg loss 0.00116398, throughput 9.55372K wps
[Epoch 132 Batch 120/162] avg loss 0.00109836, throughput 9.60844K wps
[Epoch 132 Batch 150/162] avg loss 0.0010202, throughput 9.32723K wps
Begin Testing...
[Epoch 132] train avg loss 0.00109524, dev acc 0.9267, dev avg loss 0.204818, throughput 9.4826K wps
[Epoch 133 Batch 30/162] avg loss 0.00102277, throughput 9.4083K wps
[Epoch 133 Batch 60/162] avg loss 0.000997839, throughput 9.46614K wps
[Epoch 133 Batch 90/162] avg loss 0.00100032, throughput 9.46387K wps
[Epoch 133 Batch 120/162] avg loss 0.00106547, throughput 9.43764K wps
[Epoch 133 Batch 150/162] avg loss 0.00116402, throughput 9.23098K wps
Begin Testing...
[Epoch 133] train avg loss 0.00106587, dev acc 0.9267, dev avg loss 0.204948, throughput 9.39379K wps
[Epoch 134 Batch 30/162] avg loss 0.00105663, throughput 9.62409K wps
[Epoch 134 Batch 60/162] avg loss 0.0011933, throughput 9.33625K wps
[Epoch 134 Batch 90/162] avg loss 0.00107912, throughput 9.5041K wps
[Epoch 134 Batch 120/162] avg loss 0.00106092, throughput 9.41616K wps
[Epoch 134 Batch 150/162] avg loss 0.000999577, throughput 9.31532K wps
Begin Testing...
[Epoch 134] train avg loss 0.00107234, dev acc 0.9244, dev avg loss 0.205479, throughput 9.44828K wps
[Epoch 135 Batch 30/162] avg loss 0.00107099, throughput 9.75594K wps
[Epoch 135 Batch 60/162] avg loss 0.00112744, throughput 9.51418K wps
[Epoch 135 Batch 90/162] avg loss 0.00108513, throughput 9.50769K wps
[Epoch 135 Batch 120/162] avg loss 0.00101637, throughput 9.2922K wps
[Epoch 135 Batch 150/162] avg loss 0.000931342, throughput 9.38479K wps
Begin Testing...
[Epoch 135] train avg loss 0.00104465, dev acc 0.9278, dev avg loss 0.21125, throughput 9.47741K wps
[Epoch 136 Batch 30/162] avg loss 0.00100281, throughput 9.67769K wps
[Epoch 136 Batch 60/162] avg loss 0.00102594, throughput 9.60204K wps
[Epoch 136 Batch 90/162] avg loss 0.00112902, throughput 9.38025K wps
[Epoch 136 Batch 120/162] avg loss 0.00109966, throughput 9.54946K wps
[Epoch 136 Batch 150/162] avg loss 0.00101322, throughput 9.31063K wps
Begin Testing...
[Epoch 136] train avg loss 0.0010547, dev acc 0.9289, dev avg loss 0.205828, throughput 9.49261K wps
[Epoch 137 Batch 30/162] avg loss 0.00103059, throughput 9.60185K wps
[Epoch 137 Batch 60/162] avg loss 0.00106634, throughput 9.52127K wps
[Epoch 137 Batch 90/162] avg loss 0.00109773, throughput 9.34847K wps
[Epoch 137 Batch 120/162] avg loss 0.00106444, throughput 9.63085K wps
[Epoch 137 Batch 150/162] avg loss 0.000942552, throughput 9.50762K wps
Begin Testing...
[Epoch 137] train avg loss 0.00104561, dev acc 0.9322, dev avg loss 0.206864, throughput 9.50197K wps
Observed Improvement.
Begin Testing...
[Epoch 138 Batch 30/162] avg loss 0.00108039, throughput 9.63981K wps
[Epoch 138 Batch 60/162] avg loss 0.000928163, throughput 9.22343K wps
[Epoch 138 Batch 90/162] avg loss 0.00112476, throughput 9.44039K wps
[Epoch 138 Batch 120/162] avg loss 0.000972664, throughput 9.43594K wps
[Epoch 138 Batch 150/162] avg loss 0.00106086, throughput 9.47911K wps
Begin Testing...
[Epoch 138] train avg loss 0.00103877, dev acc 0.9322, dev avg loss 0.207329, throughput 9.42277K wps
Observed Improvement.
Begin Testing...
[Epoch 139 Batch 30/162] avg loss 0.000921321, throughput 9.50083K wps
[Epoch 139 Batch 60/162] avg loss 0.00117006, throughput 9.49219K wps
[Epoch 139 Batch 90/162] avg loss 0.00104691, throughput 9.46558K wps
[Epoch 139 Batch 120/162] avg loss 0.000911981, throughput 9.58023K wps
[Epoch 139 Batch 150/162] avg loss 0.000925314, throughput 9.24772K wps
Begin Testing...
[Epoch 139] train avg loss 0.000992025, dev acc 0.9311, dev avg loss 0.206755, throughput 9.47003K wps
[Epoch 140 Batch 30/162] avg loss 0.00114545, throughput 9.71335K wps
[Epoch 140 Batch 60/162] avg loss 0.000984249, throughput 9.33462K wps
[Epoch 140 Batch 90/162] avg loss 0.000981308, throughput 9.61263K wps
[Epoch 140 Batch 120/162] avg loss 0.000972837, throughput 9.39843K wps
[Epoch 140 Batch 150/162] avg loss 0.000997537, throughput 9.21936K wps
Begin Testing...
[Epoch 140] train avg loss 0.00100955, dev acc 0.9311, dev avg loss 0.207087, throughput 9.46209K wps
[Epoch 141 Batch 30/162] avg loss 0.00106212, throughput 9.6012K wps
[Epoch 141 Batch 60/162] avg loss 0.000926665, throughput 9.56276K wps
[Epoch 141 Batch 90/162] avg loss 0.000942565, throughput 9.3576K wps
[Epoch 141 Batch 120/162] avg loss 0.00097107, throughput 9.34861K wps
[Epoch 141 Batch 150/162] avg loss 0.000938971, throughput 9.5307K wps
Begin Testing...
[Epoch 141] train avg loss 0.000976918, dev acc 0.9233, dev avg loss 0.205859, throughput 9.4582K wps
[Epoch 142 Batch 30/162] avg loss 0.0010335, throughput 9.43355K wps
[Epoch 142 Batch 60/162] avg loss 0.00102742, throughput 9.33467K wps
[Epoch 142 Batch 90/162] avg loss 0.000960278, throughput 9.49943K wps
[Epoch 142 Batch 120/162] avg loss 0.000926676, throughput 9.62473K wps
[Epoch 142 Batch 150/162] avg loss 0.00107193, throughput 9.30903K wps
Begin Testing...
[Epoch 142] train avg loss 0.000996268, dev acc 0.9278, dev avg loss 0.206189, throughput 9.44869K wps
[Epoch 143 Batch 30/162] avg loss 0.000988357, throughput 9.75837K wps
[Epoch 143 Batch 60/162] avg loss 0.000930853, throughput 9.44455K wps
[Epoch 143 Batch 90/162] avg loss 0.000866374, throughput 9.38079K wps
[Epoch 143 Batch 120/162] avg loss 0.000911915, throughput 9.39714K wps
[Epoch 143 Batch 150/162] avg loss 0.000982973, throughput 9.57111K wps
Begin Testing...
[Epoch 143] train avg loss 0.000949267, dev acc 0.9311, dev avg loss 0.20794, throughput 9.4974K wps
[Epoch 144 Batch 30/162] avg loss 0.000918662, throughput 9.52234K wps
[Epoch 144 Batch 60/162] avg loss 0.0009974, throughput 9.3135K wps
[Epoch 144 Batch 90/162] avg loss 0.00104768, throughput 9.32049K wps
[Epoch 144 Batch 120/162] avg loss 0.000910811, throughput 9.59731K wps
[Epoch 144 Batch 150/162] avg loss 0.00094092, throughput 9.31257K wps
Begin Testing...
[Epoch 144] train avg loss 0.000963589, dev acc 0.9244, dev avg loss 0.206333, throughput 9.43464K wps
[Epoch 145 Batch 30/162] avg loss 0.000873784, throughput 9.52366K wps
[Epoch 145 Batch 60/162] avg loss 0.000966652, throughput 9.42417K wps
[Epoch 145 Batch 90/162] avg loss 0.000989104, throughput 9.55852K wps
[Epoch 145 Batch 120/162] avg loss 0.000996362, throughput 9.38341K wps
[Epoch 145 Batch 150/162] avg loss 0.000972402, throughput 9.45512K wps
Begin Testing...
[Epoch 145] train avg loss 0.00095145, dev acc 0.9267, dev avg loss 0.206423, throughput 9.45226K wps
[Epoch 146 Batch 30/162] avg loss 0.000888272, throughput 9.67792K wps
[Epoch 146 Batch 60/162] avg loss 0.000851266, throughput 9.42893K wps
[Epoch 146 Batch 90/162] avg loss 0.000879877, throughput 9.38014K wps
[Epoch 146 Batch 120/162] avg loss 0.000994568, throughput 9.51497K wps
[Epoch 146 Batch 150/162] avg loss 0.000815247, throughput 9.4778K wps
Begin Testing...
[Epoch 146] train avg loss 0.000879662, dev acc 0.9267, dev avg loss 0.206844, throughput 9.48311K wps
[Epoch 147 Batch 30/162] avg loss 0.000841456, throughput 9.56263K wps
[Epoch 147 Batch 60/162] avg loss 0.000942986, throughput 9.43979K wps
[Epoch 147 Batch 90/162] avg loss 0.000876734, throughput 9.34722K wps
[Epoch 147 Batch 120/162] avg loss 0.000967253, throughput 9.4942K wps
[Epoch 147 Batch 150/162] avg loss 0.00103534, throughput 9.36437K wps
Begin Testing...
[Epoch 147] train avg loss 0.000923816, dev acc 0.9267, dev avg loss 0.207681, throughput 9.42605K wps
[Epoch 148 Batch 30/162] avg loss 0.000905358, throughput 9.65712K wps
[Epoch 148 Batch 60/162] avg loss 0.00107609, throughput 9.40492K wps
[Epoch 148 Batch 90/162] avg loss 0.00087812, throughput 9.41856K wps
[Epoch 148 Batch 120/162] avg loss 0.00088526, throughput 9.38063K wps
[Epoch 148 Batch 150/162] avg loss 0.000851569, throughput 9.3986K wps
Begin Testing...
[Epoch 148] train avg loss 0.00091754, dev acc 0.9322, dev avg loss 0.208332, throughput 9.43942K wps
Observed Improvement.
Begin Testing...
[Epoch 149 Batch 30/162] avg loss 0.000783728, throughput 9.50271K wps
[Epoch 149 Batch 60/162] avg loss 0.000945813, throughput 9.3832K wps
[Epoch 149 Batch 90/162] avg loss 0.000885649, throughput 9.35248K wps
[Epoch 149 Batch 120/162] avg loss 0.000891654, throughput 9.52465K wps
[Epoch 149 Batch 150/162] avg loss 0.000932909, throughput 9.31763K wps
Begin Testing...
[Epoch 149] train avg loss 0.000884067, dev acc 0.9322, dev avg loss 0.208596, throughput 9.42054K wps
Observed Improvement.
Begin Testing...
[Epoch 150 Batch 30/162] avg loss 0.00089138, throughput 9.60841K wps
[Epoch 150 Batch 60/162] avg loss 0.000797745, throughput 9.39194K wps
[Epoch 150 Batch 90/162] avg loss 0.000808908, throughput 9.46476K wps
[Epoch 150 Batch 120/162] avg loss 0.000945538, throughput 9.48622K wps
[Epoch 150 Batch 150/162] avg loss 0.00111477, throughput 9.36436K wps
Begin Testing...
[Epoch 150] train avg loss 0.000909956, dev acc 0.9244, dev avg loss 0.207124, throughput 9.47785K wps
[Epoch 151 Batch 30/162] avg loss 0.00084703, throughput 9.71896K wps
[Epoch 151 Batch 60/162] avg loss 0.000807382, throughput 9.41173K wps
[Epoch 151 Batch 90/162] avg loss 0.000813106, throughput 9.2957K wps
[Epoch 151 Batch 120/162] avg loss 0.0009072, throughput 9.55517K wps
[Epoch 151 Batch 150/162] avg loss 0.000894305, throughput 9.57697K wps
Begin Testing...
[Epoch 151] train avg loss 0.000864847, dev acc 0.9311, dev avg loss 0.209318, throughput 9.48399K wps
[Epoch 152 Batch 30/162] avg loss 0.000932019, throughput 9.7217K wps
[Epoch 152 Batch 60/162] avg loss 0.000956861, throughput 9.28493K wps
[Epoch 152 Batch 90/162] avg loss 0.000832195, throughput 9.52803K wps
[Epoch 152 Batch 120/162] avg loss 0.000952767, throughput 9.34234K wps
[Epoch 152 Batch 150/162] avg loss 0.000849126, throughput 9.46579K wps
Begin Testing...
[Epoch 152] train avg loss 0.00089712, dev acc 0.9300, dev avg loss 0.208925, throughput 9.47273K wps
[Epoch 153 Batch 30/162] avg loss 0.000885167, throughput 9.63919K wps
[Epoch 153 Batch 60/162] avg loss 0.000880176, throughput 9.36374K wps
[Epoch 153 Batch 90/162] avg loss 0.000808268, throughput 9.30946K wps
[Epoch 153 Batch 120/162] avg loss 0.000869969, throughput 9.52152K wps
[Epoch 153 Batch 150/162] avg loss 0.000894881, throughput 9.21274K wps
Begin Testing...
[Epoch 153] train avg loss 0.000857002, dev acc 0.9300, dev avg loss 0.210177, throughput 9.42385K wps
[Epoch 154 Batch 30/162] avg loss 0.000887949, throughput 9.70552K wps
[Epoch 154 Batch 60/162] avg loss 0.000864549, throughput 9.52723K wps
[Epoch 154 Batch 90/162] avg loss 0.00083949, throughput 9.34817K wps
[Epoch 154 Batch 120/162] avg loss 0.000851244, throughput 9.41092K wps
[Epoch 154 Batch 150/162] avg loss 0.000816988, throughput 9.43089K wps
Begin Testing...
[Epoch 154] train avg loss 0.000855057, dev acc 0.9300, dev avg loss 0.209018, throughput 9.4855K wps
[Epoch 155 Batch 30/162] avg loss 0.000875852, throughput 9.70829K wps
[Epoch 155 Batch 60/162] avg loss 0.000809426, throughput 9.37225K wps
[Epoch 155 Batch 90/162] avg loss 0.000884458, throughput 9.5543K wps
[Epoch 155 Batch 120/162] avg loss 0.000794363, throughput 9.52988K wps
[Epoch 155 Batch 150/162] avg loss 0.000871111, throughput 9.33332K wps
Begin Testing...
[Epoch 155] train avg loss 0.000851296, dev acc 0.9333, dev avg loss 0.208595, throughput 9.50161K wps
Observed Improvement.
Begin Testing...
[Epoch 156 Batch 30/162] avg loss 0.000785985, throughput 9.68545K wps
[Epoch 156 Batch 60/162] avg loss 0.000824204, throughput 9.36398K wps
[Epoch 156 Batch 90/162] avg loss 0.000871296, throughput 9.37758K wps
[Epoch 156 Batch 120/162] avg loss 0.000858668, throughput 9.3443K wps
[Epoch 156 Batch 150/162] avg loss 0.000911469, throughput 9.37841K wps
Begin Testing...
[Epoch 156] train avg loss 0.00085598, dev acc 0.9300, dev avg loss 0.209687, throughput 9.40992K wps
[Epoch 157 Batch 30/162] avg loss 0.000774689, throughput 9.48316K wps
[Epoch 157 Batch 60/162] avg loss 0.000933474, throughput 9.4289K wps
[Epoch 157 Batch 90/162] avg loss 0.000846877, throughput 9.58464K wps
[Epoch 157 Batch 120/162] avg loss 0.000806663, throughput 9.36784K wps
[Epoch 157 Batch 150/162] avg loss 0.000780336, throughput 9.58956K wps
Begin Testing...
[Epoch 157] train avg loss 0.000834758, dev acc 0.9289, dev avg loss 0.210351, throughput 9.47255K wps
[Epoch 158 Batch 30/162] avg loss 0.000860296, throughput 9.56351K wps
[Epoch 158 Batch 60/162] avg loss 0.000788706, throughput 9.3729K wps
[Epoch 158 Batch 90/162] avg loss 0.000806336, throughput 9.53732K wps
[Epoch 158 Batch 120/162] avg loss 0.000914871, throughput 9.38795K wps
[Epoch 158 Batch 150/162] avg loss 0.000761012, throughput 9.37813K wps
Begin Testing...
[Epoch 158] train avg loss 0.000823487, dev acc 0.9289, dev avg loss 0.210461, throughput 9.4434K wps
[Epoch 159 Batch 30/162] avg loss 0.000773153, throughput 9.74385K wps
[Epoch 159 Batch 60/162] avg loss 0.000847269, throughput 9.60629K wps
[Epoch 159 Batch 90/162] avg loss 0.000847657, throughput 9.43131K wps
[Epoch 159 Batch 120/162] avg loss 0.00078649, throughput 9.58769K wps
[Epoch 159 Batch 150/162] avg loss 0.000910715, throughput 9.32189K wps
Begin Testing...
[Epoch 159] train avg loss 0.000830412, dev acc 0.9311, dev avg loss 0.210394, throughput 9.53513K wps
[Epoch 160 Batch 30/162] avg loss 0.000754895, throughput 9.62348K wps
[Epoch 160 Batch 60/162] avg loss 0.000799484, throughput 9.62182K wps
[Epoch 160 Batch 90/162] avg loss 0.000867502, throughput 9.54308K wps
[Epoch 160 Batch 120/162] avg loss 0.000777253, throughput 9.58816K wps
[Epoch 160 Batch 150/162] avg loss 0.000862022, throughput 9.32431K wps
Begin Testing...
[Epoch 160] train avg loss 0.000812389, dev acc 0.9322, dev avg loss 0.21004, throughput 9.51667K wps
[Epoch 161 Batch 30/162] avg loss 0.000772623, throughput 9.80789K wps
[Epoch 161 Batch 60/162] avg loss 0.000846794, throughput 9.28624K wps
[Epoch 161 Batch 90/162] avg loss 0.000734095, throughput 9.3992K wps
[Epoch 161 Batch 120/162] avg loss 0.000743982, throughput 9.4238K wps
[Epoch 161 Batch 150/162] avg loss 0.00087136, throughput 9.43065K wps
Begin Testing...
[Epoch 161] train avg loss 0.000797339, dev acc 0.9300, dev avg loss 0.211194, throughput 9.47018K wps
[Epoch 162 Batch 30/162] avg loss 0.000825885, throughput 9.55719K wps
[Epoch 162 Batch 60/162] avg loss 0.00088123, throughput 9.34066K wps
[Epoch 162 Batch 90/162] avg loss 0.00076485, throughput 9.49918K wps
[Epoch 162 Batch 120/162] avg loss 0.000775405, throughput 9.41827K wps
[Epoch 162 Batch 150/162] avg loss 0.000785462, throughput 9.56143K wps
Begin Testing...
[Epoch 162] train avg loss 0.000796468, dev acc 0.9322, dev avg loss 0.211644, throughput 9.47993K wps
[Epoch 163 Batch 30/162] avg loss 0.00081531, throughput 9.62265K wps
[Epoch 163 Batch 60/162] avg loss 0.000731377, throughput 9.33086K wps
[Epoch 163 Batch 90/162] avg loss 0.000742062, throughput 9.36019K wps
[Epoch 163 Batch 120/162] avg loss 0.000798287, throughput 9.39763K wps
[Epoch 163 Batch 150/162] avg loss 0.000848569, throughput 9.47662K wps
Begin Testing...
[Epoch 163] train avg loss 0.000797465, dev acc 0.9267, dev avg loss 0.209602, throughput 9.44227K wps
[Epoch 164 Batch 30/162] avg loss 0.00087002, throughput 9.55999K wps
[Epoch 164 Batch 60/162] avg loss 0.000777993, throughput 9.3746K wps
[Epoch 164 Batch 90/162] avg loss 0.00075153, throughput 9.37108K wps
[Epoch 164 Batch 120/162] avg loss 0.00072447, throughput 9.65132K wps
[Epoch 164 Batch 150/162] avg loss 0.000771806, throughput 9.55404K wps
Begin Testing...
[Epoch 164] train avg loss 0.000786239, dev acc 0.9278, dev avg loss 0.210404, throughput 9.4878K wps
[Epoch 165 Batch 30/162] avg loss 0.000789808, throughput 9.67482K wps
[Epoch 165 Batch 60/162] avg loss 0.000779157, throughput 9.51742K wps
[Epoch 165 Batch 90/162] avg loss 0.000790593, throughput 9.49024K wps
[Epoch 165 Batch 120/162] avg loss 0.000825362, throughput 9.49705K wps
[Epoch 165 Batch 150/162] avg loss 0.000756317, throughput 9.52084K wps
Begin Testing...
[Epoch 165] train avg loss 0.000794623, dev acc 0.9289, dev avg loss 0.211018, throughput 9.53726K wps
[Epoch 166 Batch 30/162] avg loss 0.000822924, throughput 9.72664K wps
[Epoch 166 Batch 60/162] avg loss 0.000717983, throughput 9.47922K wps
[Epoch 166 Batch 90/162] avg loss 0.00079494, throughput 9.51166K wps
[Epoch 166 Batch 120/162] avg loss 0.000858493, throughput 9.42722K wps
[Epoch 166 Batch 150/162] avg loss 0.000755342, throughput 9.29247K wps
Begin Testing...
[Epoch 166] train avg loss 0.000787149, dev acc 0.9322, dev avg loss 0.211128, throughput 9.47453K wps
[Epoch 167 Batch 30/162] avg loss 0.000789909, throughput 9.58568K wps
[Epoch 167 Batch 60/162] avg loss 0.000754604, throughput 9.44013K wps
[Epoch 167 Batch 90/162] avg loss 0.000759372, throughput 9.57767K wps
[Epoch 167 Batch 120/162] avg loss 0.000775806, throughput 9.43504K wps
[Epoch 167 Batch 150/162] avg loss 0.000733013, throughput 9.46101K wps
Begin Testing...
[Epoch 167] train avg loss 0.000759399, dev acc 0.9300, dev avg loss 0.21092, throughput 9.50956K wps
[Epoch 168 Batch 30/162] avg loss 0.000715303, throughput 9.55771K wps
[Epoch 168 Batch 60/162] avg loss 0.000878609, throughput 9.65598K wps
[Epoch 168 Batch 90/162] avg loss 0.000747658, throughput 9.45439K wps
[Epoch 168 Batch 120/162] avg loss 0.000729606, throughput 9.22393K wps
[Epoch 168 Batch 150/162] avg loss 0.000736736, throughput 9.36167K wps
Begin Testing...
[Epoch 168] train avg loss 0.000773096, dev acc 0.9233, dev avg loss 0.209812, throughput 9.46253K wps
[Epoch 169 Batch 30/162] avg loss 0.000689768, throughput 9.43679K wps
[Epoch 169 Batch 60/162] avg loss 0.000838235, throughput 9.3361K wps
[Epoch 169 Batch 90/162] avg loss 0.0008769, throughput 9.54471K wps
[Epoch 169 Batch 120/162] avg loss 0.000738192, throughput 9.42164K wps
[Epoch 169 Batch 150/162] avg loss 0.000733081, throughput 9.28735K wps
Begin Testing...
[Epoch 169] train avg loss 0.000771699, dev acc 0.9267, dev avg loss 0.210849, throughput 9.41279K wps
[Epoch 170 Batch 30/162] avg loss 0.000701067, throughput 9.40337K wps
[Epoch 170 Batch 60/162] avg loss 0.000608046, throughput 9.32043K wps
[Epoch 170 Batch 90/162] avg loss 0.000713109, throughput 9.43K wps
[Epoch 170 Batch 120/162] avg loss 0.000749368, throughput 9.51685K wps
[Epoch 170 Batch 150/162] avg loss 0.000788219, throughput 9.41057K wps
Begin Testing...
[Epoch 170] train avg loss 0.000709359, dev acc 0.9289, dev avg loss 0.211579, throughput 9.42061K wps
[Epoch 171 Batch 30/162] avg loss 0.000767763, throughput 9.56408K wps
[Epoch 171 Batch 60/162] avg loss 0.000659662, throughput 9.39395K wps
[Epoch 171 Batch 90/162] avg loss 0.000731328, throughput 9.46394K wps
[Epoch 171 Batch 120/162] avg loss 0.000754827, throughput 9.38139K wps
[Epoch 171 Batch 150/162] avg loss 0.000631687, throughput 9.4503K wps
Begin Testing...
[Epoch 171] train avg loss 0.000715866, dev acc 0.9267, dev avg loss 0.211792, throughput 9.44402K wps
[Epoch 172 Batch 30/162] avg loss 0.000788123, throughput 9.4978K wps
[Epoch 172 Batch 60/162] avg loss 0.00070615, throughput 9.31188K wps
[Epoch 172 Batch 90/162] avg loss 0.000687915, throughput 9.55517K wps
[Epoch 172 Batch 120/162] avg loss 0.000711886, throughput 9.228K wps
[Epoch 172 Batch 150/162] avg loss 0.000710511, throughput 9.26312K wps
Begin Testing...
[Epoch 172] train avg loss 0.000722513, dev acc 0.9244, dev avg loss 0.211368, throughput 9.36684K wps
[Epoch 173 Batch 30/162] avg loss 0.000621936, throughput 9.63745K wps
[Epoch 173 Batch 60/162] avg loss 0.000729872, throughput 9.30012K wps
[Epoch 173 Batch 90/162] avg loss 0.000677665, throughput 9.55323K wps
[Epoch 173 Batch 120/162] avg loss 0.000705925, throughput 9.50556K wps
[Epoch 173 Batch 150/162] avg loss 0.000757457, throughput 9.48975K wps
Begin Testing...
[Epoch 173] train avg loss 0.000690702, dev acc 0.9278, dev avg loss 0.212417, throughput 9.50256K wps
[Epoch 174 Batch 30/162] avg loss 0.000650361, throughput 9.785K wps
[Epoch 174 Batch 60/162] avg loss 0.00068233, throughput 9.21393K wps
[Epoch 174 Batch 90/162] avg loss 0.000724919, throughput 9.37205K wps
[Epoch 174 Batch 120/162] avg loss 0.000743209, throughput 9.31768K wps
[Epoch 174 Batch 150/162] avg loss 0.000691725, throughput 9.26703K wps
Begin Testing...
[Epoch 174] train avg loss 0.000697366, dev acc 0.9300, dev avg loss 0.214583, throughput 9.3838K wps
[Epoch 175 Batch 30/162] avg loss 0.000713723, throughput 9.57816K wps
[Epoch 175 Batch 60/162] avg loss 0.000767976, throughput 9.38262K wps
[Epoch 175 Batch 90/162] avg loss 0.000735441, throughput 9.43323K wps
[Epoch 175 Batch 120/162] avg loss 0.00065166, throughput 9.50879K wps
[Epoch 175 Batch 150/162] avg loss 0.000722867, throughput 9.2783K wps
Begin Testing...
[Epoch 175] train avg loss 0.000709878, dev acc 0.9311, dev avg loss 0.212932, throughput 9.4571K wps
[Epoch 176 Batch 30/162] avg loss 0.000733071, throughput 9.50096K wps
[Epoch 176 Batch 60/162] avg loss 0.000674443, throughput 9.44009K wps
[Epoch 176 Batch 90/162] avg loss 0.000728356, throughput 9.43785K wps
[Epoch 176 Batch 120/162] avg loss 0.000693816, throughput 9.57494K wps
[Epoch 176 Batch 150/162] avg loss 0.000620967, throughput 9.32863K wps
Begin Testing...
[Epoch 176] train avg loss 0.000695175, dev acc 0.9256, dev avg loss 0.211455, throughput 9.45538K wps
[Epoch 177 Batch 30/162] avg loss 0.000764278, throughput 9.51641K wps
[Epoch 177 Batch 60/162] avg loss 0.00063597, throughput 9.27038K wps
[Epoch 177 Batch 90/162] avg loss 0.000708808, throughput 9.55452K wps
[Epoch 177 Batch 120/162] avg loss 0.000694744, throughput 9.33968K wps
[Epoch 177 Batch 150/162] avg loss 0.000665785, throughput 9.35496K wps
Begin Testing...
[Epoch 177] train avg loss 0.000686991, dev acc 0.9311, dev avg loss 0.213967, throughput 9.40391K wps
[Epoch 178 Batch 30/162] avg loss 0.0006718, throughput 9.5395K wps
[Epoch 178 Batch 60/162] avg loss 0.000661736, throughput 9.35891K wps
[Epoch 178 Batch 90/162] avg loss 0.000749727, throughput 9.51239K wps
[Epoch 178 Batch 120/162] avg loss 0.000700749, throughput 9.43395K wps
[Epoch 178 Batch 150/162] avg loss 0.000621561, throughput 9.3119K wps
Begin Testing...
[Epoch 178] train avg loss 0.000684948, dev acc 0.9289, dev avg loss 0.212564, throughput 9.44204K wps
[Epoch 179 Batch 30/162] avg loss 0.000663017, throughput 9.63372K wps
[Epoch 179 Batch 60/162] avg loss 0.000666347, throughput 9.61487K wps
[Epoch 179 Batch 90/162] avg loss 0.000635737, throughput 9.45966K wps
[Epoch 179 Batch 120/162] avg loss 0.000661275, throughput 9.41749K wps
[Epoch 179 Batch 150/162] avg loss 0.000658978, throughput 9.37285K wps
Begin Testing...
[Epoch 179] train avg loss 0.000662383, dev acc 0.9278, dev avg loss 0.212473, throughput 9.50665K wps
[Epoch 180 Batch 30/162] avg loss 0.000637753, throughput 9.58036K wps
[Epoch 180 Batch 60/162] avg loss 0.000657706, throughput 9.51694K wps
[Epoch 180 Batch 90/162] avg loss 0.00064768, throughput 9.4089K wps
[Epoch 180 Batch 120/162] avg loss 0.00062582, throughput 9.53953K wps
[Epoch 180 Batch 150/162] avg loss 0.000647787, throughput 9.60374K wps
Begin Testing...
[Epoch 180] train avg loss 0.000642957, dev acc 0.9300, dev avg loss 0.214426, throughput 9.51277K wps
[Epoch 181 Batch 30/162] avg loss 0.000672703, throughput 9.63747K wps
[Epoch 181 Batch 60/162] avg loss 0.000555308, throughput 9.38713K wps
[Epoch 181 Batch 90/162] avg loss 0.000674902, throughput 9.38794K wps
[Epoch 181 Batch 120/162] avg loss 0.000753623, throughput 9.63968K wps
[Epoch 181 Batch 150/162] avg loss 0.000654547, throughput 9.39681K wps
Begin Testing...
[Epoch 181] train avg loss 0.000667244, dev acc 0.9300, dev avg loss 0.21533, throughput 9.49348K wps
[Epoch 182 Batch 30/162] avg loss 0.000616437, throughput 9.74595K wps
[Epoch 182 Batch 60/162] avg loss 0.000686848, throughput 9.58492K wps
[Epoch 182 Batch 90/162] avg loss 0.000610409, throughput 9.48064K wps
[Epoch 182 Batch 120/162] avg loss 0.000656857, throughput 9.53063K wps
[Epoch 182 Batch 150/162] avg loss 0.000765077, throughput 9.43315K wps
Begin Testing...
[Epoch 182] train avg loss 0.00065919, dev acc 0.9289, dev avg loss 0.213144, throughput 9.56769K wps
[Epoch 183 Batch 30/162] avg loss 0.000672605, throughput 9.84553K wps
[Epoch 183 Batch 60/162] avg loss 0.000742733, throughput 9.45568K wps
[Epoch 183 Batch 90/162] avg loss 0.000646728, throughput 9.57952K wps
[Epoch 183 Batch 120/162] avg loss 0.00064278, throughput 9.37548K wps
[Epoch 183 Batch 150/162] avg loss 0.00067401, throughput 9.38206K wps
Begin Testing...
[Epoch 183] train avg loss 0.00067619, dev acc 0.9244, dev avg loss 0.212656, throughput 9.50326K wps
[Epoch 184 Batch 30/162] avg loss 0.000633549, throughput 9.50594K wps
[Epoch 184 Batch 60/162] avg loss 0.000595861, throughput 9.4139K wps
[Epoch 184 Batch 90/162] avg loss 0.000696915, throughput 9.41755K wps
[Epoch 184 Batch 120/162] avg loss 0.00058606, throughput 9.45155K wps
[Epoch 184 Batch 150/162] avg loss 0.000786677, throughput 9.31722K wps
Begin Testing...
[Epoch 184] train avg loss 0.000655573, dev acc 0.9300, dev avg loss 0.215816, throughput 9.42873K wps
[Epoch 185 Batch 30/162] avg loss 0.000580663, throughput 9.6533K wps
[Epoch 185 Batch 60/162] avg loss 0.000724272, throughput 9.33917K wps
[Epoch 185 Batch 90/162] avg loss 0.000659099, throughput 9.43881K wps
[Epoch 185 Batch 120/162] avg loss 0.000532653, throughput 9.57378K wps
[Epoch 185 Batch 150/162] avg loss 0.000705388, throughput 9.37384K wps
Begin Testing...
[Epoch 185] train avg loss 0.000639377, dev acc 0.9333, dev avg loss 0.216384, throughput 9.47256K wps
Observed Improvement.
Begin Testing...
[Epoch 186 Batch 30/162] avg loss 0.000650346, throughput 9.6444K wps
[Epoch 186 Batch 60/162] avg loss 0.000619059, throughput 9.42988K wps
[Epoch 186 Batch 90/162] avg loss 0.000662279, throughput 9.55971K wps
[Epoch 186 Batch 120/162] avg loss 0.000619642, throughput 9.5253K wps
[Epoch 186 Batch 150/162] avg loss 0.000647278, throughput 9.42924K wps
Begin Testing...
[Epoch 186] train avg loss 0.000642925, dev acc 0.9278, dev avg loss 0.214248, throughput 9.51682K wps
[Epoch 187 Batch 30/162] avg loss 0.000599771, throughput 9.80778K wps
[Epoch 187 Batch 60/162] avg loss 0.00064741, throughput 9.50865K wps
[Epoch 187 Batch 90/162] avg loss 0.00059695, throughput 9.61843K wps
[Epoch 187 Batch 120/162] avg loss 0.000695354, throughput 9.30248K wps
[Epoch 187 Batch 150/162] avg loss 0.000531065, throughput 9.34733K wps
Begin Testing...
[Epoch 187] train avg loss 0.000608167, dev acc 0.9278, dev avg loss 0.214214, throughput 9.50627K wps
[Epoch 188 Batch 30/162] avg loss 0.000682783, throughput 9.7396K wps
[Epoch 188 Batch 60/162] avg loss 0.000704196, throughput 9.47002K wps
[Epoch 188 Batch 90/162] avg loss 0.000667785, throughput 9.29913K wps
[Epoch 188 Batch 120/162] avg loss 0.000684991, throughput 9.35816K wps
[Epoch 188 Batch 150/162] avg loss 0.000600866, throughput 9.43201K wps
Begin Testing...
[Epoch 188] train avg loss 0.000667922, dev acc 0.9267, dev avg loss 0.213364, throughput 9.45233K wps
[Epoch 189 Batch 30/162] avg loss 0.000670477, throughput 9.67909K wps
[Epoch 189 Batch 60/162] avg loss 0.000669962, throughput 9.38256K wps
[Epoch 189 Batch 90/162] avg loss 0.000576029, throughput 9.42598K wps
[Epoch 189 Batch 120/162] avg loss 0.000531804, throughput 9.53223K wps
[Epoch 189 Batch 150/162] avg loss 0.000570375, throughput 9.39953K wps
Begin Testing...
[Epoch 189] train avg loss 0.000608752, dev acc 0.9278, dev avg loss 0.213787, throughput 9.48878K wps
[Epoch 190 Batch 30/162] avg loss 0.000639585, throughput 9.61631K wps
[Epoch 190 Batch 60/162] avg loss 0.000628595, throughput 9.55332K wps
[Epoch 190 Batch 90/162] avg loss 0.000625134, throughput 9.50199K wps
[Epoch 190 Batch 120/162] avg loss 0.000631411, throughput 9.38946K wps
[Epoch 190 Batch 150/162] avg loss 0.000619457, throughput 9.2976K wps
Begin Testing...
[Epoch 190] train avg loss 0.000632243, dev acc 0.9256, dev avg loss 0.213457, throughput 9.47133K wps
[Epoch 191 Batch 30/162] avg loss 0.000610998, throughput 9.85661K wps
[Epoch 191 Batch 60/162] avg loss 0.000616073, throughput 9.56251K wps
[Epoch 191 Batch 90/162] avg loss 0.000665106, throughput 9.41684K wps
[Epoch 191 Batch 120/162] avg loss 0.000656122, throughput 9.53455K wps
[Epoch 191 Batch 150/162] avg loss 0.000612434, throughput 9.38199K wps
Begin Testing...
[Epoch 191] train avg loss 0.000633959, dev acc 0.9256, dev avg loss 0.213583, throughput 9.51713K wps
[Epoch 192 Batch 30/162] avg loss 0.000594883, throughput 9.60736K wps
[Epoch 192 Batch 60/162] avg loss 0.000628445, throughput 9.36835K wps
[Epoch 192 Batch 90/162] avg loss 0.000619293, throughput 9.39803K wps
[Epoch 192 Batch 120/162] avg loss 0.000601872, throughput 9.53465K wps
[Epoch 192 Batch 150/162] avg loss 0.000592265, throughput 9.34482K wps
Begin Testing...
[Epoch 192] train avg loss 0.00060202, dev acc 0.9289, dev avg loss 0.214972, throughput 9.43321K wps
[Epoch 193 Batch 30/162] avg loss 0.000608482, throughput 9.52484K wps
[Epoch 193 Batch 60/162] avg loss 0.00061439, throughput 9.61417K wps
[Epoch 193 Batch 90/162] avg loss 0.000595505, throughput 9.38707K wps
[Epoch 193 Batch 120/162] avg loss 0.000638385, throughput 9.3687K wps
[Epoch 193 Batch 150/162] avg loss 0.000612881, throughput 9.45881K wps
Begin Testing...
[Epoch 193] train avg loss 0.000613987, dev acc 0.9289, dev avg loss 0.215258, throughput 9.45975K wps
[Epoch 194 Batch 30/162] avg loss 0.000615531, throughput 9.51574K wps
[Epoch 194 Batch 60/162] avg loss 0.000608506, throughput 9.41543K wps
[Epoch 194 Batch 90/162] avg loss 0.00063444, throughput 9.48629K wps
[Epoch 194 Batch 120/162] avg loss 0.000622747, throughput 9.35317K wps
[Epoch 194 Batch 150/162] avg loss 0.000665975, throughput 9.622K wps
Begin Testing...
[Epoch 194] train avg loss 0.00063014, dev acc 0.9267, dev avg loss 0.214462, throughput 9.46296K wps
[Epoch 195 Batch 30/162] avg loss 0.000652382, throughput 9.52073K wps
[Epoch 195 Batch 60/162] avg loss 0.000532771, throughput 9.61387K wps
[Epoch 195 Batch 90/162] avg loss 0.000605353, throughput 9.54039K wps
[Epoch 195 Batch 120/162] avg loss 0.000601139, throughput 9.31905K wps
[Epoch 195 Batch 150/162] avg loss 0.000541415, throughput 9.39632K wps
Begin Testing...
[Epoch 195] train avg loss 0.000583237, dev acc 0.9267, dev avg loss 0.215054, throughput 9.46826K wps
[Epoch 196 Batch 30/162] avg loss 0.000564425, throughput 9.69749K wps
[Epoch 196 Batch 60/162] avg loss 0.000576881, throughput 8.47535K wps
[Epoch 196 Batch 90/162] avg loss 0.000569951, throughput 9.47385K wps
[Epoch 196 Batch 120/162] avg loss 0.000584691, throughput 9.43181K wps
[Epoch 196 Batch 150/162] avg loss 0.000600867, throughput 9.45184K wps
Begin Testing...
[Epoch 196] train avg loss 0.000592074, dev acc 0.9300, dev avg loss 0.216933, throughput 9.2961K wps
[Epoch 197 Batch 30/162] avg loss 0.000593346, throughput 9.50722K wps
[Epoch 197 Batch 60/162] avg loss 0.000610816, throughput 9.3231K wps
[Epoch 197 Batch 90/162] avg loss 0.000581374, throughput 9.3243K wps
[Epoch 197 Batch 120/162] avg loss 0.000570871, throughput 9.34086K wps
[Epoch 197 Batch 150/162] avg loss 0.000564062, throughput 9.32971K wps
Begin Testing...
[Epoch 197] train avg loss 0.000586054, dev acc 0.9289, dev avg loss 0.216824, throughput 9.36766K wps
[Epoch 198 Batch 30/162] avg loss 0.000566127, throughput 9.69731K wps
[Epoch 198 Batch 60/162] avg loss 0.000594536, throughput 9.42669K wps
[Epoch 198 Batch 90/162] avg loss 0.00054169, throughput 9.5459K wps
[Epoch 198 Batch 120/162] avg loss 0.000630613, throughput 9.61596K wps
[Epoch 198 Batch 150/162] avg loss 0.000598119, throughput 9.5651K wps
Begin Testing...
[Epoch 198] train avg loss 0.00058499, dev acc 0.9289, dev avg loss 0.215713, throughput 9.56247K wps
[Epoch 199 Batch 30/162] avg loss 0.00065372, throughput 9.49196K wps
[Epoch 199 Batch 60/162] avg loss 0.00053798, throughput 9.53689K wps
[Epoch 199 Batch 90/162] avg loss 0.000609997, throughput 9.41142K wps
[Epoch 199 Batch 120/162] avg loss 0.000558827, throughput 9.31306K wps
[Epoch 199 Batch 150/162] avg loss 0.000620211, throughput 9.37389K wps
Begin Testing...
[Epoch 199] train avg loss 0.00059247, dev acc 0.9289, dev avg loss 0.217183, throughput 9.41815K wps
Test loss 0.239984, test acc 0.9090
Total time cost 443.80s
[Epoch 0 Batch 30/162] avg loss 0.0139825, throughput 7.49568K wps
[Epoch 0 Batch 60/162] avg loss 0.0139651, throughput 9.45568K wps
[Epoch 0 Batch 90/162] avg loss 0.0137258, throughput 9.27212K wps
[Epoch 0 Batch 120/162] avg loss 0.0134256, throughput 9.39426K wps
[Epoch 0 Batch 150/162] avg loss 0.0134632, throughput 9.46151K wps
Begin Testing...
[Epoch 0] train avg loss 0.0136888, dev acc 0.7711, dev avg loss 0.661879, throughput 8.97186K wps
Observed Improvement.
Begin Testing...
[Epoch 1 Batch 30/162] avg loss 0.0131691, throughput 9.46866K wps
[Epoch 1 Batch 60/162] avg loss 0.0130881, throughput 9.32578K wps
[Epoch 1 Batch 90/162] avg loss 0.0130308, throughput 9.32991K wps
[Epoch 1 Batch 120/162] avg loss 0.0128969, throughput 9.42728K wps
[Epoch 1 Batch 150/162] avg loss 0.0129661, throughput 9.38434K wps
Begin Testing...
[Epoch 1] train avg loss 0.0130119, dev acc 0.8444, dev avg loss 0.629765, throughput 9.40809K wps
Observed Improvement.
Begin Testing...
[Epoch 2 Batch 30/162] avg loss 0.0124943, throughput 9.707K wps
[Epoch 2 Batch 60/162] avg loss 0.0126056, throughput 9.29032K wps
[Epoch 2 Batch 90/162] avg loss 0.0122539, throughput 9.57191K wps
[Epoch 2 Batch 120/162] avg loss 0.012186, throughput 9.57154K wps
[Epoch 2 Batch 150/162] avg loss 0.0119895, throughput 9.28012K wps
Begin Testing...
[Epoch 2] train avg loss 0.0122994, dev acc 0.8500, dev avg loss 0.595159, throughput 9.47594K wps
Observed Improvement.
Begin Testing...
[Epoch 3 Batch 30/162] avg loss 0.0119634, throughput 9.68522K wps
[Epoch 3 Batch 60/162] avg loss 0.0117744, throughput 9.27191K wps
[Epoch 3 Batch 90/162] avg loss 0.0117152, throughput 9.48255K wps
[Epoch 3 Batch 120/162] avg loss 0.0113713, throughput 9.39051K wps
[Epoch 3 Batch 150/162] avg loss 0.0113462, throughput 9.42604K wps
Begin Testing...
[Epoch 3] train avg loss 0.0116014, dev acc 0.8778, dev avg loss 0.554865, throughput 9.44573K wps
Observed Improvement.
Begin Testing...
[Epoch 4 Batch 30/162] avg loss 0.011099, throughput 9.56813K wps
[Epoch 4 Batch 60/162] avg loss 0.0109573, throughput 9.44231K wps
[Epoch 4 Batch 90/162] avg loss 0.0107478, throughput 9.44103K wps
[Epoch 4 Batch 120/162] avg loss 0.0106736, throughput 9.40841K wps
[Epoch 4 Batch 150/162] avg loss 0.0105485, throughput 9.4218K wps
Begin Testing...
[Epoch 4] train avg loss 0.0107648, dev acc 0.8744, dev avg loss 0.512721, throughput 9.46314K wps
[Epoch 5 Batch 30/162] avg loss 0.0103198, throughput 9.83499K wps
[Epoch 5 Batch 60/162] avg loss 0.0100045, throughput 9.25673K wps
[Epoch 5 Batch 90/162] avg loss 0.00991426, throughput 9.53359K wps
[Epoch 5 Batch 120/162] avg loss 0.00969016, throughput 9.53257K wps
[Epoch 5 Batch 150/162] avg loss 0.00990291, throughput 9.42876K wps
Begin Testing...
[Epoch 5] train avg loss 0.00993815, dev acc 0.8689, dev avg loss 0.473028, throughput 9.5068K wps
[Epoch 6 Batch 30/162] avg loss 0.00955886, throughput 9.50377K wps
[Epoch 6 Batch 60/162] avg loss 0.00928337, throughput 9.48807K wps
[Epoch 6 Batch 90/162] avg loss 0.00910475, throughput 9.46616K wps
[Epoch 6 Batch 120/162] avg loss 0.00892793, throughput 9.34813K wps
[Epoch 6 Batch 150/162] avg loss 0.00896287, throughput 9.57356K wps
Begin Testing...
[Epoch 6] train avg loss 0.00916664, dev acc 0.8711, dev avg loss 0.437, throughput 9.44691K wps
[Epoch 7 Batch 30/162] avg loss 0.00874671, throughput 9.53498K wps
[Epoch 7 Batch 60/162] avg loss 0.00879596, throughput 9.29917K wps
[Epoch 7 Batch 90/162] avg loss 0.00842875, throughput 9.47195K wps
[Epoch 7 Batch 120/162] avg loss 0.00823087, throughput 9.62861K wps
[Epoch 7 Batch 150/162] avg loss 0.00840739, throughput 9.32881K wps
Begin Testing...
[Epoch 7] train avg loss 0.00852621, dev acc 0.8744, dev avg loss 0.406936, throughput 9.43931K wps
[Epoch 8 Batch 30/162] avg loss 0.0082697, throughput 9.63425K wps
[Epoch 8 Batch 60/162] avg loss 0.00806563, throughput 9.32298K wps
[Epoch 8 Batch 90/162] avg loss 0.00794349, throughput 9.57906K wps
[Epoch 8 Batch 120/162] avg loss 0.00776054, throughput 9.48384K wps
[Epoch 8 Batch 150/162] avg loss 0.00776879, throughput 9.60791K wps
Begin Testing...
[Epoch 8] train avg loss 0.00795234, dev acc 0.8700, dev avg loss 0.383764, throughput 9.49578K wps
[Epoch 9 Batch 30/162] avg loss 0.00772339, throughput 9.78151K wps
[Epoch 9 Batch 60/162] avg loss 0.00758351, throughput 9.28226K wps
[Epoch 9 Batch 90/162] avg loss 0.00761131, throughput 9.33112K wps
[Epoch 9 Batch 120/162] avg loss 0.007647, throughput 9.41242K wps
[Epoch 9 Batch 150/162] avg loss 0.00740406, throughput 9.37446K wps
Begin Testing...
[Epoch 9] train avg loss 0.00755243, dev acc 0.8744, dev avg loss 0.365921, throughput 9.42955K wps
[Epoch 10 Batch 30/162] avg loss 0.00700638, throughput 9.81318K wps
[Epoch 10 Batch 60/162] avg loss 0.00734812, throughput 9.39867K wps
[Epoch 10 Batch 90/162] avg loss 0.00707648, throughput 9.57023K wps
[Epoch 10 Batch 120/162] avg loss 0.00708918, throughput 9.34713K wps
[Epoch 10 Batch 150/162] avg loss 0.00694796, throughput 9.41912K wps
Begin Testing...
[Epoch 10] train avg loss 0.00708014, dev acc 0.8789, dev avg loss 0.35189, throughput 9.47399K wps
Observed Improvement.
Begin Testing...
[Epoch 11 Batch 30/162] avg loss 0.00678814, throughput 9.69034K wps
[Epoch 11 Batch 60/162] avg loss 0.00685432, throughput 9.25433K wps
[Epoch 11 Batch 90/162] avg loss 0.00692997, throughput 9.34382K wps
[Epoch 11 Batch 120/162] avg loss 0.00665797, throughput 9.5126K wps
[Epoch 11 Batch 150/162] avg loss 0.00713154, throughput 9.47836K wps
Begin Testing...
[Epoch 11] train avg loss 0.00684713, dev acc 0.8789, dev avg loss 0.338836, throughput 9.46128K wps
Observed Improvement.
Begin Testing...
[Epoch 12 Batch 30/162] avg loss 0.00638532, throughput 9.61886K wps
[Epoch 12 Batch 60/162] avg loss 0.0068088, throughput 9.50748K wps
[Epoch 12 Batch 90/162] avg loss 0.00655689, throughput 9.41578K wps
[Epoch 12 Batch 120/162] avg loss 0.00661975, throughput 9.57296K wps
[Epoch 12 Batch 150/162] avg loss 0.00695985, throughput 9.42294K wps
Begin Testing...
[Epoch 12] train avg loss 0.00664812, dev acc 0.8811, dev avg loss 0.329401, throughput 9.48713K wps
Observed Improvement.
Begin Testing...
[Epoch 13 Batch 30/162] avg loss 0.00636439, throughput 9.45308K wps
[Epoch 13 Batch 60/162] avg loss 0.00613926, throughput 9.33687K wps
[Epoch 13 Batch 90/162] avg loss 0.00629012, throughput 9.46292K wps
[Epoch 13 Batch 120/162] avg loss 0.00650297, throughput 9.61223K wps
[Epoch 13 Batch 150/162] avg loss 0.00625137, throughput 9.55976K wps
Begin Testing...
[Epoch 13] train avg loss 0.00634779, dev acc 0.8856, dev avg loss 0.321806, throughput 9.48035K wps
Observed Improvement.
Begin Testing...
[Epoch 14 Batch 30/162] avg loss 0.00636438, throughput 9.51257K wps
[Epoch 14 Batch 60/162] avg loss 0.00661898, throughput 9.43619K wps
[Epoch 14 Batch 90/162] avg loss 0.00604907, throughput 9.56816K wps
[Epoch 14 Batch 120/162] avg loss 0.00593193, throughput 9.2529K wps
[Epoch 14 Batch 150/162] avg loss 0.00647751, throughput 9.55003K wps
Begin Testing...
[Epoch 14] train avg loss 0.00630644, dev acc 0.8856, dev avg loss 0.31409, throughput 9.48076K wps
Observed Improvement.
Begin Testing...
[Epoch 15 Batch 30/162] avg loss 0.00589865, throughput 9.75357K wps
[Epoch 15 Batch 60/162] avg loss 0.00594977, throughput 9.39137K wps
[Epoch 15 Batch 90/162] avg loss 0.00643593, throughput 9.67226K wps
[Epoch 15 Batch 120/162] avg loss 0.00610794, throughput 9.44808K wps
[Epoch 15 Batch 150/162] avg loss 0.00615393, throughput 9.47335K wps
Begin Testing...
[Epoch 15] train avg loss 0.00609366, dev acc 0.8867, dev avg loss 0.308299, throughput 9.54036K wps
Observed Improvement.
Begin Testing...
[Epoch 16 Batch 30/162] avg loss 0.0058601, throughput 9.51494K wps
[Epoch 16 Batch 60/162] avg loss 0.00574209, throughput 9.24332K wps
[Epoch 16 Batch 90/162] avg loss 0.00590704, throughput 9.23454K wps
[Epoch 16 Batch 120/162] avg loss 0.00639577, throughput 9.42839K wps
[Epoch 16 Batch 150/162] avg loss 0.00623476, throughput 9.30049K wps
Begin Testing...
[Epoch 16] train avg loss 0.00599568, dev acc 0.8867, dev avg loss 0.302604, throughput 9.33867K wps
Observed Improvement.
Begin Testing...
[Epoch 17 Batch 30/162] avg loss 0.00610861, throughput 9.75514K wps
[Epoch 17 Batch 60/162] avg loss 0.00553997, throughput 9.37239K wps
[Epoch 17 Batch 90/162] avg loss 0.00583785, throughput 9.53448K wps
[Epoch 17 Batch 120/162] avg loss 0.00589014, throughput 9.42792K wps
[Epoch 17 Batch 150/162] avg loss 0.00553139, throughput 9.30895K wps
Begin Testing...
[Epoch 17] train avg loss 0.00579106, dev acc 0.8856, dev avg loss 0.298033, throughput 9.46292K wps
[Epoch 18 Batch 30/162] avg loss 0.00560322, throughput 9.81478K wps
[Epoch 18 Batch 60/162] avg loss 0.0055875, throughput 9.41633K wps
[Epoch 18 Batch 90/162] avg loss 0.00551438, throughput 9.54682K wps
[Epoch 18 Batch 120/162] avg loss 0.00578135, throughput 9.60124K wps
[Epoch 18 Batch 150/162] avg loss 0.00561666, throughput 9.41518K wps
Begin Testing...
[Epoch 18] train avg loss 0.00565916, dev acc 0.8867, dev avg loss 0.293867, throughput 9.55434K wps
Observed Improvement.
Begin Testing...
[Epoch 19 Batch 30/162] avg loss 0.00532398, throughput 9.62494K wps
[Epoch 19 Batch 60/162] avg loss 0.00561211, throughput 9.30376K wps
[Epoch 19 Batch 90/162] avg loss 0.00558686, throughput 9.52264K wps
[Epoch 19 Batch 120/162] avg loss 0.00547382, throughput 9.37215K wps
[Epoch 19 Batch 150/162] avg loss 0.0056365, throughput 9.43533K wps
Begin Testing...
[Epoch 19] train avg loss 0.00555668, dev acc 0.8878, dev avg loss 0.288584, throughput 9.43819K wps
Observed Improvement.
Begin Testing...
[Epoch 20 Batch 30/162] avg loss 0.00549226, throughput 9.68109K wps
[Epoch 20 Batch 60/162] avg loss 0.0053993, throughput 9.2231K wps
[Epoch 20 Batch 90/162] avg loss 0.00500895, throughput 9.34271K wps
[Epoch 20 Batch 120/162] avg loss 0.00539971, throughput 9.39154K wps
[Epoch 20 Batch 150/162] avg loss 0.00550359, throughput 9.36563K wps
Begin Testing...
[Epoch 20] train avg loss 0.00536894, dev acc 0.8889, dev avg loss 0.286125, throughput 9.39952K wps
Observed Improvement.
Begin Testing...
[Epoch 21 Batch 30/162] avg loss 0.00529462, throughput 9.65366K wps
[Epoch 21 Batch 60/162] avg loss 0.00527439, throughput 9.41792K wps
[Epoch 21 Batch 90/162] avg loss 0.00529978, throughput 9.42731K wps
[Epoch 21 Batch 120/162] avg loss 0.00509908, throughput 9.39327K wps
[Epoch 21 Batch 150/162] avg loss 0.0054078, throughput 9.54949K wps
Begin Testing...
[Epoch 21] train avg loss 0.0052816, dev acc 0.8889, dev avg loss 0.28378, throughput 9.47776K wps
Observed Improvement.
Begin Testing...
[Epoch 22 Batch 30/162] avg loss 0.00513871, throughput 9.85171K wps
[Epoch 22 Batch 60/162] avg loss 0.00512184, throughput 9.53841K wps
[Epoch 22 Batch 90/162] avg loss 0.00485204, throughput 9.45253K wps
[Epoch 22 Batch 120/162] avg loss 0.00506282, throughput 9.55082K wps
[Epoch 22 Batch 150/162] avg loss 0.00534495, throughput 9.37351K wps
Begin Testing...
[Epoch 22] train avg loss 0.00515489, dev acc 0.8911, dev avg loss 0.278044, throughput 9.54811K wps
Observed Improvement.
Begin Testing...
[Epoch 23 Batch 30/162] avg loss 0.00533986, throughput 9.37403K wps
[Epoch 23 Batch 60/162] avg loss 0.00485363, throughput 9.59638K wps
[Epoch 23 Batch 90/162] avg loss 0.00528768, throughput 9.34739K wps
[Epoch 23 Batch 120/162] avg loss 0.00516451, throughput 9.40663K wps
[Epoch 23 Batch 150/162] avg loss 0.00490921, throughput 9.60681K wps
Begin Testing...
[Epoch 23] train avg loss 0.00511814, dev acc 0.8900, dev avg loss 0.275524, throughput 9.44133K wps
[Epoch 24 Batch 30/162] avg loss 0.00484805, throughput 9.50491K wps
[Epoch 24 Batch 60/162] avg loss 0.00509497, throughput 9.21635K wps
[Epoch 24 Batch 90/162] avg loss 0.00515729, throughput 9.34863K wps
[Epoch 24 Batch 120/162] avg loss 0.00516947, throughput 9.52362K wps
[Epoch 24 Batch 150/162] avg loss 0.00505064, throughput 9.28155K wps
Begin Testing...
[Epoch 24] train avg loss 0.00503716, dev acc 0.8956, dev avg loss 0.273401, throughput 9.38515K wps
Observed Improvement.
Begin Testing...
[Epoch 25 Batch 30/162] avg loss 0.00494949, throughput 9.48114K wps
[Epoch 25 Batch 60/162] avg loss 0.00500692, throughput 9.45946K wps
[Epoch 25 Batch 90/162] avg loss 0.00483521, throughput 9.43769K wps
[Epoch 25 Batch 120/162] avg loss 0.00488613, throughput 9.28741K wps
[Epoch 25 Batch 150/162] avg loss 0.00454593, throughput 9.28422K wps
Begin Testing...
[Epoch 25] train avg loss 0.00486793, dev acc 0.8967, dev avg loss 0.271452, throughput 9.40458K wps
Observed Improvement.
Begin Testing...
[Epoch 26 Batch 30/162] avg loss 0.00469414, throughput 9.68995K wps
[Epoch 26 Batch 60/162] avg loss 0.00477638, throughput 9.37909K wps
[Epoch 26 Batch 90/162] avg loss 0.00481066, throughput 9.55688K wps
[Epoch 26 Batch 120/162] avg loss 0.00507507, throughput 9.52388K wps
[Epoch 26 Batch 150/162] avg loss 0.00467645, throughput 9.29694K wps
Begin Testing...
[Epoch 26] train avg loss 0.00481303, dev acc 0.8922, dev avg loss 0.267085, throughput 9.4951K wps
[Epoch 27 Batch 30/162] avg loss 0.00485451, throughput 9.5806K wps
[Epoch 27 Batch 60/162] avg loss 0.00507631, throughput 9.36162K wps
[Epoch 27 Batch 90/162] avg loss 0.00516835, throughput 9.34375K wps
[Epoch 27 Batch 120/162] avg loss 0.00464131, throughput 9.51578K wps
[Epoch 27 Batch 150/162] avg loss 0.00457097, throughput 9.40366K wps
Begin Testing...
[Epoch 27] train avg loss 0.00480405, dev acc 0.8922, dev avg loss 0.265026, throughput 9.43954K wps
[Epoch 28 Batch 30/162] avg loss 0.00447128, throughput 9.65968K wps
[Epoch 28 Batch 60/162] avg loss 0.00477829, throughput 9.38523K wps
[Epoch 28 Batch 90/162] avg loss 0.00497192, throughput 9.63155K wps
[Epoch 28 Batch 120/162] avg loss 0.00504486, throughput 9.51035K wps
[Epoch 28 Batch 150/162] avg loss 0.00425027, throughput 9.44191K wps
Begin Testing...
[Epoch 28] train avg loss 0.00473219, dev acc 0.8956, dev avg loss 0.263031, throughput 9.50015K wps
[Epoch 29 Batch 30/162] avg loss 0.00454172, throughput 9.64833K wps
[Epoch 29 Batch 60/162] avg loss 0.00466184, throughput 9.23645K wps
[Epoch 29 Batch 90/162] avg loss 0.00439524, throughput 9.33389K wps
[Epoch 29 Batch 120/162] avg loss 0.00460785, throughput 9.58203K wps
[Epoch 29 Batch 150/162] avg loss 0.00487905, throughput 9.4848K wps
Begin Testing...
[Epoch 29] train avg loss 0.00460513, dev acc 0.8967, dev avg loss 0.261012, throughput 9.45248K wps
Observed Improvement.
Begin Testing...
[Epoch 30 Batch 30/162] avg loss 0.00433349, throughput 9.59143K wps
[Epoch 30 Batch 60/162] avg loss 0.00450857, throughput 9.38907K wps
[Epoch 30 Batch 90/162] avg loss 0.00449303, throughput 9.55036K wps
[Epoch 30 Batch 120/162] avg loss 0.00478618, throughput 9.36993K wps
[Epoch 30 Batch 150/162] avg loss 0.00439054, throughput 9.50454K wps
Begin Testing...
[Epoch 30] train avg loss 0.0045078, dev acc 0.8989, dev avg loss 0.259309, throughput 9.46403K wps
Observed Improvement.
Begin Testing...
[Epoch 31 Batch 30/162] avg loss 0.00437112, throughput 9.67787K wps
[Epoch 31 Batch 60/162] avg loss 0.00461422, throughput 9.39901K wps
[Epoch 31 Batch 90/162] avg loss 0.00416385, throughput 9.3712K wps
[Epoch 31 Batch 120/162] avg loss 0.00456719, throughput 9.41549K wps
[Epoch 31 Batch 150/162] avg loss 0.00429625, throughput 9.39842K wps
Begin Testing...
[Epoch 31] train avg loss 0.00436447, dev acc 0.8956, dev avg loss 0.257309, throughput 9.46284K wps
[Epoch 32 Batch 30/162] avg loss 0.00420146, throughput 9.4407K wps
[Epoch 32 Batch 60/162] avg loss 0.0045089, throughput 9.41183K wps
[Epoch 32 Batch 90/162] avg loss 0.00421182, throughput 9.49974K wps
[Epoch 32 Batch 120/162] avg loss 0.0041965, throughput 9.25489K wps
[Epoch 32 Batch 150/162] avg loss 0.00434442, throughput 9.39498K wps
Begin Testing...
[Epoch 32] train avg loss 0.00429094, dev acc 0.9000, dev avg loss 0.255726, throughput 9.39431K wps
Observed Improvement.
Begin Testing...
[Epoch 33 Batch 30/162] avg loss 0.0045111, throughput 9.57958K wps
[Epoch 33 Batch 60/162] avg loss 0.00390907, throughput 9.59449K wps
[Epoch 33 Batch 90/162] avg loss 0.00463363, throughput 9.52117K wps
[Epoch 33 Batch 120/162] avg loss 0.00399018, throughput 9.52201K wps
[Epoch 33 Batch 150/162] avg loss 0.00448386, throughput 9.51103K wps
Begin Testing...
[Epoch 33] train avg loss 0.00428787, dev acc 0.8989, dev avg loss 0.253939, throughput 9.54755K wps
[Epoch 34 Batch 30/162] avg loss 0.00427952, throughput 9.57592K wps
[Epoch 34 Batch 60/162] avg loss 0.00433383, throughput 9.45584K wps
[Epoch 34 Batch 90/162] avg loss 0.0044711, throughput 9.38643K wps
[Epoch 34 Batch 120/162] avg loss 0.00416733, throughput 9.46268K wps
[Epoch 34 Batch 150/162] avg loss 0.00399448, throughput 9.44453K wps
Begin Testing...
[Epoch 34] train avg loss 0.00423558, dev acc 0.9000, dev avg loss 0.252395, throughput 9.45075K wps
Observed Improvement.
Begin Testing...
[Epoch 35 Batch 30/162] avg loss 0.00419712, throughput 9.65921K wps
[Epoch 35 Batch 60/162] avg loss 0.00381078, throughput 9.37006K wps
[Epoch 35 Batch 90/162] avg loss 0.00433777, throughput 9.20708K wps
[Epoch 35 Batch 120/162] avg loss 0.00408114, throughput 9.47532K wps
[Epoch 35 Batch 150/162] avg loss 0.0041096, throughput 9.36738K wps
Begin Testing...
[Epoch 35] train avg loss 0.00412409, dev acc 0.9000, dev avg loss 0.252585, throughput 9.4217K wps
Observed Improvement.
Begin Testing...
[Epoch 36 Batch 30/162] avg loss 0.00422937, throughput 9.643K wps
[Epoch 36 Batch 60/162] avg loss 0.00414452, throughput 9.54626K wps
[Epoch 36 Batch 90/162] avg loss 0.00421936, throughput 9.31018K wps
[Epoch 36 Batch 120/162] avg loss 0.00389043, throughput 9.40739K wps
[Epoch 36 Batch 150/162] avg loss 0.00387, throughput 9.5888K wps
Begin Testing...
[Epoch 36] train avg loss 0.00404626, dev acc 0.9033, dev avg loss 0.250451, throughput 9.49592K wps
Observed Improvement.
Begin Testing...
[Epoch 37 Batch 30/162] avg loss 0.00400134, throughput 9.60234K wps
[Epoch 37 Batch 60/162] avg loss 0.00412893, throughput 9.2961K wps
[Epoch 37 Batch 90/162] avg loss 0.00404126, throughput 9.4004K wps
[Epoch 37 Batch 120/162] avg loss 0.00394111, throughput 9.4628K wps
[Epoch 37 Batch 150/162] avg loss 0.00404071, throughput 9.35582K wps
Begin Testing...
[Epoch 37] train avg loss 0.00402468, dev acc 0.9033, dev avg loss 0.250145, throughput 9.41519K wps
Observed Improvement.
Begin Testing...
[Epoch 38 Batch 30/162] avg loss 0.00396422, throughput 9.42721K wps
[Epoch 38 Batch 60/162] avg loss 0.00385977, throughput 9.44969K wps
[Epoch 38 Batch 90/162] avg loss 0.00371367, throughput 9.43732K wps
[Epoch 38 Batch 120/162] avg loss 0.00393169, throughput 9.30524K wps
[Epoch 38 Batch 150/162] avg loss 0.00394827, throughput 9.5173K wps
Begin Testing...
[Epoch 38] train avg loss 0.00391715, dev acc 0.8989, dev avg loss 0.249179, throughput 9.44386K wps
[Epoch 39 Batch 30/162] avg loss 0.00403767, throughput 9.59589K wps
[Epoch 39 Batch 60/162] avg loss 0.00346819, throughput 9.2794K wps
[Epoch 39 Batch 90/162] avg loss 0.00378384, throughput 9.40224K wps
[Epoch 39 Batch 120/162] avg loss 0.00416939, throughput 9.30031K wps
[Epoch 39 Batch 150/162] avg loss 0.00393703, throughput 9.60249K wps
Begin Testing...
[Epoch 39] train avg loss 0.00387017, dev acc 0.9044, dev avg loss 0.245716, throughput 9.44634K wps
Observed Improvement.
Begin Testing...
[Epoch 40 Batch 30/162] avg loss 0.00370537, throughput 9.81247K wps
[Epoch 40 Batch 60/162] avg loss 0.00389293, throughput 9.38896K wps
[Epoch 40 Batch 90/162] avg loss 0.00398956, throughput 9.54858K wps
[Epoch 40 Batch 120/162] avg loss 0.00386417, throughput 9.59153K wps
[Epoch 40 Batch 150/162] avg loss 0.00389862, throughput 9.57753K wps
Begin Testing...
[Epoch 40] train avg loss 0.00387233, dev acc 0.9044, dev avg loss 0.245867, throughput 9.55984K wps
Observed Improvement.
Begin Testing...
[Epoch 41 Batch 30/162] avg loss 0.00379977, throughput 9.48701K wps
[Epoch 41 Batch 60/162] avg loss 0.00401791, throughput 9.55183K wps
[Epoch 41 Batch 90/162] avg loss 0.00366033, throughput 9.51701K wps
[Epoch 41 Batch 120/162] avg loss 0.00381382, throughput 9.39403K wps
[Epoch 41 Batch 150/162] avg loss 0.0035558, throughput 9.34592K wps
Begin Testing...
[Epoch 41] train avg loss 0.00374727, dev acc 0.9044, dev avg loss 0.243624, throughput 9.47396K wps
Observed Improvement.
Begin Testing...
[Epoch 42 Batch 30/162] avg loss 0.00372216, throughput 9.53202K wps
[Epoch 42 Batch 60/162] avg loss 0.00350659, throughput 9.45383K wps
[Epoch 42 Batch 90/162] avg loss 0.0038096, throughput 9.30459K wps
[Epoch 42 Batch 120/162] avg loss 0.00374028, throughput 9.44914K wps
[Epoch 42 Batch 150/162] avg loss 0.00368857, throughput 9.52651K wps
Begin Testing...
[Epoch 42] train avg loss 0.00367876, dev acc 0.9044, dev avg loss 0.244026, throughput 9.46508K wps
Observed Improvement.
Begin Testing...
[Epoch 43 Batch 30/162] avg loss 0.0033733, throughput 9.65816K wps
[Epoch 43 Batch 60/162] avg loss 0.00356203, throughput 9.34764K wps
[Epoch 43 Batch 90/162] avg loss 0.0038331, throughput 9.40315K wps
[Epoch 43 Batch 120/162] avg loss 0.00366062, throughput 9.59496K wps
[Epoch 43 Batch 150/162] avg loss 0.00396006, throughput 9.45607K wps
Begin Testing...
[Epoch 43] train avg loss 0.00367319, dev acc 0.9044, dev avg loss 0.242102, throughput 9.48128K wps
Observed Improvement.
Begin Testing...
[Epoch 44 Batch 30/162] avg loss 0.00392299, throughput 9.63738K wps
[Epoch 44 Batch 60/162] avg loss 0.0032435, throughput 9.46274K wps
[Epoch 44 Batch 90/162] avg loss 0.003735, throughput 9.5543K wps
[Epoch 44 Batch 120/162] avg loss 0.00369374, throughput 9.44834K wps
[Epoch 44 Batch 150/162] avg loss 0.00356028, throughput 9.5557K wps
Begin Testing...
[Epoch 44] train avg loss 0.00362115, dev acc 0.9044, dev avg loss 0.241104, throughput 9.51832K wps
Observed Improvement.
Begin Testing...
[Epoch 45 Batch 30/162] avg loss 0.00352714, throughput 9.38143K wps
[Epoch 45 Batch 60/162] avg loss 0.00341902, throughput 9.44006K wps
[Epoch 45 Batch 90/162] avg loss 0.00377772, throughput 9.3354K wps
[Epoch 45 Batch 120/162] avg loss 0.00339493, throughput 9.44309K wps
[Epoch 45 Batch 150/162] avg loss 0.00366732, throughput 9.42469K wps
Begin Testing...
[Epoch 45] train avg loss 0.00354357, dev acc 0.9033, dev avg loss 0.240778, throughput 9.39922K wps
[Epoch 46 Batch 30/162] avg loss 0.00368652, throughput 9.72567K wps
[Epoch 46 Batch 60/162] avg loss 0.00336749, throughput 9.47993K wps
[Epoch 46 Batch 90/162] avg loss 0.00344873, throughput 9.34147K wps
[Epoch 46 Batch 120/162] avg loss 0.0035183, throughput 9.48268K wps
[Epoch 46 Batch 150/162] avg loss 0.00345836, throughput 9.35019K wps
Begin Testing...
[Epoch 46] train avg loss 0.0034463, dev acc 0.9056, dev avg loss 0.238973, throughput 9.48535K wps
Observed Improvement.
Begin Testing...
[Epoch 47 Batch 30/162] avg loss 0.00339077, throughput 9.74177K wps
[Epoch 47 Batch 60/162] avg loss 0.00336536, throughput 9.42236K wps
[Epoch 47 Batch 90/162] avg loss 0.00340757, throughput 9.36432K wps
[Epoch 47 Batch 120/162] avg loss 0.0034393, throughput 9.46413K wps
[Epoch 47 Batch 150/162] avg loss 0.00355601, throughput 9.38071K wps
Begin Testing...
[Epoch 47] train avg loss 0.00342528, dev acc 0.9033, dev avg loss 0.240823, throughput 9.46974K wps
[Epoch 48 Batch 30/162] avg loss 0.00322189, throughput 9.66421K wps
[Epoch 48 Batch 60/162] avg loss 0.00320269, throughput 9.45993K wps
[Epoch 48 Batch 90/162] avg loss 0.0032689, throughput 9.4536K wps
[Epoch 48 Batch 120/162] avg loss 0.00379154, throughput 9.49119K wps
[Epoch 48 Batch 150/162] avg loss 0.00371524, throughput 9.42834K wps
Begin Testing...
[Epoch 48] train avg loss 0.00341801, dev acc 0.9056, dev avg loss 0.239078, throughput 9.50927K wps
Observed Improvement.
Begin Testing...
[Epoch 49 Batch 30/162] avg loss 0.00338094, throughput 9.58416K wps
[Epoch 49 Batch 60/162] avg loss 0.00354519, throughput 9.36248K wps
[Epoch 49 Batch 90/162] avg loss 0.00297501, throughput 9.58056K wps
[Epoch 49 Batch 120/162] avg loss 0.0032967, throughput 9.35219K wps
[Epoch 49 Batch 150/162] avg loss 0.0033296, throughput 9.52741K wps
Begin Testing...
[Epoch 49] train avg loss 0.00330891, dev acc 0.9044, dev avg loss 0.236495, throughput 9.4503K wps
[Epoch 50 Batch 30/162] avg loss 0.0030225, throughput 9.69499K wps
[Epoch 50 Batch 60/162] avg loss 0.00327383, throughput 9.50718K wps
[Epoch 50 Batch 90/162] avg loss 0.00313939, throughput 9.26712K wps
[Epoch 50 Batch 120/162] avg loss 0.00361152, throughput 9.31892K wps
[Epoch 50 Batch 150/162] avg loss 0.00322057, throughput 9.52716K wps
Begin Testing...
[Epoch 50] train avg loss 0.00325964, dev acc 0.9067, dev avg loss 0.23636, throughput 9.45268K wps
Observed Improvement.
Begin Testing...
[Epoch 51 Batch 30/162] avg loss 0.00355772, throughput 9.67919K wps
[Epoch 51 Batch 60/162] avg loss 0.0031992, throughput 9.38083K wps
[Epoch 51 Batch 90/162] avg loss 0.00295139, throughput 9.4255K wps
[Epoch 51 Batch 120/162] avg loss 0.00328806, throughput 9.61017K wps
[Epoch 51 Batch 150/162] avg loss 0.00325621, throughput 9.35851K wps
Begin Testing...
[Epoch 51] train avg loss 0.00324921, dev acc 0.9044, dev avg loss 0.236424, throughput 9.47955K wps
[Epoch 52 Batch 30/162] avg loss 0.00314386, throughput 9.66327K wps
[Epoch 52 Batch 60/162] avg loss 0.0029504, throughput 9.38874K wps
[Epoch 52 Batch 90/162] avg loss 0.00315145, throughput 9.42455K wps
[Epoch 52 Batch 120/162] avg loss 0.00349968, throughput 9.45921K wps
[Epoch 52 Batch 150/162] avg loss 0.00335827, throughput 9.33461K wps
Begin Testing...
[Epoch 52] train avg loss 0.00321288, dev acc 0.9089, dev avg loss 0.233904, throughput 9.45871K wps
Observed Improvement.
Begin Testing...
[Epoch 53 Batch 30/162] avg loss 0.00337183, throughput 9.50362K wps
[Epoch 53 Batch 60/162] avg loss 0.00301729, throughput 9.35619K wps
[Epoch 53 Batch 90/162] avg loss 0.0032495, throughput 9.3928K wps
[Epoch 53 Batch 120/162] avg loss 0.00285333, throughput 9.32451K wps
[Epoch 53 Batch 150/162] avg loss 0.00305276, throughput 9.36361K wps
Begin Testing...
[Epoch 53] train avg loss 0.00310959, dev acc 0.9089, dev avg loss 0.233115, throughput 9.40642K wps
Observed Improvement.
Begin Testing...
[Epoch 54 Batch 30/162] avg loss 0.00337602, throughput 9.62012K wps
[Epoch 54 Batch 60/162] avg loss 0.00293776, throughput 9.51614K wps
[Epoch 54 Batch 90/162] avg loss 0.00315104, throughput 9.36396K wps
[Epoch 54 Batch 120/162] avg loss 0.00317715, throughput 9.59708K wps
[Epoch 54 Batch 150/162] avg loss 0.00295044, throughput 9.24181K wps
Begin Testing...
[Epoch 54] train avg loss 0.00312607, dev acc 0.9044, dev avg loss 0.234564, throughput 9.45582K wps
[Epoch 55 Batch 30/162] avg loss 0.00289326, throughput 9.64173K wps
[Epoch 55 Batch 60/162] avg loss 0.00308962, throughput 9.46413K wps
[Epoch 55 Batch 90/162] avg loss 0.00319154, throughput 9.48583K wps
[Epoch 55 Batch 120/162] avg loss 0.00295428, throughput 9.36807K wps
[Epoch 55 Batch 150/162] avg loss 0.00343454, throughput 9.49509K wps
Begin Testing...
[Epoch 55] train avg loss 0.00311881, dev acc 0.9056, dev avg loss 0.232507, throughput 9.49586K wps
[Epoch 56 Batch 30/162] avg loss 0.00282086, throughput 9.7074K wps
[Epoch 56 Batch 60/162] avg loss 0.00311448, throughput 9.33717K wps
[Epoch 56 Batch 90/162] avg loss 0.00292301, throughput 9.31939K wps
[Epoch 56 Batch 120/162] avg loss 0.00323353, throughput 9.38096K wps
[Epoch 56 Batch 150/162] avg loss 0.00290342, throughput 9.56886K wps
Begin Testing...
[Epoch 56] train avg loss 0.0030422, dev acc 0.9100, dev avg loss 0.231817, throughput 9.44877K wps
Observed Improvement.
Begin Testing...
[Epoch 57 Batch 30/162] avg loss 0.00276019, throughput 9.47032K wps
[Epoch 57 Batch 60/162] avg loss 0.00307052, throughput 9.34332K wps
[Epoch 57 Batch 90/162] avg loss 0.00313161, throughput 9.66873K wps
[Epoch 57 Batch 120/162] avg loss 0.00292836, throughput 9.38455K wps
[Epoch 57 Batch 150/162] avg loss 0.00315751, throughput 9.38015K wps
Begin Testing...
[Epoch 57] train avg loss 0.00301593, dev acc 0.9078, dev avg loss 0.231714, throughput 9.44985K wps
[Epoch 58 Batch 30/162] avg loss 0.00257349, throughput 9.76415K wps
[Epoch 58 Batch 60/162] avg loss 0.00300848, throughput 9.37982K wps
[Epoch 58 Batch 90/162] avg loss 0.00290385, throughput 9.38917K wps
[Epoch 58 Batch 120/162] avg loss 0.0030307, throughput 9.29048K wps
[Epoch 58 Batch 150/162] avg loss 0.00284335, throughput 9.49495K wps
Begin Testing...
[Epoch 58] train avg loss 0.00289931, dev acc 0.9089, dev avg loss 0.230422, throughput 9.46315K wps
[Epoch 59 Batch 30/162] avg loss 0.00292745, throughput 9.63754K wps
[Epoch 59 Batch 60/162] avg loss 0.00313592, throughput 9.33221K wps
[Epoch 59 Batch 90/162] avg loss 0.00289726, throughput 9.49064K wps
[Epoch 59 Batch 120/162] avg loss 0.00299202, throughput 9.32261K wps
[Epoch 59 Batch 150/162] avg loss 0.00281195, throughput 9.38516K wps
Begin Testing...
[Epoch 59] train avg loss 0.00294905, dev acc 0.9100, dev avg loss 0.230025, throughput 9.42549K wps
Observed Improvement.
Begin Testing...
[Epoch 60 Batch 30/162] avg loss 0.00290384, throughput 9.53357K wps
[Epoch 60 Batch 60/162] avg loss 0.00289049, throughput 9.35384K wps
[Epoch 60 Batch 90/162] avg loss 0.00290897, throughput 9.52855K wps
[Epoch 60 Batch 120/162] avg loss 0.0030105, throughput 9.41865K wps
[Epoch 60 Batch 150/162] avg loss 0.00292313, throughput 9.33795K wps
Begin Testing...
[Epoch 60] train avg loss 0.00290948, dev acc 0.9100, dev avg loss 0.229548, throughput 9.42695K wps
Observed Improvement.
Begin Testing...
[Epoch 61 Batch 30/162] avg loss 0.00297252, throughput 9.64192K wps
[Epoch 61 Batch 60/162] avg loss 0.00282943, throughput 9.42059K wps
[Epoch 61 Batch 90/162] avg loss 0.00267761, throughput 9.35949K wps
[Epoch 61 Batch 120/162] avg loss 0.00284388, throughput 9.34975K wps
[Epoch 61 Batch 150/162] avg loss 0.00288134, throughput 9.48142K wps
Begin Testing...
[Epoch 61] train avg loss 0.00281883, dev acc 0.9089, dev avg loss 0.230125, throughput 9.45092K wps
[Epoch 62 Batch 30/162] avg loss 0.00246717, throughput 9.70041K wps
[Epoch 62 Batch 60/162] avg loss 0.00307574, throughput 9.51913K wps
[Epoch 62 Batch 90/162] avg loss 0.00269469, throughput 9.58892K wps
[Epoch 62 Batch 120/162] avg loss 0.00297206, throughput 9.29189K wps
[Epoch 62 Batch 150/162] avg loss 0.00265812, throughput 9.38103K wps
Begin Testing...
[Epoch 62] train avg loss 0.0027563, dev acc 0.9100, dev avg loss 0.228795, throughput 9.50347K wps
Observed Improvement.
Begin Testing...
[Epoch 63 Batch 30/162] avg loss 0.00258846, throughput 9.59087K wps
[Epoch 63 Batch 60/162] avg loss 0.00294047, throughput 9.17447K wps
[Epoch 63 Batch 90/162] avg loss 0.00280207, throughput 9.71011K wps
[Epoch 63 Batch 120/162] avg loss 0.00268736, throughput 9.50271K wps
[Epoch 63 Batch 150/162] avg loss 0.00258904, throughput 9.49908K wps
Begin Testing...
[Epoch 63] train avg loss 0.00274539, dev acc 0.9089, dev avg loss 0.228536, throughput 9.48636K wps
[Epoch 64 Batch 30/162] avg loss 0.00262524, throughput 9.42954K wps
[Epoch 64 Batch 60/162] avg loss 0.00245183, throughput 9.48672K wps
[Epoch 64 Batch 90/162] avg loss 0.00281167, throughput 9.45471K wps
[Epoch 64 Batch 120/162] avg loss 0.00268571, throughput 9.32386K wps
[Epoch 64 Batch 150/162] avg loss 0.00296333, throughput 9.41571K wps
Begin Testing...
[Epoch 64] train avg loss 0.0026994, dev acc 0.9089, dev avg loss 0.228003, throughput 9.43291K wps
[Epoch 65 Batch 30/162] avg loss 0.00275917, throughput 9.50582K wps
[Epoch 65 Batch 60/162] avg loss 0.00260699, throughput 9.51252K wps
[Epoch 65 Batch 90/162] avg loss 0.00271515, throughput 9.51646K wps
[Epoch 65 Batch 120/162] avg loss 0.00275958, throughput 9.50542K wps
[Epoch 65 Batch 150/162] avg loss 0.00267258, throughput 9.5255K wps
Begin Testing...
[Epoch 65] train avg loss 0.00272376, dev acc 0.9067, dev avg loss 0.228016, throughput 9.48875K wps
[Epoch 66 Batch 30/162] avg loss 0.00263177, throughput 9.71139K wps
[Epoch 66 Batch 60/162] avg loss 0.00263791, throughput 9.2868K wps
[Epoch 66 Batch 90/162] avg loss 0.00257487, throughput 9.39998K wps
[Epoch 66 Batch 120/162] avg loss 0.00269154, throughput 9.2347K wps
[Epoch 66 Batch 150/162] avg loss 0.00270904, throughput 9.38043K wps
Begin Testing...
[Epoch 66] train avg loss 0.00264418, dev acc 0.9111, dev avg loss 0.227126, throughput 9.40462K wps
Observed Improvement.
Begin Testing...
[Epoch 67 Batch 30/162] avg loss 0.00275374, throughput 9.47258K wps
[Epoch 67 Batch 60/162] avg loss 0.0026151, throughput 9.5266K wps
[Epoch 67 Batch 90/162] avg loss 0.00241788, throughput 9.56929K wps
[Epoch 67 Batch 120/162] avg loss 0.00259874, throughput 9.27978K wps
[Epoch 67 Batch 150/162] avg loss 0.00270398, throughput 9.32192K wps
Begin Testing...
[Epoch 67] train avg loss 0.00261561, dev acc 0.9111, dev avg loss 0.226928, throughput 9.42287K wps
Observed Improvement.
Begin Testing...
[Epoch 68 Batch 30/162] avg loss 0.00256686, throughput 9.64532K wps
[Epoch 68 Batch 60/162] avg loss 0.00262526, throughput 9.23282K wps
[Epoch 68 Batch 90/162] avg loss 0.00271918, throughput 9.46017K wps
[Epoch 68 Batch 120/162] avg loss 0.00246705, throughput 9.27295K wps
[Epoch 68 Batch 150/162] avg loss 0.00260137, throughput 9.70299K wps
Begin Testing...
[Epoch 68] train avg loss 0.00258476, dev acc 0.9111, dev avg loss 0.226276, throughput 9.46618K wps
Observed Improvement.
Begin Testing...
[Epoch 69 Batch 30/162] avg loss 0.00251701, throughput 9.55871K wps
[Epoch 69 Batch 60/162] avg loss 0.0025385, throughput 9.33942K wps
[Epoch 69 Batch 90/162] avg loss 0.0025573, throughput 9.41148K wps
[Epoch 69 Batch 120/162] avg loss 0.00234348, throughput 9.37491K wps
[Epoch 69 Batch 150/162] avg loss 0.00267169, throughput 9.36295K wps
Begin Testing...
[Epoch 69] train avg loss 0.0025176, dev acc 0.9100, dev avg loss 0.225957, throughput 9.41098K wps
[Epoch 70 Batch 30/162] avg loss 0.00253908, throughput 9.7031K wps
[Epoch 70 Batch 60/162] avg loss 0.00261232, throughput 9.27576K wps
[Epoch 70 Batch 90/162] avg loss 0.00249504, throughput 9.51337K wps
[Epoch 70 Batch 120/162] avg loss 0.0026178, throughput 9.50653K wps
[Epoch 70 Batch 150/162] avg loss 0.00213221, throughput 9.43976K wps
Begin Testing...
[Epoch 70] train avg loss 0.0024684, dev acc 0.9122, dev avg loss 0.225585, throughput 9.47026K wps
Observed Improvement.
Begin Testing...
[Epoch 71 Batch 30/162] avg loss 0.00259165, throughput 9.72889K wps
[Epoch 71 Batch 60/162] avg loss 0.00247668, throughput 9.30801K wps
[Epoch 71 Batch 90/162] avg loss 0.00247152, throughput 9.48814K wps
[Epoch 71 Batch 120/162] avg loss 0.00253448, throughput 9.3957K wps
[Epoch 71 Batch 150/162] avg loss 0.00231425, throughput 9.37061K wps
Begin Testing...
[Epoch 71] train avg loss 0.00244365, dev acc 0.9089, dev avg loss 0.226053, throughput 9.44894K wps
[Epoch 72 Batch 30/162] avg loss 0.00247284, throughput 9.67976K wps
[Epoch 72 Batch 60/162] avg loss 0.00239089, throughput 9.4674K wps
[Epoch 72 Batch 90/162] avg loss 0.00224955, throughput 9.47567K wps
[Epoch 72 Batch 120/162] avg loss 0.00236024, throughput 9.3556K wps
[Epoch 72 Batch 150/162] avg loss 0.00249529, throughput 9.52992K wps
Begin Testing...
[Epoch 72] train avg loss 0.00241203, dev acc 0.9111, dev avg loss 0.225074, throughput 9.48214K wps
[Epoch 73 Batch 30/162] avg loss 0.00247017, throughput 9.57709K wps
[Epoch 73 Batch 60/162] avg loss 0.00254371, throughput 9.21256K wps
[Epoch 73 Batch 90/162] avg loss 0.00221639, throughput 9.48683K wps
[Epoch 73 Batch 120/162] avg loss 0.00221465, throughput 9.31865K wps
[Epoch 73 Batch 150/162] avg loss 0.00236916, throughput 9.42793K wps
Begin Testing...
[Epoch 73] train avg loss 0.00235228, dev acc 0.9100, dev avg loss 0.224577, throughput 9.38474K wps
[Epoch 74 Batch 30/162] avg loss 0.0024305, throughput 9.6846K wps
[Epoch 74 Batch 60/162] avg loss 0.00235505, throughput 9.35732K wps
[Epoch 74 Batch 90/162] avg loss 0.00236726, throughput 9.34874K wps
[Epoch 74 Batch 120/162] avg loss 0.00250422, throughput 9.3398K wps
[Epoch 74 Batch 150/162] avg loss 0.00226989, throughput 9.53487K wps
Begin Testing...
[Epoch 74] train avg loss 0.00237898, dev acc 0.9100, dev avg loss 0.225316, throughput 9.45828K wps
[Epoch 75 Batch 30/162] avg loss 0.00228304, throughput 9.5532K wps
[Epoch 75 Batch 60/162] avg loss 0.00213432, throughput 9.32872K wps
[Epoch 75 Batch 90/162] avg loss 0.00239232, throughput 9.46167K wps
[Epoch 75 Batch 120/162] avg loss 0.00230707, throughput 9.54129K wps
[Epoch 75 Batch 150/162] avg loss 0.00228749, throughput 9.42594K wps
Begin Testing...
[Epoch 75] train avg loss 0.00230046, dev acc 0.9056, dev avg loss 0.22571, throughput 9.43726K wps
[Epoch 76 Batch 30/162] avg loss 0.00224714, throughput 9.45192K wps
[Epoch 76 Batch 60/162] avg loss 0.00231544, throughput 9.39437K wps
[Epoch 76 Batch 90/162] avg loss 0.00205823, throughput 9.443K wps
[Epoch 76 Batch 120/162] avg loss 0.00209177, throughput 9.38991K wps
[Epoch 76 Batch 150/162] avg loss 0.00237868, throughput 9.57005K wps
Begin Testing...
[Epoch 76] train avg loss 0.00220901, dev acc 0.9111, dev avg loss 0.224357, throughput 9.44794K wps
[Epoch 77 Batch 30/162] avg loss 0.00220209, throughput 9.55915K wps
[Epoch 77 Batch 60/162] avg loss 0.00222217, throughput 9.50527K wps
[Epoch 77 Batch 90/162] avg loss 0.00199414, throughput 9.35501K wps
[Epoch 77 Batch 120/162] avg loss 0.00254181, throughput 9.47514K wps
[Epoch 77 Batch 150/162] avg loss 0.00224727, throughput 9.51065K wps
Begin Testing...
[Epoch 77] train avg loss 0.00223525, dev acc 0.9078, dev avg loss 0.225648, throughput 9.47539K wps
[Epoch 78 Batch 30/162] avg loss 0.00253816, throughput 9.64244K wps
[Epoch 78 Batch 60/162] avg loss 0.00205734, throughput 9.54058K wps
[Epoch 78 Batch 90/162] avg loss 0.00220292, throughput 9.44492K wps
[Epoch 78 Batch 120/162] avg loss 0.00226932, throughput 9.46821K wps
[Epoch 78 Batch 150/162] avg loss 0.00221516, throughput 9.5915K wps
Begin Testing...
[Epoch 78] train avg loss 0.00224516, dev acc 0.9133, dev avg loss 0.223769, throughput 9.52174K wps
Observed Improvement.
Begin Testing...
[Epoch 79 Batch 30/162] avg loss 0.00216032, throughput 9.60881K wps
[Epoch 79 Batch 60/162] avg loss 0.00235953, throughput 9.34424K wps
[Epoch 79 Batch 90/162] avg loss 0.00205525, throughput 9.35477K wps
[Epoch 79 Batch 120/162] avg loss 0.00187817, throughput 9.43884K wps
[Epoch 79 Batch 150/162] avg loss 0.00221021, throughput 9.5554K wps
Begin Testing...
[Epoch 79] train avg loss 0.00216147, dev acc 0.9122, dev avg loss 0.22369, throughput 9.45645K wps
[Epoch 80 Batch 30/162] avg loss 0.00207867, throughput 9.65489K wps
[Epoch 80 Batch 60/162] avg loss 0.00225942, throughput 9.2852K wps
[Epoch 80 Batch 90/162] avg loss 0.00236887, throughput 9.59068K wps
[Epoch 80 Batch 120/162] avg loss 0.00220089, throughput 9.55084K wps
[Epoch 80 Batch 150/162] avg loss 0.0021572, throughput 9.36314K wps
Begin Testing...
[Epoch 80] train avg loss 0.00218079, dev acc 0.9133, dev avg loss 0.223047, throughput 9.48836K wps
Observed Improvement.
Begin Testing...
[Epoch 81 Batch 30/162] avg loss 0.00207491, throughput 9.46962K wps
[Epoch 81 Batch 60/162] avg loss 0.00213306, throughput 9.26397K wps
[Epoch 81 Batch 90/162] avg loss 0.00220052, throughput 9.49139K wps
[Epoch 81 Batch 120/162] avg loss 0.00208682, throughput 9.33806K wps
[Epoch 81 Batch 150/162] avg loss 0.00209385, throughput 9.42478K wps
Begin Testing...
[Epoch 81] train avg loss 0.00212402, dev acc 0.9100, dev avg loss 0.225884, throughput 9.4045K wps
[Epoch 82 Batch 30/162] avg loss 0.00190511, throughput 9.40148K wps
[Epoch 82 Batch 60/162] avg loss 0.0021001, throughput 9.37453K wps
[Epoch 82 Batch 90/162] avg loss 0.00214368, throughput 9.33718K wps
[Epoch 82 Batch 120/162] avg loss 0.00220985, throughput 9.44888K wps
[Epoch 82 Batch 150/162] avg loss 0.00189709, throughput 9.47173K wps
Begin Testing...
[Epoch 82] train avg loss 0.0020473, dev acc 0.9133, dev avg loss 0.222767, throughput 9.39326K wps
Observed Improvement.
Begin Testing...
[Epoch 83 Batch 30/162] avg loss 0.00202925, throughput 9.58715K wps
[Epoch 83 Batch 60/162] avg loss 0.00197542, throughput 9.35232K wps
[Epoch 83 Batch 90/162] avg loss 0.00190326, throughput 9.39897K wps
[Epoch 83 Batch 120/162] avg loss 0.00236555, throughput 9.30035K wps
[Epoch 83 Batch 150/162] avg loss 0.00230602, throughput 9.45884K wps
Begin Testing...
[Epoch 83] train avg loss 0.00209648, dev acc 0.9122, dev avg loss 0.222183, throughput 9.40048K wps
[Epoch 84 Batch 30/162] avg loss 0.00203792, throughput 9.54848K wps
[Epoch 84 Batch 60/162] avg loss 0.00198852, throughput 9.43485K wps
[Epoch 84 Batch 90/162] avg loss 0.00190257, throughput 9.54312K wps
[Epoch 84 Batch 120/162] avg loss 0.00198403, throughput 9.33641K wps
[Epoch 84 Batch 150/162] avg loss 0.00206311, throughput 9.34842K wps
Begin Testing...
[Epoch 84] train avg loss 0.00198992, dev acc 0.9122, dev avg loss 0.221822, throughput 9.44791K wps
[Epoch 85 Batch 30/162] avg loss 0.0019478, throughput 9.74634K wps
[Epoch 85 Batch 60/162] avg loss 0.00199411, throughput 9.25331K wps
[Epoch 85 Batch 90/162] avg loss 0.00179936, throughput 9.50767K wps
[Epoch 85 Batch 120/162] avg loss 0.00210383, throughput 9.28287K wps
[Epoch 85 Batch 150/162] avg loss 0.00225755, throughput 9.46509K wps
Begin Testing...
[Epoch 85] train avg loss 0.00201538, dev acc 0.9078, dev avg loss 0.221921, throughput 9.44935K wps
[Epoch 86 Batch 30/162] avg loss 0.00194055, throughput 9.53472K wps
[Epoch 86 Batch 60/162] avg loss 0.00219943, throughput 9.56271K wps
[Epoch 86 Batch 90/162] avg loss 0.00219676, throughput 9.49602K wps
[Epoch 86 Batch 120/162] avg loss 0.00191981, throughput 9.59609K wps
[Epoch 86 Batch 150/162] avg loss 0.00194707, throughput 9.46138K wps
Begin Testing...
[Epoch 86] train avg loss 0.00202743, dev acc 0.9133, dev avg loss 0.221779, throughput 9.5143K wps
Observed Improvement.
Begin Testing...
[Epoch 87 Batch 30/162] avg loss 0.00204262, throughput 9.525K wps
[Epoch 87 Batch 60/162] avg loss 0.0017773, throughput 9.2505K wps
[Epoch 87 Batch 90/162] avg loss 0.00195111, throughput 9.31364K wps
[Epoch 87 Batch 120/162] avg loss 0.00199924, throughput 9.29252K wps
[Epoch 87 Batch 150/162] avg loss 0.00176251, throughput 9.31483K wps
Begin Testing...
[Epoch 87] train avg loss 0.00193348, dev acc 0.9133, dev avg loss 0.221594, throughput 9.33351K wps
Observed Improvement.
Begin Testing...
[Epoch 88 Batch 30/162] avg loss 0.00192227, throughput 9.72592K wps
[Epoch 88 Batch 60/162] avg loss 0.00188778, throughput 9.33491K wps
[Epoch 88 Batch 90/162] avg loss 0.00203899, throughput 9.40392K wps
[Epoch 88 Batch 120/162] avg loss 0.00171242, throughput 9.47379K wps
[Epoch 88 Batch 150/162] avg loss 0.002246, throughput 9.35862K wps
Begin Testing...
[Epoch 88] train avg loss 0.00195975, dev acc 0.9100, dev avg loss 0.222178, throughput 9.4506K wps
[Epoch 89 Batch 30/162] avg loss 0.00177319, throughput 9.59091K wps
[Epoch 89 Batch 60/162] avg loss 0.00209377, throughput 9.44849K wps
[Epoch 89 Batch 90/162] avg loss 0.00178207, throughput 9.2254K wps
[Epoch 89 Batch 120/162] avg loss 0.00191529, throughput 9.44955K wps
[Epoch 89 Batch 150/162] avg loss 0.00182595, throughput 9.37694K wps
Begin Testing...
[Epoch 89] train avg loss 0.00186672, dev acc 0.9122, dev avg loss 0.222913, throughput 9.41074K wps
[Epoch 90 Batch 30/162] avg loss 0.00205714, throughput 9.72563K wps
[Epoch 90 Batch 60/162] avg loss 0.00198082, throughput 9.2028K wps
[Epoch 90 Batch 90/162] avg loss 0.00186353, throughput 9.43993K wps
[Epoch 90 Batch 120/162] avg loss 0.00199619, throughput 9.27534K wps
[Epoch 90 Batch 150/162] avg loss 0.00175283, throughput 9.25673K wps
Begin Testing...
[Epoch 90] train avg loss 0.00191809, dev acc 0.9133, dev avg loss 0.221372, throughput 9.37943K wps
Observed Improvement.
Begin Testing...
[Epoch 91 Batch 30/162] avg loss 0.00178068, throughput 9.50519K wps
[Epoch 91 Batch 60/162] avg loss 0.00189745, throughput 9.52216K wps
[Epoch 91 Batch 90/162] avg loss 0.00167091, throughput 9.33696K wps
[Epoch 91 Batch 120/162] avg loss 0.00191668, throughput 9.61392K wps
[Epoch 91 Batch 150/162] avg loss 0.00195061, throughput 9.42081K wps
Begin Testing...
[Epoch 91] train avg loss 0.00186935, dev acc 0.9133, dev avg loss 0.221206, throughput 9.48099K wps
Observed Improvement.
Begin Testing...
[Epoch 92 Batch 30/162] avg loss 0.00169665, throughput 9.54121K wps
[Epoch 92 Batch 60/162] avg loss 0.00180584, throughput 9.3826K wps
[Epoch 92 Batch 90/162] avg loss 0.00209973, throughput 9.36879K wps
[Epoch 92 Batch 120/162] avg loss 0.00187617, throughput 9.47832K wps
[Epoch 92 Batch 150/162] avg loss 0.00166508, throughput 9.45659K wps
Begin Testing...
[Epoch 92] train avg loss 0.0018442, dev acc 0.9100, dev avg loss 0.220846, throughput 9.44129K wps
[Epoch 93 Batch 30/162] avg loss 0.0017494, throughput 9.52524K wps
[Epoch 93 Batch 60/162] avg loss 0.0017066, throughput 9.49617K wps
[Epoch 93 Batch 90/162] avg loss 0.00184144, throughput 9.41577K wps
[Epoch 93 Batch 120/162] avg loss 0.00172695, throughput 9.56218K wps
[Epoch 93 Batch 150/162] avg loss 0.00200898, throughput 9.43064K wps
Begin Testing...
[Epoch 93] train avg loss 0.00180885, dev acc 0.9122, dev avg loss 0.221514, throughput 9.48772K wps
[Epoch 94 Batch 30/162] avg loss 0.00180003, throughput 9.55219K wps
[Epoch 94 Batch 60/162] avg loss 0.00166697, throughput 9.37779K wps
[Epoch 94 Batch 90/162] avg loss 0.00169316, throughput 9.37023K wps
[Epoch 94 Batch 120/162] avg loss 0.00160552, throughput 9.39516K wps
[Epoch 94 Batch 150/162] avg loss 0.00176401, throughput 9.37656K wps
Begin Testing...
[Epoch 94] train avg loss 0.00171374, dev acc 0.9078, dev avg loss 0.221657, throughput 9.42344K wps
[Epoch 95 Batch 30/162] avg loss 0.00181362, throughput 9.6223K wps
[Epoch 95 Batch 60/162] avg loss 0.00184647, throughput 9.40739K wps
[Epoch 95 Batch 90/162] avg loss 0.0016597, throughput 9.26303K wps
[Epoch 95 Batch 120/162] avg loss 0.00182766, throughput 9.4012K wps
[Epoch 95 Batch 150/162] avg loss 0.00173316, throughput 9.63212K wps
Begin Testing...
[Epoch 95] train avg loss 0.00179003, dev acc 0.9133, dev avg loss 0.221095, throughput 9.45708K wps
Observed Improvement.
Begin Testing...
[Epoch 96 Batch 30/162] avg loss 0.00156056, throughput 9.46833K wps
[Epoch 96 Batch 60/162] avg loss 0.00210087, throughput 9.38327K wps
[Epoch 96 Batch 90/162] avg loss 0.00167795, throughput 9.29705K wps
[Epoch 96 Batch 120/162] avg loss 0.00186134, throughput 9.30113K wps
[Epoch 96 Batch 150/162] avg loss 0.00165538, throughput 9.52666K wps
Begin Testing...
[Epoch 96] train avg loss 0.0017698, dev acc 0.9133, dev avg loss 0.220608, throughput 9.39005K wps
Observed Improvement.
Begin Testing...
[Epoch 97 Batch 30/162] avg loss 0.0017643, throughput 9.56351K wps
[Epoch 97 Batch 60/162] avg loss 0.00172027, throughput 9.30563K wps
[Epoch 97 Batch 90/162] avg loss 0.00168552, throughput 9.6811K wps
[Epoch 97 Batch 120/162] avg loss 0.00159191, throughput 9.45585K wps
[Epoch 97 Batch 150/162] avg loss 0.00174208, throughput 9.49504K wps
Begin Testing...
[Epoch 97] train avg loss 0.00168874, dev acc 0.9133, dev avg loss 0.220206, throughput 9.48811K wps
Observed Improvement.
Begin Testing...
[Epoch 98 Batch 30/162] avg loss 0.00164827, throughput 9.62847K wps
[Epoch 98 Batch 60/162] avg loss 0.00191954, throughput 9.41796K wps
[Epoch 98 Batch 90/162] avg loss 0.00171277, throughput 9.39381K wps
[Epoch 98 Batch 120/162] avg loss 0.00165838, throughput 9.40584K wps
[Epoch 98 Batch 150/162] avg loss 0.00172756, throughput 9.40523K wps
Begin Testing...
[Epoch 98] train avg loss 0.00173469, dev acc 0.9122, dev avg loss 0.220559, throughput 9.44944K wps
[Epoch 99 Batch 30/162] avg loss 0.00164538, throughput 9.60262K wps
[Epoch 99 Batch 60/162] avg loss 0.0016509, throughput 9.45207K wps
[Epoch 99 Batch 90/162] avg loss 0.00176076, throughput 9.52299K wps
[Epoch 99 Batch 120/162] avg loss 0.00147589, throughput 9.32321K wps
[Epoch 99 Batch 150/162] avg loss 0.00164025, throughput 9.45877K wps
Begin Testing...
[Epoch 99] train avg loss 0.00162806, dev acc 0.9078, dev avg loss 0.220188, throughput 9.45944K wps
[Epoch 100 Batch 30/162] avg loss 0.00163632, throughput 9.38995K wps
[Epoch 100 Batch 60/162] avg loss 0.00146837, throughput 9.47208K wps
[Epoch 100 Batch 90/162] avg loss 0.00168574, throughput 9.42968K wps
[Epoch 100 Batch 120/162] avg loss 0.00155409, throughput 9.36397K wps
[Epoch 100 Batch 150/162] avg loss 0.0015598, throughput 9.48158K wps
Begin Testing...
[Epoch 100] train avg loss 0.00160004, dev acc 0.9100, dev avg loss 0.220788, throughput 9.42108K wps
[Epoch 101 Batch 30/162] avg loss 0.00137179, throughput 9.83386K wps
[Epoch 101 Batch 60/162] avg loss 0.00154931, throughput 9.25186K wps
[Epoch 101 Batch 90/162] avg loss 0.0017936, throughput 9.42156K wps
[Epoch 101 Batch 120/162] avg loss 0.00148038, throughput 9.47936K wps
[Epoch 101 Batch 150/162] avg loss 0.00177874, throughput 9.34887K wps
Begin Testing...
[Epoch 101] train avg loss 0.00160404, dev acc 0.9144, dev avg loss 0.220437, throughput 9.45488K wps
Observed Improvement.
Begin Testing...
[Epoch 102 Batch 30/162] avg loss 0.00157739, throughput 9.55005K wps
[Epoch 102 Batch 60/162] avg loss 0.001658, throughput 9.31624K wps
[Epoch 102 Batch 90/162] avg loss 0.00152966, throughput 9.28802K wps
[Epoch 102 Batch 120/162] avg loss 0.00140791, throughput 9.33692K wps
[Epoch 102 Batch 150/162] avg loss 0.00169811, throughput 9.34028K wps
Begin Testing...
[Epoch 102] train avg loss 0.00159247, dev acc 0.9122, dev avg loss 0.219555, throughput 9.36021K wps
[Epoch 103 Batch 30/162] avg loss 0.00168681, throughput 9.49328K wps
[Epoch 103 Batch 60/162] avg loss 0.00154863, throughput 9.34576K wps
[Epoch 103 Batch 90/162] avg loss 0.00156526, throughput 9.2873K wps
[Epoch 103 Batch 120/162] avg loss 0.00152021, throughput 9.25464K wps
[Epoch 103 Batch 150/162] avg loss 0.00157635, throughput 9.41008K wps
Begin Testing...
[Epoch 103] train avg loss 0.00160338, dev acc 0.9111, dev avg loss 0.21961, throughput 9.36781K wps
[Epoch 104 Batch 30/162] avg loss 0.00156989, throughput 9.45905K wps
[Epoch 104 Batch 60/162] avg loss 0.00164398, throughput 9.56919K wps
[Epoch 104 Batch 90/162] avg loss 0.00155207, throughput 9.29887K wps
[Epoch 104 Batch 120/162] avg loss 0.00139914, throughput 9.36164K wps
[Epoch 104 Batch 150/162] avg loss 0.00162402, throughput 9.54015K wps
Begin Testing...
[Epoch 104] train avg loss 0.00154726, dev acc 0.9078, dev avg loss 0.219858, throughput 9.42267K wps
[Epoch 105 Batch 30/162] avg loss 0.0014797, throughput 9.48179K wps
[Epoch 105 Batch 60/162] avg loss 0.0015317, throughput 9.44771K wps
[Epoch 105 Batch 90/162] avg loss 0.00153842, throughput 9.37849K wps
[Epoch 105 Batch 120/162] avg loss 0.00146306, throughput 9.48313K wps
[Epoch 105 Batch 150/162] avg loss 0.00154514, throughput 9.48049K wps
Begin Testing...
[Epoch 105] train avg loss 0.00151947, dev acc 0.9144, dev avg loss 0.220351, throughput 9.43629K wps
Observed Improvement.
Begin Testing...
[Epoch 106 Batch 30/162] avg loss 0.0014729, throughput 9.68706K wps
[Epoch 106 Batch 60/162] avg loss 0.00149937, throughput 9.43221K wps
[Epoch 106 Batch 90/162] avg loss 0.00155456, throughput 9.35849K wps
[Epoch 106 Batch 120/162] avg loss 0.00156519, throughput 9.61776K wps
[Epoch 106 Batch 150/162] avg loss 0.00154412, throughput 9.62167K wps
Begin Testing...
[Epoch 106] train avg loss 0.00153836, dev acc 0.9156, dev avg loss 0.220173, throughput 9.54362K wps
Observed Improvement.
Begin Testing...
[Epoch 107 Batch 30/162] avg loss 0.00158925, throughput 9.7098K wps
[Epoch 107 Batch 60/162] avg loss 0.00158619, throughput 9.24939K wps
[Epoch 107 Batch 90/162] avg loss 0.00155721, throughput 9.51523K wps
[Epoch 107 Batch 120/162] avg loss 0.0014287, throughput 9.36544K wps
[Epoch 107 Batch 150/162] avg loss 0.00142872, throughput 9.3352K wps
Begin Testing...
[Epoch 107] train avg loss 0.00151437, dev acc 0.9089, dev avg loss 0.219802, throughput 9.44402K wps
[Epoch 108 Batch 30/162] avg loss 0.00132171, throughput 9.56079K wps
[Epoch 108 Batch 60/162] avg loss 0.00153351, throughput 9.4275K wps
[Epoch 108 Batch 90/162] avg loss 0.00143467, throughput 9.58358K wps
[Epoch 108 Batch 120/162] avg loss 0.00136403, throughput 9.37377K wps
[Epoch 108 Batch 150/162] avg loss 0.00154609, throughput 9.35953K wps
Begin Testing...
[Epoch 108] train avg loss 0.00145714, dev acc 0.9144, dev avg loss 0.22054, throughput 9.44533K wps
[Epoch 109 Batch 30/162] avg loss 0.00138341, throughput 9.4966K wps
[Epoch 109 Batch 60/162] avg loss 0.00148892, throughput 9.52737K wps
[Epoch 109 Batch 90/162] avg loss 0.00149885, throughput 9.40805K wps
[Epoch 109 Batch 120/162] avg loss 0.0014062, throughput 9.42508K wps
[Epoch 109 Batch 150/162] avg loss 0.00146541, throughput 9.40863K wps
Begin Testing...
[Epoch 109] train avg loss 0.00145382, dev acc 0.9144, dev avg loss 0.220203, throughput 9.43655K wps
[Epoch 110 Batch 30/162] avg loss 0.00140546, throughput 9.5131K wps
[Epoch 110 Batch 60/162] avg loss 0.0014141, throughput 9.40464K wps
[Epoch 110 Batch 90/162] avg loss 0.00144893, throughput 9.50049K wps
[Epoch 110 Batch 120/162] avg loss 0.00127264, throughput 9.4795K wps
[Epoch 110 Batch 150/162] avg loss 0.00156091, throughput 9.30454K wps
Begin Testing...
[Epoch 110] train avg loss 0.0014432, dev acc 0.9122, dev avg loss 0.219828, throughput 9.43764K wps
[Epoch 111 Batch 30/162] avg loss 0.00126812, throughput 9.76429K wps
[Epoch 111 Batch 60/162] avg loss 0.001481, throughput 9.26524K wps
[Epoch 111 Batch 90/162] avg loss 0.00142283, throughput 9.45379K wps
[Epoch 111 Batch 120/162] avg loss 0.00151125, throughput 9.47254K wps
[Epoch 111 Batch 150/162] avg loss 0.00161306, throughput 9.54763K wps
Begin Testing...
[Epoch 111] train avg loss 0.00145727, dev acc 0.9067, dev avg loss 0.219801, throughput 9.49981K wps
[Epoch 112 Batch 30/162] avg loss 0.00143015, throughput 9.77562K wps
[Epoch 112 Batch 60/162] avg loss 0.00126704, throughput 9.45097K wps
[Epoch 112 Batch 90/162] avg loss 0.00147221, throughput 9.66198K wps
[Epoch 112 Batch 120/162] avg loss 0.0014228, throughput 9.37787K wps
[Epoch 112 Batch 150/162] avg loss 0.00138302, throughput 9.35288K wps
Begin Testing...
[Epoch 112] train avg loss 0.00139075, dev acc 0.9100, dev avg loss 0.219645, throughput 9.52359K wps
[Epoch 113 Batch 30/162] avg loss 0.00151858, throughput 9.57525K wps
[Epoch 113 Batch 60/162] avg loss 0.0014457, throughput 9.59229K wps
[Epoch 113 Batch 90/162] avg loss 0.00128213, throughput 9.42381K wps
[Epoch 113 Batch 120/162] avg loss 0.00141966, throughput 9.50089K wps
[Epoch 113 Batch 150/162] avg loss 0.0015507, throughput 9.28234K wps
Begin Testing...
[Epoch 113] train avg loss 0.00142586, dev acc 0.9111, dev avg loss 0.219339, throughput 9.46573K wps
[Epoch 114 Batch 30/162] avg loss 0.00154838, throughput 9.45241K wps
[Epoch 114 Batch 60/162] avg loss 0.00123374, throughput 9.26755K wps
[Epoch 114 Batch 90/162] avg loss 0.00133265, throughput 9.49219K wps
[Epoch 114 Batch 120/162] avg loss 0.00139602, throughput 9.60441K wps
[Epoch 114 Batch 150/162] avg loss 0.00133207, throughput 9.39899K wps
Begin Testing...
[Epoch 114] train avg loss 0.00137247, dev acc 0.9078, dev avg loss 0.219903, throughput 9.42962K wps
[Epoch 115 Batch 30/162] avg loss 0.00144079, throughput 9.59114K wps
[Epoch 115 Batch 60/162] avg loss 0.00127496, throughput 9.58601K wps
[Epoch 115 Batch 90/162] avg loss 0.00140628, throughput 9.30951K wps
[Epoch 115 Batch 120/162] avg loss 0.00124593, throughput 9.30126K wps
[Epoch 115 Batch 150/162] avg loss 0.00133964, throughput 9.41743K wps
Begin Testing...
[Epoch 115] train avg loss 0.00133649, dev acc 0.9111, dev avg loss 0.2194, throughput 9.44693K wps
[Epoch 116 Batch 30/162] avg loss 0.00134933, throughput 9.56761K wps
[Epoch 116 Batch 60/162] avg loss 0.00132524, throughput 9.59585K wps
[Epoch 116 Batch 90/162] avg loss 0.00124989, throughput 9.58627K wps
[Epoch 116 Batch 120/162] avg loss 0.00133711, throughput 9.52803K wps
[Epoch 116 Batch 150/162] avg loss 0.00138527, throughput 9.39278K wps
Begin Testing...
[Epoch 116] train avg loss 0.00133463, dev acc 0.9133, dev avg loss 0.219502, throughput 9.52418K wps
[Epoch 117 Batch 30/162] avg loss 0.00131677, throughput 9.57254K wps
[Epoch 117 Batch 60/162] avg loss 0.00125643, throughput 9.32787K wps
[Epoch 117 Batch 90/162] avg loss 0.00131054, throughput 9.58304K wps
[Epoch 117 Batch 120/162] avg loss 0.00149667, throughput 9.5742K wps
[Epoch 117 Batch 150/162] avg loss 0.00131855, throughput 9.43964K wps
Begin Testing...
[Epoch 117] train avg loss 0.00135692, dev acc 0.9122, dev avg loss 0.219391, throughput 9.50832K wps
[Epoch 118 Batch 30/162] avg loss 0.00142302, throughput 9.76981K wps
[Epoch 118 Batch 60/162] avg loss 0.00128297, throughput 9.31237K wps
[Epoch 118 Batch 90/162] avg loss 0.001257, throughput 9.4077K wps
[Epoch 118 Batch 120/162] avg loss 0.00117628, throughput 9.50576K wps
[Epoch 118 Batch 150/162] avg loss 0.0014041, throughput 9.29767K wps
Begin Testing...
[Epoch 118] train avg loss 0.00132162, dev acc 0.9089, dev avg loss 0.218789, throughput 9.47114K wps
[Epoch 119 Batch 30/162] avg loss 0.00123781, throughput 9.47365K wps
[Epoch 119 Batch 60/162] avg loss 0.00127138, throughput 9.40968K wps
[Epoch 119 Batch 90/162] avg loss 0.00134882, throughput 9.34282K wps
[Epoch 119 Batch 120/162] avg loss 0.0013431, throughput 9.49522K wps
[Epoch 119 Batch 150/162] avg loss 0.00136837, throughput 9.59791K wps
Begin Testing...
[Epoch 119] train avg loss 0.00131736, dev acc 0.9144, dev avg loss 0.219578, throughput 9.45387K wps
[Epoch 120 Batch 30/162] avg loss 0.00131822, throughput 9.72831K wps
[Epoch 120 Batch 60/162] avg loss 0.00123827, throughput 9.33246K wps
[Epoch 120 Batch 90/162] avg loss 0.0012106, throughput 9.62128K wps
[Epoch 120 Batch 120/162] avg loss 0.00116641, throughput 9.59085K wps
[Epoch 120 Batch 150/162] avg loss 0.00126853, throughput 9.382K wps
Begin Testing...
[Epoch 120] train avg loss 0.00122656, dev acc 0.9078, dev avg loss 0.219664, throughput 9.53627K wps
[Epoch 121 Batch 30/162] avg loss 0.00118849, throughput 9.52357K wps
[Epoch 121 Batch 60/162] avg loss 0.00126839, throughput 9.3824K wps
[Epoch 121 Batch 90/162] avg loss 0.00133301, throughput 9.43849K wps
[Epoch 121 Batch 120/162] avg loss 0.00116984, throughput 9.44006K wps
[Epoch 121 Batch 150/162] avg loss 0.00122038, throughput 9.21095K wps
Begin Testing...
[Epoch 121] train avg loss 0.00123239, dev acc 0.9111, dev avg loss 0.219559, throughput 9.39174K wps
[Epoch 122 Batch 30/162] avg loss 0.00121644, throughput 9.75044K wps
[Epoch 122 Batch 60/162] avg loss 0.00128541, throughput 9.57868K wps
[Epoch 122 Batch 90/162] avg loss 0.00136915, throughput 9.28362K wps
[Epoch 122 Batch 120/162] avg loss 0.00126478, throughput 9.52206K wps
[Epoch 122 Batch 150/162] avg loss 0.00130148, throughput 9.61719K wps
Begin Testing...
[Epoch 122] train avg loss 0.00128584, dev acc 0.9156, dev avg loss 0.220542, throughput 9.5466K wps
Observed Improvement.
Begin Testing...
[Epoch 123 Batch 30/162] avg loss 0.00127959, throughput 9.56209K wps
[Epoch 123 Batch 60/162] avg loss 0.00116058, throughput 9.42487K wps
[Epoch 123 Batch 90/162] avg loss 0.00128762, throughput 9.62366K wps
[Epoch 123 Batch 120/162] avg loss 0.00122316, throughput 9.41996K wps
[Epoch 123 Batch 150/162] avg loss 0.00118883, throughput 9.48103K wps
Begin Testing...
[Epoch 123] train avg loss 0.00122761, dev acc 0.9144, dev avg loss 0.220538, throughput 9.51578K wps
[Epoch 124 Batch 30/162] avg loss 0.00132204, throughput 9.66663K wps
[Epoch 124 Batch 60/162] avg loss 0.00120255, throughput 9.39891K wps
[Epoch 124 Batch 90/162] avg loss 0.0012825, throughput 9.26954K wps
[Epoch 124 Batch 120/162] avg loss 0.00110025, throughput 9.46425K wps
[Epoch 124 Batch 150/162] avg loss 0.00126249, throughput 9.32674K wps
Begin Testing...
[Epoch 124] train avg loss 0.00122207, dev acc 0.9067, dev avg loss 0.21926, throughput 9.4117K wps
[Epoch 125 Batch 30/162] avg loss 0.00131609, throughput 9.65801K wps
[Epoch 125 Batch 60/162] avg loss 0.00113491, throughput 9.43369K wps
[Epoch 125 Batch 90/162] avg loss 0.00116254, throughput 9.36659K wps
[Epoch 125 Batch 120/162] avg loss 0.00111946, throughput 9.4092K wps
[Epoch 125 Batch 150/162] avg loss 0.00130979, throughput 9.49023K wps
Begin Testing...
[Epoch 125] train avg loss 0.001205, dev acc 0.9122, dev avg loss 0.219927, throughput 9.46099K wps
[Epoch 126 Batch 30/162] avg loss 0.00112208, throughput 9.56578K wps
[Epoch 126 Batch 60/162] avg loss 0.00120425, throughput 9.49972K wps
[Epoch 126 Batch 90/162] avg loss 0.00115062, throughput 9.42056K wps
[Epoch 126 Batch 120/162] avg loss 0.00117599, throughput 9.38923K wps
[Epoch 126 Batch 150/162] avg loss 0.00112888, throughput 9.4137K wps
Begin Testing...
[Epoch 126] train avg loss 0.00114657, dev acc 0.9078, dev avg loss 0.220101, throughput 9.46217K wps
[Epoch 127 Batch 30/162] avg loss 0.00125397, throughput 9.69285K wps
[Epoch 127 Batch 60/162] avg loss 0.00123518, throughput 9.52042K wps
[Epoch 127 Batch 90/162] avg loss 0.00115117, throughput 9.53733K wps
[Epoch 127 Batch 120/162] avg loss 0.00116034, throughput 9.59682K wps
[Epoch 127 Batch 150/162] avg loss 0.00120292, throughput 9.49642K wps
Begin Testing...
[Epoch 127] train avg loss 0.00119446, dev acc 0.9078, dev avg loss 0.219672, throughput 9.54187K wps
[Epoch 128 Batch 30/162] avg loss 0.0012921, throughput 9.44665K wps
[Epoch 128 Batch 60/162] avg loss 0.00107005, throughput 9.40098K wps
[Epoch 128 Batch 90/162] avg loss 0.0011908, throughput 9.46992K wps
[Epoch 128 Batch 120/162] avg loss 0.00119708, throughput 9.59131K wps
[Epoch 128 Batch 150/162] avg loss 0.0011899, throughput 9.39935K wps
Begin Testing...
[Epoch 128] train avg loss 0.00116649, dev acc 0.9111, dev avg loss 0.219769, throughput 9.47267K wps
[Epoch 129 Batch 30/162] avg loss 0.00124188, throughput 9.60468K wps
[Epoch 129 Batch 60/162] avg loss 0.00121275, throughput 9.41388K wps
[Epoch 129 Batch 90/162] avg loss 0.00117834, throughput 9.45437K wps
[Epoch 129 Batch 120/162] avg loss 0.00110095, throughput 9.40895K wps
[Epoch 129 Batch 150/162] avg loss 0.00111869, throughput 9.24687K wps
Begin Testing...
[Epoch 129] train avg loss 0.00115475, dev acc 0.9089, dev avg loss 0.219779, throughput 9.44739K wps
[Epoch 130 Batch 30/162] avg loss 0.00118842, throughput 9.55506K wps
[Epoch 130 Batch 60/162] avg loss 0.00124075, throughput 9.45539K wps
[Epoch 130 Batch 90/162] avg loss 0.00104706, throughput 9.34926K wps
[Epoch 130 Batch 120/162] avg loss 0.00127253, throughput 9.34936K wps
[Epoch 130 Batch 150/162] avg loss 0.00123864, throughput 9.3919K wps
Begin Testing...
[Epoch 130] train avg loss 0.00119707, dev acc 0.9122, dev avg loss 0.219851, throughput 9.44479K wps
[Epoch 131 Batch 30/162] avg loss 0.00108787, throughput 9.65318K wps
[Epoch 131 Batch 60/162] avg loss 0.00119411, throughput 9.6316K wps
[Epoch 131 Batch 90/162] avg loss 0.0011501, throughput 9.52952K wps
[Epoch 131 Batch 120/162] avg loss 0.00103977, throughput 9.37207K wps
[Epoch 131 Batch 150/162] avg loss 0.0011603, throughput 9.3998K wps
Begin Testing...
[Epoch 131] train avg loss 0.00113879, dev acc 0.9078, dev avg loss 0.219764, throughput 9.48907K wps
[Epoch 132 Batch 30/162] avg loss 0.0010541, throughput 9.65972K wps
[Epoch 132 Batch 60/162] avg loss 0.0011474, throughput 9.40473K wps
[Epoch 132 Batch 90/162] avg loss 0.000971857, throughput 9.5115K wps
[Epoch 132 Batch 120/162] avg loss 0.00115822, throughput 9.55887K wps
[Epoch 132 Batch 150/162] avg loss 0.00122277, throughput 9.61346K wps
Begin Testing...
[Epoch 132] train avg loss 0.00110062, dev acc 0.9067, dev avg loss 0.220075, throughput 9.5485K wps
[Epoch 133 Batch 30/162] avg loss 0.00106813, throughput 9.53076K wps
[Epoch 133 Batch 60/162] avg loss 0.00117516, throughput 9.63114K wps
[Epoch 133 Batch 90/162] avg loss 0.00111729, throughput 9.27934K wps
[Epoch 133 Batch 120/162] avg loss 0.00110987, throughput 9.39327K wps
[Epoch 133 Batch 150/162] avg loss 0.00109996, throughput 9.44437K wps
Begin Testing...
[Epoch 133] train avg loss 0.00110224, dev acc 0.9111, dev avg loss 0.220022, throughput 9.44338K wps
[Epoch 134 Batch 30/162] avg loss 0.00113081, throughput 9.54194K wps
[Epoch 134 Batch 60/162] avg loss 0.00115107, throughput 9.36903K wps
[Epoch 134 Batch 90/162] avg loss 0.00100009, throughput 9.3978K wps
[Epoch 134 Batch 120/162] avg loss 0.00099557, throughput 9.51164K wps
[Epoch 134 Batch 150/162] avg loss 0.0011232, throughput 9.48115K wps
Begin Testing...
[Epoch 134] train avg loss 0.00107355, dev acc 0.9122, dev avg loss 0.220098, throughput 9.4545K wps
[Epoch 135 Batch 30/162] avg loss 0.00107626, throughput 9.45257K wps
[Epoch 135 Batch 60/162] avg loss 0.00107904, throughput 9.33924K wps
[Epoch 135 Batch 90/162] avg loss 0.00105174, throughput 9.62181K wps
[Epoch 135 Batch 120/162] avg loss 0.0010537, throughput 9.37143K wps
[Epoch 135 Batch 150/162] avg loss 0.00110219, throughput 9.52821K wps
Begin Testing...
[Epoch 135] train avg loss 0.00105428, dev acc 0.9067, dev avg loss 0.220296, throughput 9.4559K wps
[Epoch 136 Batch 30/162] avg loss 0.000933952, throughput 9.55002K wps
[Epoch 136 Batch 60/162] avg loss 0.00102818, throughput 9.47346K wps
[Epoch 136 Batch 90/162] avg loss 0.00108939, throughput 9.54636K wps
[Epoch 136 Batch 120/162] avg loss 0.00105633, throughput 9.4461K wps
[Epoch 136 Batch 150/162] avg loss 0.00103049, throughput 9.54069K wps
Begin Testing...
[Epoch 136] train avg loss 0.00102739, dev acc 0.9078, dev avg loss 0.220204, throughput 9.51216K wps
[Epoch 137 Batch 30/162] avg loss 0.00110266, throughput 9.69864K wps
[Epoch 137 Batch 60/162] avg loss 0.000973101, throughput 9.40466K wps
[Epoch 137 Batch 90/162] avg loss 0.0010731, throughput 9.30898K wps
[Epoch 137 Batch 120/162] avg loss 0.00126709, throughput 9.36971K wps
[Epoch 137 Batch 150/162] avg loss 0.000987691, throughput 9.35402K wps
Begin Testing...
[Epoch 137] train avg loss 0.00109909, dev acc 0.9078, dev avg loss 0.220081, throughput 9.42002K wps
[Epoch 138 Batch 30/162] avg loss 0.00089346, throughput 9.5258K wps
[Epoch 138 Batch 60/162] avg loss 0.00114187, throughput 9.32363K wps
[Epoch 138 Batch 90/162] avg loss 0.00102766, throughput 9.40083K wps
[Epoch 138 Batch 120/162] avg loss 0.00110682, throughput 9.33673K wps
[Epoch 138 Batch 150/162] avg loss 0.00107044, throughput 9.36919K wps
Begin Testing...
[Epoch 138] train avg loss 0.00104644, dev acc 0.9111, dev avg loss 0.219958, throughput 9.37719K wps
[Epoch 139 Batch 30/162] avg loss 0.00104312, throughput 9.71104K wps
[Epoch 139 Batch 60/162] avg loss 0.0012444, throughput 9.43846K wps
[Epoch 139 Batch 90/162] avg loss 0.00109496, throughput 9.40156K wps
[Epoch 139 Batch 120/162] avg loss 0.00100424, throughput 9.37954K wps
[Epoch 139 Batch 150/162] avg loss 0.000900157, throughput 9.34134K wps
Begin Testing...
[Epoch 139] train avg loss 0.00106394, dev acc 0.9078, dev avg loss 0.219874, throughput 9.44352K wps
[Epoch 140 Batch 30/162] avg loss 0.000968843, throughput 9.77129K wps
[Epoch 140 Batch 60/162] avg loss 0.000931253, throughput 9.51714K wps
[Epoch 140 Batch 90/162] avg loss 0.000988257, throughput 9.57002K wps
[Epoch 140 Batch 120/162] avg loss 0.000940298, throughput 9.40069K wps
[Epoch 140 Batch 150/162] avg loss 0.00111845, throughput 9.34752K wps
Begin Testing...
[Epoch 140] train avg loss 0.000988854, dev acc 0.9133, dev avg loss 0.221037, throughput 9.50292K wps
[Epoch 141 Batch 30/162] avg loss 0.00108098, throughput 9.40264K wps
[Epoch 141 Batch 60/162] avg loss 0.00103573, throughput 9.32846K wps
[Epoch 141 Batch 90/162] avg loss 0.000877055, throughput 9.5338K wps
[Epoch 141 Batch 120/162] avg loss 0.00104161, throughput 9.32561K wps
[Epoch 141 Batch 150/162] avg loss 0.00103155, throughput 9.47594K wps
Begin Testing...
[Epoch 141] train avg loss 0.00101487, dev acc 0.9078, dev avg loss 0.220254, throughput 9.42946K wps
[Epoch 142 Batch 30/162] avg loss 0.000981613, throughput 9.56863K wps
[Epoch 142 Batch 60/162] avg loss 0.00103449, throughput 9.50074K wps
[Epoch 142 Batch 90/162] avg loss 0.00111434, throughput 9.23638K wps
[Epoch 142 Batch 120/162] avg loss 0.000987619, throughput 9.38717K wps
[Epoch 142 Batch 150/162] avg loss 0.000886862, throughput 9.32943K wps
Begin Testing...
[Epoch 142] train avg loss 0.000998839, dev acc 0.9078, dev avg loss 0.220178, throughput 9.40887K wps
[Epoch 143 Batch 30/162] avg loss 0.00101698, throughput 9.79183K wps
[Epoch 143 Batch 60/162] avg loss 0.00103757, throughput 9.35397K wps
[Epoch 143 Batch 90/162] avg loss 0.000886873, throughput 9.54378K wps
[Epoch 143 Batch 120/162] avg loss 0.000943636, throughput 9.54642K wps
[Epoch 143 Batch 150/162] avg loss 0.000882586, throughput 9.39435K wps
Begin Testing...
[Epoch 143] train avg loss 0.000943254, dev acc 0.9078, dev avg loss 0.22051, throughput 9.50232K wps
[Epoch 144 Batch 30/162] avg loss 0.000930551, throughput 9.45535K wps
[Epoch 144 Batch 60/162] avg loss 0.00105166, throughput 9.34576K wps
[Epoch 144 Batch 90/162] avg loss 0.0010426, throughput 9.27998K wps
[Epoch 144 Batch 120/162] avg loss 0.000907255, throughput 9.32557K wps
[Epoch 144 Batch 150/162] avg loss 0.000960305, throughput 9.39966K wps
Begin Testing...
[Epoch 144] train avg loss 0.000975253, dev acc 0.9133, dev avg loss 0.220997, throughput 9.35999K wps
[Epoch 145 Batch 30/162] avg loss 0.000933937, throughput 9.68351K wps
[Epoch 145 Batch 60/162] avg loss 0.000873202, throughput 9.51818K wps
[Epoch 145 Batch 90/162] avg loss 0.000882687, throughput 9.45972K wps
[Epoch 145 Batch 120/162] avg loss 0.00108388, throughput 9.4953K wps
[Epoch 145 Batch 150/162] avg loss 0.000962535, throughput 9.35097K wps
Begin Testing...
[Epoch 145] train avg loss 0.000945921, dev acc 0.9133, dev avg loss 0.22144, throughput 9.51604K wps
[Epoch 146 Batch 30/162] avg loss 0.000900444, throughput 9.79411K wps
[Epoch 146 Batch 60/162] avg loss 0.000915936, throughput 9.46413K wps
[Epoch 146 Batch 90/162] avg loss 0.00093405, throughput 9.49961K wps
[Epoch 146 Batch 120/162] avg loss 0.000969471, throughput 9.33601K wps
[Epoch 146 Batch 150/162] avg loss 0.000880569, throughput 9.28886K wps
Begin Testing...
[Epoch 146] train avg loss 0.00091543, dev acc 0.9078, dev avg loss 0.221498, throughput 9.49154K wps
[Epoch 147 Batch 30/162] avg loss 0.0010145, throughput 9.65614K wps
[Epoch 147 Batch 60/162] avg loss 0.00102354, throughput 9.30678K wps
[Epoch 147 Batch 90/162] avg loss 0.00092605, throughput 9.34924K wps
[Epoch 147 Batch 120/162] avg loss 0.000888605, throughput 9.53988K wps
[Epoch 147 Batch 150/162] avg loss 0.000935584, throughput 9.60056K wps
Begin Testing...
[Epoch 147] train avg loss 0.000963579, dev acc 0.9111, dev avg loss 0.220958, throughput 9.49438K wps
[Epoch 148 Batch 30/162] avg loss 0.000797505, throughput 9.57573K wps
[Epoch 148 Batch 60/162] avg loss 0.000957491, throughput 9.33564K wps
[Epoch 148 Batch 90/162] avg loss 0.000911221, throughput 9.4911K wps
[Epoch 148 Batch 120/162] avg loss 0.00104742, throughput 9.59456K wps
[Epoch 148 Batch 150/162] avg loss 0.000934006, throughput 9.32693K wps
Begin Testing...
[Epoch 148] train avg loss 0.000927827, dev acc 0.9089, dev avg loss 0.220843, throughput 9.4576K wps
[Epoch 149 Batch 30/162] avg loss 0.000910532, throughput 9.74795K wps
[Epoch 149 Batch 60/162] avg loss 0.000856237, throughput 9.51833K wps
[Epoch 149 Batch 90/162] avg loss 0.000956357, throughput 9.56239K wps
[Epoch 149 Batch 120/162] avg loss 0.00100068, throughput 9.33471K wps
[Epoch 149 Batch 150/162] avg loss 0.000847839, throughput 9.37958K wps
Begin Testing...
[Epoch 149] train avg loss 0.000904084, dev acc 0.9122, dev avg loss 0.22173, throughput 9.49518K wps
[Epoch 150 Batch 30/162] avg loss 0.000925322, throughput 9.69069K wps
[Epoch 150 Batch 60/162] avg loss 0.000879622, throughput 9.40093K wps
[Epoch 150 Batch 90/162] avg loss 0.000916732, throughput 9.34217K wps
[Epoch 150 Batch 120/162] avg loss 0.000805112, throughput 9.5748K wps
[Epoch 150 Batch 150/162] avg loss 0.00093403, throughput 9.33849K wps
Begin Testing...
[Epoch 150] train avg loss 0.000891882, dev acc 0.9122, dev avg loss 0.221685, throughput 9.45923K wps
[Epoch 151 Batch 30/162] avg loss 0.00087063, throughput 9.6895K wps
[Epoch 151 Batch 60/162] avg loss 0.000898002, throughput 9.41477K wps
[Epoch 151 Batch 90/162] avg loss 0.000932744, throughput 9.57455K wps
[Epoch 151 Batch 120/162] avg loss 0.000937104, throughput 9.42937K wps
[Epoch 151 Batch 150/162] avg loss 0.000879115, throughput 9.36753K wps
Begin Testing...
[Epoch 151] train avg loss 0.000902502, dev acc 0.9111, dev avg loss 0.221468, throughput 9.5055K wps
[Epoch 152 Batch 30/162] avg loss 0.000946764, throughput 9.72877K wps
[Epoch 152 Batch 60/162] avg loss 0.000906099, throughput 9.13762K wps
[Epoch 152 Batch 90/162] avg loss 0.000928593, throughput 9.63237K wps
[Epoch 152 Batch 120/162] avg loss 0.000942469, throughput 9.54023K wps
[Epoch 152 Batch 150/162] avg loss 0.000816662, throughput 9.4516K wps
Begin Testing...
[Epoch 152] train avg loss 0.000904355, dev acc 0.9111, dev avg loss 0.221937, throughput 9.4925K wps
[Epoch 153 Batch 30/162] avg loss 0.000906176, throughput 9.48927K wps
[Epoch 153 Batch 60/162] avg loss 0.000926806, throughput 9.41031K wps
[Epoch 153 Batch 90/162] avg loss 0.000892875, throughput 9.32381K wps
[Epoch 153 Batch 120/162] avg loss 0.000836417, throughput 9.39694K wps
[Epoch 153 Batch 150/162] avg loss 0.000779267, throughput 9.40247K wps
Begin Testing...
[Epoch 153] train avg loss 0.000863946, dev acc 0.9156, dev avg loss 0.22433, throughput 9.39799K wps
Observed Improvement.
Begin Testing...
[Epoch 154 Batch 30/162] avg loss 0.000815208, throughput 9.56522K wps
[Epoch 154 Batch 60/162] avg loss 0.00082075, throughput 9.26665K wps
[Epoch 154 Batch 90/162] avg loss 0.000982625, throughput 9.35945K wps
[Epoch 154 Batch 120/162] avg loss 0.000809518, throughput 9.48083K wps
[Epoch 154 Batch 150/162] avg loss 0.000871546, throughput 9.54019K wps
Begin Testing...
[Epoch 154] train avg loss 0.000854411, dev acc 0.9133, dev avg loss 0.222581, throughput 9.45378K wps
[Epoch 155 Batch 30/162] avg loss 0.000798028, throughput 9.70175K wps
[Epoch 155 Batch 60/162] avg loss 0.000842432, throughput 9.54731K wps
[Epoch 155 Batch 90/162] avg loss 0.000917182, throughput 9.24669K wps
[Epoch 155 Batch 120/162] avg loss 0.000925908, throughput 9.35276K wps
[Epoch 155 Batch 150/162] avg loss 0.000793847, throughput 9.28849K wps
Begin Testing...
[Epoch 155] train avg loss 0.000860058, dev acc 0.9133, dev avg loss 0.221923, throughput 9.42126K wps
[Epoch 156 Batch 30/162] avg loss 0.00083538, throughput 9.54522K wps
[Epoch 156 Batch 60/162] avg loss 0.000976914, throughput 9.40021K wps
[Epoch 156 Batch 90/162] avg loss 0.000816123, throughput 9.41036K wps
[Epoch 156 Batch 120/162] avg loss 0.000843734, throughput 9.24142K wps
[Epoch 156 Batch 150/162] avg loss 0.000986658, throughput 9.47231K wps
Begin Testing...
[Epoch 156] train avg loss 0.00088194, dev acc 0.9089, dev avg loss 0.221888, throughput 9.4116K wps
[Epoch 157 Batch 30/162] avg loss 0.000933942, throughput 9.48427K wps
[Epoch 157 Batch 60/162] avg loss 0.000740034, throughput 9.53995K wps
[Epoch 157 Batch 90/162] avg loss 0.000919311, throughput 9.41147K wps
[Epoch 157 Batch 120/162] avg loss 0.000820311, throughput 9.30493K wps
[Epoch 157 Batch 150/162] avg loss 0.000855674, throughput 9.4801K wps
Begin Testing...
[Epoch 157] train avg loss 0.000849839, dev acc 0.9078, dev avg loss 0.222877, throughput 9.44435K wps
[Epoch 158 Batch 30/162] avg loss 0.000770352, throughput 9.48617K wps
[Epoch 158 Batch 60/162] avg loss 0.00089964, throughput 9.40175K wps
[Epoch 158 Batch 90/162] avg loss 0.000905003, throughput 9.42491K wps
[Epoch 158 Batch 120/162] avg loss 0.000916712, throughput 9.30171K wps
[Epoch 158 Batch 150/162] avg loss 0.000755963, throughput 9.34157K wps
Begin Testing...
[Epoch 158] train avg loss 0.00084313, dev acc 0.9100, dev avg loss 0.222999, throughput 9.38416K wps
[Epoch 159 Batch 30/162] avg loss 0.000769246, throughput 9.48178K wps
[Epoch 159 Batch 60/162] avg loss 0.000846508, throughput 9.54082K wps
[Epoch 159 Batch 90/162] avg loss 0.000799635, throughput 9.52613K wps
[Epoch 159 Batch 120/162] avg loss 0.000787537, throughput 9.41116K wps
[Epoch 159 Batch 150/162] avg loss 0.00084803, throughput 9.20089K wps
Begin Testing...
[Epoch 159] train avg loss 0.0008172, dev acc 0.9078, dev avg loss 0.223053, throughput 9.42913K wps
[Epoch 160 Batch 30/162] avg loss 0.000854948, throughput 9.8216K wps
[Epoch 160 Batch 60/162] avg loss 0.000758233, throughput 9.36114K wps
[Epoch 160 Batch 90/162] avg loss 0.000908488, throughput 9.55161K wps
[Epoch 160 Batch 120/162] avg loss 0.000856275, throughput 9.28267K wps
[Epoch 160 Batch 150/162] avg loss 0.000853466, throughput 9.33319K wps
Begin Testing...
[Epoch 160] train avg loss 0.000839406, dev acc 0.9067, dev avg loss 0.223043, throughput 9.47226K wps
[Epoch 161 Batch 30/162] avg loss 0.000861797, throughput 9.77544K wps
[Epoch 161 Batch 60/162] avg loss 0.000856503, throughput 9.33892K wps
[Epoch 161 Batch 90/162] avg loss 0.000818889, throughput 9.63201K wps
[Epoch 161 Batch 120/162] avg loss 0.000854773, throughput 9.23421K wps
[Epoch 161 Batch 150/162] avg loss 0.000757938, throughput 9.47317K wps
Begin Testing...
[Epoch 161] train avg loss 0.000833005, dev acc 0.9122, dev avg loss 0.223081, throughput 9.47589K wps
[Epoch 162 Batch 30/162] avg loss 0.000794625, throughput 9.59568K wps
[Epoch 162 Batch 60/162] avg loss 0.000718354, throughput 9.41641K wps
[Epoch 162 Batch 90/162] avg loss 0.000866617, throughput 9.46003K wps
[Epoch 162 Batch 120/162] avg loss 0.000895189, throughput 9.41285K wps
[Epoch 162 Batch 150/162] avg loss 0.000800907, throughput 9.35352K wps
Begin Testing...
[Epoch 162] train avg loss 0.000807685, dev acc 0.9133, dev avg loss 0.222284, throughput 9.43837K wps
[Epoch 163 Batch 30/162] avg loss 0.00087021, throughput 9.41063K wps
[Epoch 163 Batch 60/162] avg loss 0.000801092, throughput 9.34096K wps
[Epoch 163 Batch 90/162] avg loss 0.000800272, throughput 9.38658K wps
[Epoch 163 Batch 120/162] avg loss 0.000874332, throughput 9.48356K wps
[Epoch 163 Batch 150/162] avg loss 0.000775821, throughput 9.54883K wps
Begin Testing...
[Epoch 163] train avg loss 0.000828231, dev acc 0.9122, dev avg loss 0.223332, throughput 9.43476K wps
[Epoch 164 Batch 30/162] avg loss 0.000843145, throughput 9.50985K wps
[Epoch 164 Batch 60/162] avg loss 0.000752993, throughput 9.31233K wps
[Epoch 164 Batch 90/162] avg loss 0.000722002, throughput 9.31879K wps
[Epoch 164 Batch 120/162] avg loss 0.000729957, throughput 9.44813K wps
[Epoch 164 Batch 150/162] avg loss 0.00078387, throughput 9.51731K wps
Begin Testing...
[Epoch 164] train avg loss 0.000780063, dev acc 0.9122, dev avg loss 0.223195, throughput 9.42575K wps
[Epoch 165 Batch 30/162] avg loss 0.000845348, throughput 9.76513K wps
[Epoch 165 Batch 60/162] avg loss 0.000766849, throughput 9.42669K wps
[Epoch 165 Batch 90/162] avg loss 0.000736091, throughput 9.5486K wps
[Epoch 165 Batch 120/162] avg loss 0.000821382, throughput 9.44054K wps
[Epoch 165 Batch 150/162] avg loss 0.000802639, throughput 9.55611K wps
Begin Testing...
[Epoch 165] train avg loss 0.000800666, dev acc 0.9144, dev avg loss 0.224786, throughput 9.54308K wps
[Epoch 166 Batch 30/162] avg loss 0.000760267, throughput 9.48677K wps
[Epoch 166 Batch 60/162] avg loss 0.000787835, throughput 9.39914K wps
[Epoch 166 Batch 90/162] avg loss 0.000738461, throughput 9.2906K wps
[Epoch 166 Batch 120/162] avg loss 0.000751568, throughput 9.18553K wps
[Epoch 166 Batch 150/162] avg loss 0.000781036, throughput 9.36325K wps
Begin Testing...
[Epoch 166] train avg loss 0.000771873, dev acc 0.9100, dev avg loss 0.223033, throughput 9.33552K wps
[Epoch 167 Batch 30/162] avg loss 0.00072483, throughput 9.64846K wps
[Epoch 167 Batch 60/162] avg loss 0.000792149, throughput 9.45172K wps
[Epoch 167 Batch 90/162] avg loss 0.000784366, throughput 9.3525K wps
[Epoch 167 Batch 120/162] avg loss 0.000792023, throughput 9.39908K wps
[Epoch 167 Batch 150/162] avg loss 0.000764713, throughput 9.47535K wps
Begin Testing...
[Epoch 167] train avg loss 0.000764733, dev acc 0.9100, dev avg loss 0.223517, throughput 9.44372K wps
[Epoch 168 Batch 30/162] avg loss 0.000768847, throughput 9.61479K wps
[Epoch 168 Batch 60/162] avg loss 0.000705714, throughput 9.52559K wps
[Epoch 168 Batch 90/162] avg loss 0.000783796, throughput 9.44379K wps
[Epoch 168 Batch 120/162] avg loss 0.000822262, throughput 9.41324K wps
[Epoch 168 Batch 150/162] avg loss 0.000763883, throughput 9.46684K wps
Begin Testing...
[Epoch 168] train avg loss 0.000763622, dev acc 0.9089, dev avg loss 0.223641, throughput 9.49832K wps
[Epoch 169 Batch 30/162] avg loss 0.000749442, throughput 9.6186K wps
[Epoch 169 Batch 60/162] avg loss 0.000700399, throughput 9.25345K wps
[Epoch 169 Batch 90/162] avg loss 0.000768257, throughput 9.47267K wps
[Epoch 169 Batch 120/162] avg loss 0.000788524, throughput 9.39892K wps
[Epoch 169 Batch 150/162] avg loss 0.000705779, throughput 9.36164K wps
Begin Testing...
[Epoch 169] train avg loss 0.000742662, dev acc 0.9122, dev avg loss 0.223275, throughput 9.40778K wps
[Epoch 170 Batch 30/162] avg loss 0.000701667, throughput 9.49988K wps
[Epoch 170 Batch 60/162] avg loss 0.000720557, throughput 9.48114K wps
[Epoch 170 Batch 90/162] avg loss 0.000679531, throughput 9.23471K wps
[Epoch 170 Batch 120/162] avg loss 0.000781517, throughput 9.38415K wps
[Epoch 170 Batch 150/162] avg loss 0.000826297, throughput 9.28661K wps
Begin Testing...
[Epoch 170] train avg loss 0.000752575, dev acc 0.9111, dev avg loss 0.223348, throughput 9.37031K wps
[Epoch 171 Batch 30/162] avg loss 0.000664671, throughput 9.67465K wps
[Epoch 171 Batch 60/162] avg loss 0.000714441, throughput 9.27813K wps
[Epoch 171 Batch 90/162] avg loss 0.000702378, throughput 9.16436K wps
[Epoch 171 Batch 120/162] avg loss 0.000710784, throughput 9.46021K wps
[Epoch 171 Batch 150/162] avg loss 0.00076889, throughput 9.48005K wps
Begin Testing...
[Epoch 171] train avg loss 0.000710598, dev acc 0.9122, dev avg loss 0.224165, throughput 9.3986K wps
[Epoch 172 Batch 30/162] avg loss 0.000786989, throughput 9.54528K wps
[Epoch 172 Batch 60/162] avg loss 0.000771742, throughput 9.43845K wps
[Epoch 172 Batch 90/162] avg loss 0.000726567, throughput 9.31909K wps
[Epoch 172 Batch 120/162] avg loss 0.000733667, throughput 9.60964K wps
[Epoch 172 Batch 150/162] avg loss 0.00078039, throughput 9.29542K wps
Begin Testing...
[Epoch 172] train avg loss 0.000757232, dev acc 0.9111, dev avg loss 0.223761, throughput 9.44519K wps
[Epoch 173 Batch 30/162] avg loss 0.000665419, throughput 9.78356K wps
[Epoch 173 Batch 60/162] avg loss 0.000704815, throughput 9.42306K wps
[Epoch 173 Batch 90/162] avg loss 0.000718261, throughput 9.44861K wps
[Epoch 173 Batch 120/162] avg loss 0.000707626, throughput 9.51653K wps
[Epoch 173 Batch 150/162] avg loss 0.000731501, throughput 9.49691K wps
Begin Testing...
[Epoch 173] train avg loss 0.000713299, dev acc 0.9122, dev avg loss 0.224277, throughput 9.52546K wps
[Epoch 174 Batch 30/162] avg loss 0.000721294, throughput 9.58754K wps
[Epoch 174 Batch 60/162] avg loss 0.000663579, throughput 9.5159K wps
[Epoch 174 Batch 90/162] avg loss 0.00068606, throughput 9.27005K wps
[Epoch 174 Batch 120/162] avg loss 0.000833909, throughput 9.32777K wps
[Epoch 174 Batch 150/162] avg loss 0.000684831, throughput 9.46697K wps
Begin Testing...
[Epoch 174] train avg loss 0.000718586, dev acc 0.9111, dev avg loss 0.224449, throughput 9.43909K wps
[Epoch 175 Batch 30/162] avg loss 0.00069632, throughput 9.50464K wps
[Epoch 175 Batch 60/162] avg loss 0.000740415, throughput 9.47173K wps
[Epoch 175 Batch 90/162] avg loss 0.00076376, throughput 9.34854K wps
[Epoch 175 Batch 120/162] avg loss 0.000645485, throughput 9.48883K wps
[Epoch 175 Batch 150/162] avg loss 0.000727642, throughput 9.58931K wps
Begin Testing...
[Epoch 175] train avg loss 0.000707291, dev acc 0.9100, dev avg loss 0.224365, throughput 9.4742K wps
[Epoch 176 Batch 30/162] avg loss 0.000682724, throughput 9.69655K wps
[Epoch 176 Batch 60/162] avg loss 0.000664758, throughput 9.33396K wps
[Epoch 176 Batch 90/162] avg loss 0.000654223, throughput 9.36184K wps
[Epoch 176 Batch 120/162] avg loss 0.000704454, throughput 9.30096K wps
[Epoch 176 Batch 150/162] avg loss 0.000798115, throughput 9.52817K wps
Begin Testing...
[Epoch 176] train avg loss 0.00070601, dev acc 0.9122, dev avg loss 0.224308, throughput 9.42879K wps
[Epoch 177 Batch 30/162] avg loss 0.000741427, throughput 9.63975K wps
[Epoch 177 Batch 60/162] avg loss 0.000774541, throughput 9.30244K wps
[Epoch 177 Batch 90/162] avg loss 0.000772171, throughput 9.54328K wps
[Epoch 177 Batch 120/162] avg loss 0.000650891, throughput 9.66345K wps
[Epoch 177 Batch 150/162] avg loss 0.000814629, throughput 9.48579K wps
Begin Testing...
[Epoch 177] train avg loss 0.000753663, dev acc 0.9122, dev avg loss 0.224093, throughput 9.51359K wps
[Epoch 178 Batch 30/162] avg loss 0.00072446, throughput 9.62437K wps
[Epoch 178 Batch 60/162] avg loss 0.000704409, throughput 9.23785K wps
[Epoch 178 Batch 90/162] avg loss 0.000674672, throughput 9.16163K wps
[Epoch 178 Batch 120/162] avg loss 0.000696556, throughput 9.50957K wps
[Epoch 178 Batch 150/162] avg loss 0.000709595, throughput 9.45345K wps
Begin Testing...
[Epoch 178] train avg loss 0.000690072, dev acc 0.9111, dev avg loss 0.224673, throughput 9.37504K wps
[Epoch 179 Batch 30/162] avg loss 0.0006362, throughput 9.60819K wps
[Epoch 179 Batch 60/162] avg loss 0.000723295, throughput 9.34905K wps
[Epoch 179 Batch 90/162] avg loss 0.000690111, throughput 9.48363K wps
[Epoch 179 Batch 120/162] avg loss 0.000699419, throughput 9.56811K wps
[Epoch 179 Batch 150/162] avg loss 0.000572189, throughput 9.37589K wps
Begin Testing...
[Epoch 179] train avg loss 0.00065954, dev acc 0.9111, dev avg loss 0.225463, throughput 9.46889K wps
[Epoch 180 Batch 30/162] avg loss 0.00056576, throughput 9.49448K wps
[Epoch 180 Batch 60/162] avg loss 0.000743239, throughput 9.51438K wps
[Epoch 180 Batch 90/162] avg loss 0.000818875, throughput 9.38031K wps
[Epoch 180 Batch 120/162] avg loss 0.000722525, throughput 9.47721K wps
[Epoch 180 Batch 150/162] avg loss 0.000669385, throughput 9.23761K wps
Begin Testing...
[Epoch 180] train avg loss 0.000704861, dev acc 0.9089, dev avg loss 0.225008, throughput 9.43439K wps
[Epoch 181 Batch 30/162] avg loss 0.000720951, throughput 9.59027K wps
[Epoch 181 Batch 60/162] avg loss 0.000690462, throughput 9.38114K wps
[Epoch 181 Batch 90/162] avg loss 0.000652368, throughput 9.46633K wps
[Epoch 181 Batch 120/162] avg loss 0.000736351, throughput 9.38981K wps
[Epoch 181 Batch 150/162] avg loss 0.000667496, throughput 9.37542K wps
Begin Testing...
[Epoch 181] train avg loss 0.000696991, dev acc 0.9122, dev avg loss 0.224918, throughput 9.42867K wps
[Epoch 182 Batch 30/162] avg loss 0.000589676, throughput 9.6333K wps
[Epoch 182 Batch 60/162] avg loss 0.000617912, throughput 9.25714K wps
[Epoch 182 Batch 90/162] avg loss 0.00071039, throughput 9.64251K wps
[Epoch 182 Batch 120/162] avg loss 0.000691865, throughput 9.28949K wps
[Epoch 182 Batch 150/162] avg loss 0.000658735, throughput 9.50474K wps
Begin Testing...
[Epoch 182] train avg loss 0.000660522, dev acc 0.9133, dev avg loss 0.225335, throughput 9.4447K wps
[Epoch 183 Batch 30/162] avg loss 0.000638562, throughput 9.73458K wps
[Epoch 183 Batch 60/162] avg loss 0.000661144, throughput 9.60771K wps
[Epoch 183 Batch 90/162] avg loss 0.00062944, throughput 9.48284K wps
[Epoch 183 Batch 120/162] avg loss 0.00078073, throughput 9.36984K wps
[Epoch 183 Batch 150/162] avg loss 0.00063171, throughput 9.61737K wps
Begin Testing...
[Epoch 183] train avg loss 0.000667173, dev acc 0.9100, dev avg loss 0.225294, throughput 9.56732K wps
[Epoch 184 Batch 30/162] avg loss 0.000714811, throughput 9.57889K wps
[Epoch 184 Batch 60/162] avg loss 0.000751302, throughput 9.26691K wps
[Epoch 184 Batch 90/162] avg loss 0.000651159, throughput 9.46978K wps
[Epoch 184 Batch 120/162] avg loss 0.000705927, throughput 9.59071K wps
[Epoch 184 Batch 150/162] avg loss 0.000595798, throughput 9.39618K wps
Begin Testing...
[Epoch 184] train avg loss 0.00068942, dev acc 0.9111, dev avg loss 0.225141, throughput 9.4561K wps
[Epoch 185 Batch 30/162] avg loss 0.000693823, throughput 9.59538K wps
[Epoch 185 Batch 60/162] avg loss 0.000654768, throughput 9.51187K wps
[Epoch 185 Batch 90/162] avg loss 0.000672343, throughput 9.37431K wps
[Epoch 185 Batch 120/162] avg loss 0.000662391, throughput 9.56333K wps
[Epoch 185 Batch 150/162] avg loss 0.000679335, throughput 9.60563K wps
Begin Testing...
[Epoch 185] train avg loss 0.000670789, dev acc 0.9100, dev avg loss 0.225449, throughput 9.52154K wps
[Epoch 186 Batch 30/162] avg loss 0.000600737, throughput 9.83524K wps
[Epoch 186 Batch 60/162] avg loss 0.000622091, throughput 9.56399K wps
[Epoch 186 Batch 90/162] avg loss 0.000603692, throughput 9.40667K wps
[Epoch 186 Batch 120/162] avg loss 0.000687127, throughput 9.43986K wps
[Epoch 186 Batch 150/162] avg loss 0.000662362, throughput 9.46069K wps
Begin Testing...
[Epoch 186] train avg loss 0.000636065, dev acc 0.9100, dev avg loss 0.225864, throughput 9.50665K wps
[Epoch 187 Batch 30/162] avg loss 0.000671032, throughput 9.56289K wps
[Epoch 187 Batch 60/162] avg loss 0.000664391, throughput 9.2295K wps
[Epoch 187 Batch 90/162] avg loss 0.000629071, throughput 9.51561K wps
[Epoch 187 Batch 120/162] avg loss 0.000630279, throughput 9.36473K wps
[Epoch 187 Batch 150/162] avg loss 0.000600789, throughput 9.36472K wps
Begin Testing...
[Epoch 187] train avg loss 0.000636055, dev acc 0.9133, dev avg loss 0.22612, throughput 9.40336K wps
[Epoch 188 Batch 30/162] avg loss 0.000595856, throughput 9.50273K wps
[Epoch 188 Batch 60/162] avg loss 0.000621212, throughput 9.41431K wps
[Epoch 188 Batch 90/162] avg loss 0.000629791, throughput 9.57499K wps
[Epoch 188 Batch 120/162] avg loss 0.000596043, throughput 9.33723K wps
[Epoch 188 Batch 150/162] avg loss 0.000663129, throughput 9.54882K wps
Begin Testing...
[Epoch 188] train avg loss 0.000629592, dev acc 0.9133, dev avg loss 0.226644, throughput 9.45423K wps
[Epoch 189 Batch 30/162] avg loss 0.000745781, throughput 9.78823K wps
[Epoch 189 Batch 60/162] avg loss 0.000683769, throughput 9.37894K wps
[Epoch 189 Batch 90/162] avg loss 0.000564642, throughput 9.63133K wps
[Epoch 189 Batch 120/162] avg loss 0.000539159, throughput 9.40265K wps
[Epoch 189 Batch 150/162] avg loss 0.000606861, throughput 9.46956K wps
Begin Testing...
[Epoch 189] train avg loss 0.000635219, dev acc 0.9133, dev avg loss 0.225669, throughput 9.53678K wps
[Epoch 190 Batch 30/162] avg loss 0.000703815, throughput 9.67745K wps
[Epoch 190 Batch 60/162] avg loss 0.000617522, throughput 9.50716K wps
[Epoch 190 Batch 90/162] avg loss 0.000610861, throughput 9.38346K wps
[Epoch 190 Batch 120/162] avg loss 0.000618103, throughput 9.47646K wps
[Epoch 190 Batch 150/162] avg loss 0.000679402, throughput 9.38398K wps
Begin Testing...
[Epoch 190] train avg loss 0.000635351, dev acc 0.9122, dev avg loss 0.225466, throughput 9.47271K wps
[Epoch 191 Batch 30/162] avg loss 0.000681128, throughput 9.50809K wps
[Epoch 191 Batch 60/162] avg loss 0.000532713, throughput 9.33118K wps
[Epoch 191 Batch 90/162] avg loss 0.000635811, throughput 9.51921K wps
[Epoch 191 Batch 120/162] avg loss 0.00069502, throughput 9.3735K wps
[Epoch 191 Batch 150/162] avg loss 0.000620193, throughput 9.30973K wps
Begin Testing...
[Epoch 191] train avg loss 0.000633931, dev acc 0.9111, dev avg loss 0.225917, throughput 9.417K wps
[Epoch 192 Batch 30/162] avg loss 0.000706403, throughput 9.60697K wps
[Epoch 192 Batch 60/162] avg loss 0.000623208, throughput 9.2904K wps
[Epoch 192 Batch 90/162] avg loss 0.000584558, throughput 9.45437K wps
[Epoch 192 Batch 120/162] avg loss 0.000588716, throughput 9.46762K wps
[Epoch 192 Batch 150/162] avg loss 0.000582311, throughput 9.45883K wps
Begin Testing...
[Epoch 192] train avg loss 0.000617436, dev acc 0.9111, dev avg loss 0.226298, throughput 9.46323K wps
[Epoch 193 Batch 30/162] avg loss 0.000573028, throughput 9.68002K wps
[Epoch 193 Batch 60/162] avg loss 0.000717601, throughput 9.54518K wps
[Epoch 193 Batch 90/162] avg loss 0.000538138, throughput 9.48574K wps
[Epoch 193 Batch 120/162] avg loss 0.000553734, throughput 9.61656K wps
[Epoch 193 Batch 150/162] avg loss 0.000585577, throughput 9.50974K wps
Begin Testing...
[Epoch 193] train avg loss 0.000606796, dev acc 0.9111, dev avg loss 0.226947, throughput 9.56237K wps
[Epoch 194 Batch 30/162] avg loss 0.000657662, throughput 9.62804K wps
[Epoch 194 Batch 60/162] avg loss 0.000620225, throughput 9.30668K wps
[Epoch 194 Batch 90/162] avg loss 0.000558836, throughput 9.37205K wps
[Epoch 194 Batch 120/162] avg loss 0.000594669, throughput 9.34614K wps
[Epoch 194 Batch 150/162] avg loss 0.00057954, throughput 9.30128K wps
Begin Testing...
[Epoch 194] train avg loss 0.000600465, dev acc 0.9111, dev avg loss 0.226911, throughput 9.40102K wps
[Epoch 195 Batch 30/162] avg loss 0.000592575, throughput 9.6138K wps
[Epoch 195 Batch 60/162] avg loss 0.000551, throughput 9.41282K wps
[Epoch 195 Batch 90/162] avg loss 0.000579981, throughput 9.27996K wps
[Epoch 195 Batch 120/162] avg loss 0.000576392, throughput 9.40536K wps
[Epoch 195 Batch 150/162] avg loss 0.0007301, throughput 9.2917K wps
Begin Testing...
[Epoch 195] train avg loss 0.000609726, dev acc 0.9133, dev avg loss 0.227058, throughput 9.41419K wps
[Epoch 196 Batch 30/162] avg loss 0.000533122, throughput 9.52805K wps
[Epoch 196 Batch 60/162] avg loss 0.000582433, throughput 9.38242K wps
[Epoch 196 Batch 90/162] avg loss 0.00063012, throughput 9.36112K wps
[Epoch 196 Batch 120/162] avg loss 0.000661977, throughput 9.48845K wps
[Epoch 196 Batch 150/162] avg loss 0.000560447, throughput 9.31374K wps
Begin Testing...
[Epoch 196] train avg loss 0.000592269, dev acc 0.9089, dev avg loss 0.226597, throughput 9.42342K wps
[Epoch 197 Batch 30/162] avg loss 0.000578652, throughput 9.5907K wps
[Epoch 197 Batch 60/162] avg loss 0.000628707, throughput 9.47286K wps
[Epoch 197 Batch 90/162] avg loss 0.000538222, throughput 9.36362K wps
[Epoch 197 Batch 120/162] avg loss 0.000592408, throughput 9.40729K wps
[Epoch 197 Batch 150/162] avg loss 0.000606917, throughput 9.38992K wps
Begin Testing...
[Epoch 197] train avg loss 0.000584769, dev acc 0.9111, dev avg loss 0.226831, throughput 9.44147K wps
[Epoch 198 Batch 30/162] avg loss 0.000589252, throughput 9.77139K wps
[Epoch 198 Batch 60/162] avg loss 0.000574108, throughput 9.3738K wps
[Epoch 198 Batch 90/162] avg loss 0.000555945, throughput 9.44005K wps
[Epoch 198 Batch 120/162] avg loss 0.000573489, throughput 9.46199K wps
[Epoch 198 Batch 150/162] avg loss 0.00058973, throughput 9.49571K wps
Begin Testing...
[Epoch 198] train avg loss 0.000575245, dev acc 0.9144, dev avg loss 0.226975, throughput 9.49996K wps
[Epoch 199 Batch 30/162] avg loss 0.000542852, throughput 9.77405K wps
[Epoch 199 Batch 60/162] avg loss 0.000555296, throughput 9.40481K wps
[Epoch 199 Batch 90/162] avg loss 0.000512179, throughput 9.3764K wps
[Epoch 199 Batch 120/162] avg loss 0.000641214, throughput 9.44818K wps
[Epoch 199 Batch 150/162] avg loss 0.000649096, throughput 9.44669K wps
Begin Testing...
[Epoch 199] train avg loss 0.000579453, dev acc 0.9122, dev avg loss 0.22638, throughput 9.47982K wps
Test loss 0.198094, test acc 0.9260
Total time cost 443.03s
[Epoch 0 Batch 30/162] avg loss 0.014044, throughput 7.42173K wps
[Epoch 0 Batch 60/162] avg loss 0.0139113, throughput 9.39125K wps
[Epoch 0 Batch 90/162] avg loss 0.0135511, throughput 9.37137K wps
[Epoch 0 Batch 120/162] avg loss 0.0135927, throughput 9.60509K wps
[Epoch 0 Batch 150/162] avg loss 0.0134128, throughput 9.29364K wps
Begin Testing...
[Epoch 0] train avg loss 0.0136815, dev acc 0.7644, dev avg loss 0.658921, throughput 8.95846K wps
Observed Improvement.
Begin Testing...
[Epoch 1 Batch 30/162] avg loss 0.0131006, throughput 9.2941K wps
[Epoch 1 Batch 60/162] avg loss 0.0131621, throughput 9.33647K wps
[Epoch 1 Batch 90/162] avg loss 0.0130057, throughput 9.42683K wps
[Epoch 1 Batch 120/162] avg loss 0.0127587, throughput 9.58746K wps
[Epoch 1 Batch 150/162] avg loss 0.0126817, throughput 9.44937K wps
Begin Testing...
[Epoch 1] train avg loss 0.0129114, dev acc 0.7956, dev avg loss 0.624788, throughput 9.40696K wps
Observed Improvement.
Begin Testing...
[Epoch 2 Batch 30/162] avg loss 0.0124832, throughput 9.80776K wps
[Epoch 2 Batch 60/162] avg loss 0.0123547, throughput 9.27063K wps
[Epoch 2 Batch 90/162] avg loss 0.0122773, throughput 9.34652K wps
[Epoch 2 Batch 120/162] avg loss 0.0121444, throughput 9.32669K wps
[Epoch 2 Batch 150/162] avg loss 0.0121025, throughput 9.54865K wps
Begin Testing...
[Epoch 2] train avg loss 0.0122499, dev acc 0.8444, dev avg loss 0.587188, throughput 9.44545K wps
Observed Improvement.
Begin Testing...
[Epoch 3 Batch 30/162] avg loss 0.0118746, throughput 9.70652K wps
[Epoch 3 Batch 60/162] avg loss 0.0116458, throughput 9.29088K wps
[Epoch 3 Batch 90/162] avg loss 0.0114658, throughput 9.56816K wps
[Epoch 3 Batch 120/162] avg loss 0.0112933, throughput 9.61836K wps
[Epoch 3 Batch 150/162] avg loss 0.0109971, throughput 9.29478K wps
Begin Testing...
[Epoch 3] train avg loss 0.0114139, dev acc 0.8444, dev avg loss 0.546351, throughput 9.47371K wps
Observed Improvement.
Begin Testing...
[Epoch 4 Batch 30/162] avg loss 0.0110867, throughput 9.56729K wps
[Epoch 4 Batch 60/162] avg loss 0.0107767, throughput 9.59673K wps
[Epoch 4 Batch 90/162] avg loss 0.0106545, throughput 9.43145K wps
[Epoch 4 Batch 120/162] avg loss 0.0104114, throughput 9.58966K wps
[Epoch 4 Batch 150/162] avg loss 0.0101404, throughput 9.30357K wps
Begin Testing...
[Epoch 4] train avg loss 0.0105927, dev acc 0.8556, dev avg loss 0.505141, throughput 9.46548K wps
Observed Improvement.
Begin Testing...
[Epoch 5 Batch 30/162] avg loss 0.00996177, throughput 9.71171K wps
[Epoch 5 Batch 60/162] avg loss 0.00982345, throughput 9.40168K wps
[Epoch 5 Batch 90/162] avg loss 0.00974186, throughput 9.34161K wps
[Epoch 5 Batch 120/162] avg loss 0.00998971, throughput 9.59276K wps
[Epoch 5 Batch 150/162] avg loss 0.00965488, throughput 9.31741K wps
Begin Testing...
[Epoch 5] train avg loss 0.0098063, dev acc 0.8600, dev avg loss 0.466279, throughput 9.47128K wps
Observed Improvement.
Begin Testing...
[Epoch 6 Batch 30/162] avg loss 0.00946479, throughput 9.53986K wps
[Epoch 6 Batch 60/162] avg loss 0.00938496, throughput 9.62445K wps
[Epoch 6 Batch 90/162] avg loss 0.00895814, throughput 9.33454K wps
[Epoch 6 Batch 120/162] avg loss 0.00898372, throughput 9.46644K wps
[Epoch 6 Batch 150/162] avg loss 0.00890799, throughput 9.36622K wps
Begin Testing...
[Epoch 6] train avg loss 0.00909111, dev acc 0.8644, dev avg loss 0.432737, throughput 9.4404K wps
Observed Improvement.
Begin Testing...
[Epoch 7 Batch 30/162] avg loss 0.00864902, throughput 9.53342K wps
[Epoch 7 Batch 60/162] avg loss 0.00860795, throughput 9.48564K wps
[Epoch 7 Batch 90/162] avg loss 0.00841739, throughput 9.45939K wps
[Epoch 7 Batch 120/162] avg loss 0.00820578, throughput 9.49153K wps
[Epoch 7 Batch 150/162] avg loss 0.00827831, throughput 9.35243K wps
Begin Testing...
[Epoch 7] train avg loss 0.0084143, dev acc 0.8656, dev avg loss 0.406146, throughput 9.44754K wps
Observed Improvement.
Begin Testing...
[Epoch 8 Batch 30/162] avg loss 0.00806792, throughput 9.5274K wps
[Epoch 8 Batch 60/162] avg loss 0.0078111, throughput 9.33049K wps
[Epoch 8 Batch 90/162] avg loss 0.00803015, throughput 9.32473K wps
[Epoch 8 Batch 120/162] avg loss 0.00779646, throughput 9.55925K wps
[Epoch 8 Batch 150/162] avg loss 0.00797453, throughput 9.54674K wps
Begin Testing...
[Epoch 8] train avg loss 0.00792195, dev acc 0.8689, dev avg loss 0.384192, throughput 9.43503K wps
Observed Improvement.
Begin Testing...
[Epoch 9 Batch 30/162] avg loss 0.00753628, throughput 9.71614K wps
[Epoch 9 Batch 60/162] avg loss 0.00745364, throughput 9.35141K wps
[Epoch 9 Batch 90/162] avg loss 0.00767971, throughput 9.56005K wps
[Epoch 9 Batch 120/162] avg loss 0.00736652, throughput 9.29627K wps
[Epoch 9 Batch 150/162] avg loss 0.00735072, throughput 9.3172K wps
Begin Testing...
[Epoch 9] train avg loss 0.00747043, dev acc 0.8667, dev avg loss 0.368975, throughput 9.43008K wps
[Epoch 10 Batch 30/162] avg loss 0.00716013, throughput 9.55095K wps
[Epoch 10 Batch 60/162] avg loss 0.0069575, throughput 9.27829K wps
[Epoch 10 Batch 90/162] avg loss 0.00697828, throughput 9.32152K wps
[Epoch 10 Batch 120/162] avg loss 0.00709591, throughput 9.34114K wps
[Epoch 10 Batch 150/162] avg loss 0.00728423, throughput 9.321K wps
Begin Testing...
[Epoch 10] train avg loss 0.00707409, dev acc 0.8689, dev avg loss 0.353487, throughput 9.36246K wps
Observed Improvement.
Begin Testing...
[Epoch 11 Batch 30/162] avg loss 0.00689479, throughput 9.49412K wps
[Epoch 11 Batch 60/162] avg loss 0.00690117, throughput 9.34448K wps
[Epoch 11 Batch 90/162] avg loss 0.00677828, throughput 9.32374K wps
[Epoch 11 Batch 120/162] avg loss 0.00665143, throughput 9.36285K wps
[Epoch 11 Batch 150/162] avg loss 0.00667784, throughput 9.58142K wps
Begin Testing...
[Epoch 11] train avg loss 0.00674239, dev acc 0.8700, dev avg loss 0.342794, throughput 9.40806K wps
Observed Improvement.
Begin Testing...
[Epoch 12 Batch 30/162] avg loss 0.00673683, throughput 9.61002K wps
[Epoch 12 Batch 60/162] avg loss 0.00674519, throughput 9.45597K wps
[Epoch 12 Batch 90/162] avg loss 0.00655507, throughput 9.30878K wps
[Epoch 12 Batch 120/162] avg loss 0.00661799, throughput 9.43638K wps
[Epoch 12 Batch 150/162] avg loss 0.00626557, throughput 9.4861K wps
Begin Testing...
[Epoch 12] train avg loss 0.00658661, dev acc 0.8700, dev avg loss 0.334428, throughput 9.46157K wps
Observed Improvement.
Begin Testing...
[Epoch 13 Batch 30/162] avg loss 0.00627414, throughput 9.48448K wps
[Epoch 13 Batch 60/162] avg loss 0.00667653, throughput 9.34239K wps
[Epoch 13 Batch 90/162] avg loss 0.00597704, throughput 9.60255K wps
[Epoch 13 Batch 120/162] avg loss 0.00635429, throughput 9.3137K wps
[Epoch 13 Batch 150/162] avg loss 0.00658428, throughput 9.36902K wps
Begin Testing...
[Epoch 13] train avg loss 0.00636667, dev acc 0.8689, dev avg loss 0.328312, throughput 9.43341K wps
[Epoch 14 Batch 30/162] avg loss 0.00630411, throughput 9.6987K wps
[Epoch 14 Batch 60/162] avg loss 0.00601151, throughput 9.5844K wps
[Epoch 14 Batch 90/162] avg loss 0.00628223, throughput 9.36125K wps
[Epoch 14 Batch 120/162] avg loss 0.00631605, throughput 9.44144K wps
[Epoch 14 Batch 150/162] avg loss 0.00599739, throughput 9.19567K wps
Begin Testing...
[Epoch 14] train avg loss 0.00615073, dev acc 0.8722, dev avg loss 0.321209, throughput 9.44454K wps
Observed Improvement.
Begin Testing...
[Epoch 15 Batch 30/162] avg loss 0.00593829, throughput 9.72278K wps
[Epoch 15 Batch 60/162] avg loss 0.00604746, throughput 9.56231K wps
[Epoch 15 Batch 90/162] avg loss 0.00619689, throughput 9.32084K wps
[Epoch 15 Batch 120/162] avg loss 0.00588855, throughput 9.57151K wps
[Epoch 15 Batch 150/162] avg loss 0.00582991, throughput 9.57045K wps
Begin Testing...
[Epoch 15] train avg loss 0.00598946, dev acc 0.8644, dev avg loss 0.318891, throughput 9.52353K wps
[Epoch 16 Batch 30/162] avg loss 0.00585411, throughput 9.40916K wps
[Epoch 16 Batch 60/162] avg loss 0.00605269, throughput 9.38825K wps
[Epoch 16 Batch 90/162] avg loss 0.00566332, throughput 9.4335K wps
[Epoch 16 Batch 120/162] avg loss 0.00594855, throughput 9.53239K wps
[Epoch 16 Batch 150/162] avg loss 0.00581551, throughput 9.28893K wps
Begin Testing...
[Epoch 16] train avg loss 0.00587102, dev acc 0.8722, dev avg loss 0.31129, throughput 9.40664K wps
Observed Improvement.
Begin Testing...
[Epoch 17 Batch 30/162] avg loss 0.00568156, throughput 9.52845K wps
[Epoch 17 Batch 60/162] avg loss 0.00570679, throughput 9.31733K wps
[Epoch 17 Batch 90/162] avg loss 0.00600059, throughput 9.34508K wps
[Epoch 17 Batch 120/162] avg loss 0.00567625, throughput 9.43859K wps
[Epoch 17 Batch 150/162] avg loss 0.00592487, throughput 9.58175K wps
Begin Testing...
[Epoch 17] train avg loss 0.00579657, dev acc 0.8722, dev avg loss 0.306844, throughput 9.43238K wps
Observed Improvement.
Begin Testing...
[Epoch 18 Batch 30/162] avg loss 0.00541588, throughput 9.67642K wps
[Epoch 18 Batch 60/162] avg loss 0.00564289, throughput 9.23792K wps
[Epoch 18 Batch 90/162] avg loss 0.00548197, throughput 9.3887K wps
[Epoch 18 Batch 120/162] avg loss 0.00558148, throughput 9.53388K wps
[Epoch 18 Batch 150/162] avg loss 0.00619392, throughput 9.42127K wps
Begin Testing...
[Epoch 18] train avg loss 0.0056656, dev acc 0.8722, dev avg loss 0.303698, throughput 9.44889K wps
Observed Improvement.
Begin Testing...
[Epoch 19 Batch 30/162] avg loss 0.00578652, throughput 9.64017K wps
[Epoch 19 Batch 60/162] avg loss 0.00532289, throughput 9.32835K wps
[Epoch 19 Batch 90/162] avg loss 0.00538903, throughput 9.4586K wps
[Epoch 19 Batch 120/162] avg loss 0.00559997, throughput 9.48188K wps
[Epoch 19 Batch 150/162] avg loss 0.00521889, throughput 9.28014K wps
Begin Testing...
[Epoch 19] train avg loss 0.00547204, dev acc 0.8778, dev avg loss 0.299203, throughput 9.42792K wps
Observed Improvement.
Begin Testing...
[Epoch 20 Batch 30/162] avg loss 0.00518693, throughput 9.54515K wps
[Epoch 20 Batch 60/162] avg loss 0.00538405, throughput 9.57019K wps
[Epoch 20 Batch 90/162] avg loss 0.00542486, throughput 9.42623K wps
[Epoch 20 Batch 120/162] avg loss 0.00547941, throughput 9.39161K wps
[Epoch 20 Batch 150/162] avg loss 0.00513912, throughput 9.48984K wps
Begin Testing...
[Epoch 20] train avg loss 0.00531639, dev acc 0.8756, dev avg loss 0.295829, throughput 9.47373K wps
[Epoch 21 Batch 30/162] avg loss 0.00533976, throughput 9.51534K wps
[Epoch 21 Batch 60/162] avg loss 0.00518879, throughput 9.30512K wps
[Epoch 21 Batch 90/162] avg loss 0.00541034, throughput 9.37287K wps
[Epoch 21 Batch 120/162] avg loss 0.00519912, throughput 9.48289K wps
[Epoch 21 Batch 150/162] avg loss 0.005221, throughput 9.56777K wps
Begin Testing...
[Epoch 21] train avg loss 0.00525349, dev acc 0.8789, dev avg loss 0.292809, throughput 9.43815K wps
Observed Improvement.
Begin Testing...
[Epoch 22 Batch 30/162] avg loss 0.00495187, throughput 9.6585K wps
[Epoch 22 Batch 60/162] avg loss 0.00498665, throughput 9.31166K wps
[Epoch 22 Batch 90/162] avg loss 0.00524085, throughput 9.46411K wps
[Epoch 22 Batch 120/162] avg loss 0.00541192, throughput 9.37048K wps
[Epoch 22 Batch 150/162] avg loss 0.00523464, throughput 9.20999K wps
Begin Testing...
[Epoch 22] train avg loss 0.00517254, dev acc 0.8778, dev avg loss 0.289964, throughput 9.395K wps
[Epoch 23 Batch 30/162] avg loss 0.00530496, throughput 9.7608K wps
[Epoch 23 Batch 60/162] avg loss 0.00494584, throughput 9.49525K wps
[Epoch 23 Batch 90/162] avg loss 0.00514542, throughput 9.3285K wps
[Epoch 23 Batch 120/162] avg loss 0.00495335, throughput 9.43402K wps
[Epoch 23 Batch 150/162] avg loss 0.00511127, throughput 9.42936K wps
Begin Testing...
[Epoch 23] train avg loss 0.00506485, dev acc 0.8744, dev avg loss 0.287445, throughput 9.48925K wps
[Epoch 24 Batch 30/162] avg loss 0.00469069, throughput 9.50609K wps
[Epoch 24 Batch 60/162] avg loss 0.00469443, throughput 9.27294K wps
[Epoch 24 Batch 90/162] avg loss 0.00504144, throughput 9.38166K wps
[Epoch 24 Batch 120/162] avg loss 0.00516163, throughput 9.38854K wps
[Epoch 24 Batch 150/162] avg loss 0.00483835, throughput 9.24231K wps
Begin Testing...
[Epoch 24] train avg loss 0.00489254, dev acc 0.8756, dev avg loss 0.286172, throughput 9.33765K wps
[Epoch 25 Batch 30/162] avg loss 0.00482251, throughput 9.54411K wps
[Epoch 25 Batch 60/162] avg loss 0.00478625, throughput 9.31315K wps
[Epoch 25 Batch 90/162] avg loss 0.00500493, throughput 9.41485K wps
[Epoch 25 Batch 120/162] avg loss 0.00508516, throughput 9.32673K wps
[Epoch 25 Batch 150/162] avg loss 0.0049546, throughput 9.34375K wps
Begin Testing...
[Epoch 25] train avg loss 0.00489381, dev acc 0.8789, dev avg loss 0.282281, throughput 9.41074K wps
Observed Improvement.
Begin Testing...
[Epoch 26 Batch 30/162] avg loss 0.00487656, throughput 9.76205K wps
[Epoch 26 Batch 60/162] avg loss 0.00470308, throughput 9.52115K wps
[Epoch 26 Batch 90/162] avg loss 0.0047298, throughput 9.56274K wps
[Epoch 26 Batch 120/162] avg loss 0.00451848, throughput 9.36395K wps
[Epoch 26 Batch 150/162] avg loss 0.0047719, throughput 9.62747K wps
Begin Testing...
[Epoch 26] train avg loss 0.00481447, dev acc 0.8844, dev avg loss 0.279582, throughput 9.56437K wps
Observed Improvement.
Begin Testing...
[Epoch 27 Batch 30/162] avg loss 0.00475928, throughput 9.68286K wps
[Epoch 27 Batch 60/162] avg loss 0.00472005, throughput 9.4664K wps
[Epoch 27 Batch 90/162] avg loss 0.00436842, throughput 9.54699K wps
[Epoch 27 Batch 120/162] avg loss 0.00479422, throughput 9.64384K wps
[Epoch 27 Batch 150/162] avg loss 0.00471519, throughput 9.37709K wps
Begin Testing...
[Epoch 27] train avg loss 0.00468426, dev acc 0.8789, dev avg loss 0.277895, throughput 9.54308K wps
[Epoch 28 Batch 30/162] avg loss 0.00466451, throughput 9.82038K wps
[Epoch 28 Batch 60/162] avg loss 0.00473682, throughput 9.43763K wps
[Epoch 28 Batch 90/162] avg loss 0.00477128, throughput 9.52861K wps
[Epoch 28 Batch 120/162] avg loss 0.00452268, throughput 9.46094K wps
[Epoch 28 Batch 150/162] avg loss 0.00483537, throughput 9.36514K wps
Begin Testing...
[Epoch 28] train avg loss 0.0046631, dev acc 0.8844, dev avg loss 0.275386, throughput 9.51959K wps
Observed Improvement.
Begin Testing...
[Epoch 29 Batch 30/162] avg loss 0.00410758, throughput 9.80285K wps
[Epoch 29 Batch 60/162] avg loss 0.00491941, throughput 9.33629K wps
[Epoch 29 Batch 90/162] avg loss 0.00455026, throughput 9.44047K wps
[Epoch 29 Batch 120/162] avg loss 0.00466227, throughput 9.45426K wps
[Epoch 29 Batch 150/162] avg loss 0.0047395, throughput 9.38067K wps
Begin Testing...
[Epoch 29] train avg loss 0.00459367, dev acc 0.8856, dev avg loss 0.273469, throughput 9.49249K wps
Observed Improvement.
Begin Testing...
[Epoch 30 Batch 30/162] avg loss 0.00445268, throughput 9.58011K wps
[Epoch 30 Batch 60/162] avg loss 0.00459751, throughput 9.4028K wps
[Epoch 30 Batch 90/162] avg loss 0.00421334, throughput 9.35794K wps
[Epoch 30 Batch 120/162] avg loss 0.00473479, throughput 9.32953K wps
[Epoch 30 Batch 150/162] avg loss 0.00417864, throughput 9.46082K wps
Begin Testing...
[Epoch 30] train avg loss 0.00443051, dev acc 0.8844, dev avg loss 0.27113, throughput 9.40357K wps
[Epoch 31 Batch 30/162] avg loss 0.00430618, throughput 9.55784K wps
[Epoch 31 Batch 60/162] avg loss 0.00420817, throughput 9.34793K wps
[Epoch 31 Batch 90/162] avg loss 0.00437745, throughput 9.46005K wps
[Epoch 31 Batch 120/162] avg loss 0.00431446, throughput 9.51231K wps
[Epoch 31 Batch 150/162] avg loss 0.00460008, throughput 9.40804K wps
Begin Testing...
[Epoch 31] train avg loss 0.00437635, dev acc 0.8856, dev avg loss 0.270159, throughput 9.43524K wps
Observed Improvement.
Begin Testing...
[Epoch 32 Batch 30/162] avg loss 0.00466118, throughput 9.67906K wps
[Epoch 32 Batch 60/162] avg loss 0.00428256, throughput 9.63458K wps
[Epoch 32 Batch 90/162] avg loss 0.00412493, throughput 9.31777K wps
[Epoch 32 Batch 120/162] avg loss 0.00426027, throughput 9.44553K wps
[Epoch 32 Batch 150/162] avg loss 0.00435632, throughput 9.35815K wps
Begin Testing...
[Epoch 32] train avg loss 0.00436572, dev acc 0.8867, dev avg loss 0.268102, throughput 9.46575K wps
Observed Improvement.
Begin Testing...
[Epoch 33 Batch 30/162] avg loss 0.004302, throughput 9.71293K wps
[Epoch 33 Batch 60/162] avg loss 0.00423254, throughput 9.57029K wps
[Epoch 33 Batch 90/162] avg loss 0.00418933, throughput 9.55964K wps
[Epoch 33 Batch 120/162] avg loss 0.00447605, throughput 9.5837K wps
[Epoch 33 Batch 150/162] avg loss 0.00456588, throughput 9.62051K wps
Begin Testing...
[Epoch 33] train avg loss 0.0043146, dev acc 0.8867, dev avg loss 0.266358, throughput 9.60695K wps
Observed Improvement.
Begin Testing...
[Epoch 34 Batch 30/162] avg loss 0.00437804, throughput 9.77593K wps
[Epoch 34 Batch 60/162] avg loss 0.00408379, throughput 9.25072K wps
[Epoch 34 Batch 90/162] avg loss 0.00431129, throughput 9.53552K wps
[Epoch 34 Batch 120/162] avg loss 0.00420907, throughput 9.33623K wps
[Epoch 34 Batch 150/162] avg loss 0.00405316, throughput 9.29163K wps
Begin Testing...
[Epoch 34] train avg loss 0.00415747, dev acc 0.8822, dev avg loss 0.266951, throughput 9.42688K wps
[Epoch 35 Batch 30/162] avg loss 0.00418194, throughput 9.63068K wps
[Epoch 35 Batch 60/162] avg loss 0.00410126, throughput 9.41397K wps
[Epoch 35 Batch 90/162] avg loss 0.00424382, throughput 9.3564K wps
[Epoch 35 Batch 120/162] avg loss 0.00403649, throughput 9.34821K wps
[Epoch 35 Batch 150/162] avg loss 0.00383561, throughput 9.43299K wps
Begin Testing...
[Epoch 35] train avg loss 0.00408063, dev acc 0.8867, dev avg loss 0.263723, throughput 9.41368K wps
Observed Improvement.
Begin Testing...
[Epoch 36 Batch 30/162] avg loss 0.0043861, throughput 9.69958K wps
[Epoch 36 Batch 60/162] avg loss 0.00386609, throughput 9.36463K wps
[Epoch 36 Batch 90/162] avg loss 0.00399411, throughput 9.43358K wps
[Epoch 36 Batch 120/162] avg loss 0.00392703, throughput 9.33678K wps
[Epoch 36 Batch 150/162] avg loss 0.00397728, throughput 9.51557K wps
Begin Testing...
[Epoch 36] train avg loss 0.00401802, dev acc 0.8856, dev avg loss 0.262225, throughput 9.45598K wps
[Epoch 37 Batch 30/162] avg loss 0.00405193, throughput 9.67642K wps
[Epoch 37 Batch 60/162] avg loss 0.00380926, throughput 9.56377K wps
[Epoch 37 Batch 90/162] avg loss 0.00384673, throughput 9.26923K wps
[Epoch 37 Batch 120/162] avg loss 0.00430077, throughput 9.42371K wps
[Epoch 37 Batch 150/162] avg loss 0.00415954, throughput 9.17056K wps
Begin Testing...
[Epoch 37] train avg loss 0.00399777, dev acc 0.8867, dev avg loss 0.26119, throughput 9.40794K wps
Observed Improvement.
Begin Testing...
[Epoch 38 Batch 30/162] avg loss 0.00400751, throughput 9.5346K wps
[Epoch 38 Batch 60/162] avg loss 0.003865, throughput 9.45007K wps
[Epoch 38 Batch 90/162] avg loss 0.00379978, throughput 9.3164K wps
[Epoch 38 Batch 120/162] avg loss 0.00390737, throughput 9.22388K wps
[Epoch 38 Batch 150/162] avg loss 0.00370998, throughput 9.42708K wps
Begin Testing...
[Epoch 38] train avg loss 0.00391169, dev acc 0.8856, dev avg loss 0.259909, throughput 9.41612K wps
[Epoch 39 Batch 30/162] avg loss 0.00391539, throughput 9.69715K wps
[Epoch 39 Batch 60/162] avg loss 0.00391641, throughput 9.57708K wps
[Epoch 39 Batch 90/162] avg loss 0.00382859, throughput 9.30836K wps
[Epoch 39 Batch 120/162] avg loss 0.00410479, throughput 9.39584K wps
[Epoch 39 Batch 150/162] avg loss 0.00388993, throughput 9.51438K wps
Begin Testing...
[Epoch 39] train avg loss 0.00393487, dev acc 0.8911, dev avg loss 0.258491, throughput 9.50041K wps
Observed Improvement.
Begin Testing...
[Epoch 40 Batch 30/162] avg loss 0.00359504, throughput 9.49774K wps
[Epoch 40 Batch 60/162] avg loss 0.00382095, throughput 9.54895K wps
[Epoch 40 Batch 90/162] avg loss 0.00395517, throughput 9.33982K wps
[Epoch 40 Batch 120/162] avg loss 0.00368019, throughput 9.37272K wps
[Epoch 40 Batch 150/162] avg loss 0.00377357, throughput 9.42685K wps
Begin Testing...
[Epoch 40] train avg loss 0.00377437, dev acc 0.8911, dev avg loss 0.25757, throughput 9.43651K wps
Observed Improvement.
Begin Testing...
[Epoch 41 Batch 30/162] avg loss 0.00389117, throughput 9.45299K wps
[Epoch 41 Batch 60/162] avg loss 0.00370161, throughput 9.42835K wps
[Epoch 41 Batch 90/162] avg loss 0.00365722, throughput 9.58707K wps
[Epoch 41 Batch 120/162] avg loss 0.00367804, throughput 9.41999K wps
[Epoch 41 Batch 150/162] avg loss 0.00396983, throughput 9.47142K wps
Begin Testing...
[Epoch 41] train avg loss 0.00378921, dev acc 0.8889, dev avg loss 0.258552, throughput 9.44517K wps
[Epoch 42 Batch 30/162] avg loss 0.00368975, throughput 9.42254K wps
[Epoch 42 Batch 60/162] avg loss 0.0036328, throughput 9.59433K wps
[Epoch 42 Batch 90/162] avg loss 0.0033615, throughput 9.61896K wps
[Epoch 42 Batch 120/162] avg loss 0.00402167, throughput 9.42232K wps
[Epoch 42 Batch 150/162] avg loss 0.00374175, throughput 9.31585K wps
Begin Testing...
[Epoch 42] train avg loss 0.00369251, dev acc 0.8867, dev avg loss 0.256211, throughput 9.45829K wps
[Epoch 43 Batch 30/162] avg loss 0.00355696, throughput 9.63634K wps
[Epoch 43 Batch 60/162] avg loss 0.00372876, throughput 9.44585K wps
[Epoch 43 Batch 90/162] avg loss 0.00422217, throughput 9.42368K wps
[Epoch 43 Batch 120/162] avg loss 0.00354134, throughput 9.33358K wps
[Epoch 43 Batch 150/162] avg loss 0.00330055, throughput 9.33071K wps
Begin Testing...
[Epoch 43] train avg loss 0.00365514, dev acc 0.8878, dev avg loss 0.254117, throughput 9.42607K wps
[Epoch 44 Batch 30/162] avg loss 0.00379317, throughput 9.59559K wps
[Epoch 44 Batch 60/162] avg loss 0.00378241, throughput 9.30982K wps
[Epoch 44 Batch 90/162] avg loss 0.00340175, throughput 9.40072K wps
[Epoch 44 Batch 120/162] avg loss 0.00354626, throughput 9.40817K wps
[Epoch 44 Batch 150/162] avg loss 0.00355995, throughput 9.56989K wps
Begin Testing...
[Epoch 44] train avg loss 0.00360405, dev acc 0.8900, dev avg loss 0.256393, throughput 9.4536K wps
[Epoch 45 Batch 30/162] avg loss 0.00335727, throughput 9.66868K wps
[Epoch 45 Batch 60/162] avg loss 0.00367332, throughput 9.24424K wps
[Epoch 45 Batch 90/162] avg loss 0.00362746, throughput 9.31981K wps
[Epoch 45 Batch 120/162] avg loss 0.00363027, throughput 9.46334K wps
[Epoch 45 Batch 150/162] avg loss 0.00335932, throughput 9.365K wps
Begin Testing...
[Epoch 45] train avg loss 0.00355172, dev acc 0.8900, dev avg loss 0.252198, throughput 9.42261K wps
[Epoch 46 Batch 30/162] avg loss 0.0035815, throughput 9.65131K wps
[Epoch 46 Batch 60/162] avg loss 0.00369696, throughput 9.34544K wps
[Epoch 46 Batch 90/162] avg loss 0.00320625, throughput 9.41102K wps
[Epoch 46 Batch 120/162] avg loss 0.00332812, throughput 9.42378K wps
[Epoch 46 Batch 150/162] avg loss 0.00338439, throughput 9.60584K wps
Begin Testing...
[Epoch 46] train avg loss 0.00346786, dev acc 0.8900, dev avg loss 0.251157, throughput 9.46865K wps
[Epoch 47 Batch 30/162] avg loss 0.00374016, throughput 9.81544K wps
[Epoch 47 Batch 60/162] avg loss 0.00338939, throughput 9.3376K wps
[Epoch 47 Batch 90/162] avg loss 0.00327524, throughput 9.3707K wps
[Epoch 47 Batch 120/162] avg loss 0.00368323, throughput 9.50633K wps
[Epoch 47 Batch 150/162] avg loss 0.00346617, throughput 9.3803K wps
Begin Testing...
[Epoch 47] train avg loss 0.00347953, dev acc 0.8922, dev avg loss 0.250573, throughput 9.46918K wps
Observed Improvement.
Begin Testing...
[Epoch 48 Batch 30/162] avg loss 0.00349475, throughput 9.5249K wps
[Epoch 48 Batch 60/162] avg loss 0.00362644, throughput 9.28234K wps
[Epoch 48 Batch 90/162] avg loss 0.00327666, throughput 9.52406K wps
[Epoch 48 Batch 120/162] avg loss 0.00321477, throughput 9.43887K wps
[Epoch 48 Batch 150/162] avg loss 0.00323204, throughput 9.3127K wps
Begin Testing...
[Epoch 48] train avg loss 0.0033848, dev acc 0.8900, dev avg loss 0.250694, throughput 9.40947K wps
[Epoch 49 Batch 30/162] avg loss 0.00344899, throughput 9.51234K wps
[Epoch 49 Batch 60/162] avg loss 0.00342226, throughput 9.39777K wps
[Epoch 49 Batch 90/162] avg loss 0.00356793, throughput 9.41311K wps
[Epoch 49 Batch 120/162] avg loss 0.00304314, throughput 9.37829K wps
[Epoch 49 Batch 150/162] avg loss 0.00313328, throughput 9.36329K wps
Begin Testing...
[Epoch 49] train avg loss 0.00331213, dev acc 0.8922, dev avg loss 0.24964, throughput 9.40325K wps
Observed Improvement.
Begin Testing...
[Epoch 50 Batch 30/162] avg loss 0.00311823, throughput 9.52684K wps
[Epoch 50 Batch 60/162] avg loss 0.00326753, throughput 9.32669K wps
[Epoch 50 Batch 90/162] avg loss 0.00345923, throughput 9.37942K wps
[Epoch 50 Batch 120/162] avg loss 0.00321281, throughput 9.32304K wps
[Epoch 50 Batch 150/162] avg loss 0.00355152, throughput 9.30889K wps
Begin Testing...
[Epoch 50] train avg loss 0.00330859, dev acc 0.8911, dev avg loss 0.248168, throughput 9.37791K wps
[Epoch 51 Batch 30/162] avg loss 0.00335859, throughput 9.57068K wps
[Epoch 51 Batch 60/162] avg loss 0.00326685, throughput 9.26373K wps
[Epoch 51 Batch 90/162] avg loss 0.00356555, throughput 9.46946K wps
[Epoch 51 Batch 120/162] avg loss 0.00309679, throughput 9.46453K wps
[Epoch 51 Batch 150/162] avg loss 0.00308382, throughput 9.4839K wps
Begin Testing...
[Epoch 51] train avg loss 0.00327257, dev acc 0.8889, dev avg loss 0.247431, throughput 9.43208K wps
[Epoch 52 Batch 30/162] avg loss 0.00336259, throughput 9.43547K wps
[Epoch 52 Batch 60/162] avg loss 0.00313576, throughput 9.30797K wps
[Epoch 52 Batch 90/162] avg loss 0.00318245, throughput 9.63425K wps
[Epoch 52 Batch 120/162] avg loss 0.00314882, throughput 9.2902K wps
[Epoch 52 Batch 150/162] avg loss 0.00338573, throughput 9.43571K wps
Begin Testing...
[Epoch 52] train avg loss 0.00322081, dev acc 0.8911, dev avg loss 0.246589, throughput 9.42527K wps
[Epoch 53 Batch 30/162] avg loss 0.00347903, throughput 9.69407K wps
[Epoch 53 Batch 60/162] avg loss 0.00312003, throughput 9.40587K wps
[Epoch 53 Batch 90/162] avg loss 0.00300966, throughput 9.40713K wps
[Epoch 53 Batch 120/162] avg loss 0.00305324, throughput 9.3628K wps
[Epoch 53 Batch 150/162] avg loss 0.00315616, throughput 9.57979K wps
Begin Testing...
[Epoch 53] train avg loss 0.00314161, dev acc 0.8900, dev avg loss 0.245968, throughput 9.49355K wps
[Epoch 54 Batch 30/162] avg loss 0.00284865, throughput 9.80364K wps
[Epoch 54 Batch 60/162] avg loss 0.00296437, throughput 9.45151K wps
[Epoch 54 Batch 90/162] avg loss 0.00353542, throughput 9.5735K wps
[Epoch 54 Batch 120/162] avg loss 0.00286092, throughput 9.58929K wps
[Epoch 54 Batch 150/162] avg loss 0.00317509, throughput 9.58012K wps
Begin Testing...
[Epoch 54] train avg loss 0.00305549, dev acc 0.8967, dev avg loss 0.245909, throughput 9.57642K wps
Observed Improvement.
Begin Testing...
[Epoch 55 Batch 30/162] avg loss 0.00278485, throughput 9.57212K wps
[Epoch 55 Batch 60/162] avg loss 0.0032272, throughput 9.2302K wps
[Epoch 55 Batch 90/162] avg loss 0.00324245, throughput 9.58163K wps
[Epoch 55 Batch 120/162] avg loss 0.00310262, throughput 9.62205K wps
[Epoch 55 Batch 150/162] avg loss 0.00292671, throughput 9.40323K wps
Begin Testing...
[Epoch 55] train avg loss 0.00307542, dev acc 0.8933, dev avg loss 0.245202, throughput 9.49349K wps
[Epoch 56 Batch 30/162] avg loss 0.00310052, throughput 9.71014K wps
[Epoch 56 Batch 60/162] avg loss 0.00285735, throughput 9.25508K wps
[Epoch 56 Batch 90/162] avg loss 0.0029351, throughput 9.34949K wps
[Epoch 56 Batch 120/162] avg loss 0.0029559, throughput 9.34167K wps
[Epoch 56 Batch 150/162] avg loss 0.00294718, throughput 9.49445K wps
Begin Testing...
[Epoch 56] train avg loss 0.00297822, dev acc 0.8933, dev avg loss 0.244033, throughput 9.42037K wps
[Epoch 57 Batch 30/162] avg loss 0.00307673, throughput 9.56037K wps
[Epoch 57 Batch 60/162] avg loss 0.00293636, throughput 9.40345K wps
[Epoch 57 Batch 90/162] avg loss 0.0027105, throughput 9.53369K wps
[Epoch 57 Batch 120/162] avg loss 0.00309495, throughput 9.14552K wps
[Epoch 57 Batch 150/162] avg loss 0.00294276, throughput 9.29265K wps
Begin Testing...
[Epoch 57] train avg loss 0.0029384, dev acc 0.8922, dev avg loss 0.243878, throughput 9.38426K wps
[Epoch 58 Batch 30/162] avg loss 0.00288777, throughput 9.57725K wps
[Epoch 58 Batch 60/162] avg loss 0.00298633, throughput 9.49346K wps
[Epoch 58 Batch 90/162] avg loss 0.00289822, throughput 9.28864K wps
[Epoch 58 Batch 120/162] avg loss 0.0033056, throughput 9.58549K wps
[Epoch 58 Batch 150/162] avg loss 0.00284951, throughput 9.28609K wps
Begin Testing...
[Epoch 58] train avg loss 0.00298032, dev acc 0.8911, dev avg loss 0.242596, throughput 9.44667K wps
[Epoch 59 Batch 30/162] avg loss 0.00263403, throughput 9.64945K wps
[Epoch 59 Batch 60/162] avg loss 0.00261324, throughput 9.55186K wps
[Epoch 59 Batch 90/162] avg loss 0.00287096, throughput 9.55235K wps
[Epoch 59 Batch 120/162] avg loss 0.00283413, throughput 9.48769K wps
[Epoch 59 Batch 150/162] avg loss 0.00291539, throughput 9.55205K wps
Begin Testing...
[Epoch 59] train avg loss 0.00280505, dev acc 0.8911, dev avg loss 0.242412, throughput 9.53733K wps
[Epoch 60 Batch 30/162] avg loss 0.00266162, throughput 9.50214K wps
[Epoch 60 Batch 60/162] avg loss 0.00298779, throughput 9.42528K wps
[Epoch 60 Batch 90/162] avg loss 0.00261823, throughput 9.45231K wps
[Epoch 60 Batch 120/162] avg loss 0.00292727, throughput 9.26817K wps
[Epoch 60 Batch 150/162] avg loss 0.00303968, throughput 9.55073K wps
Begin Testing...
[Epoch 60] train avg loss 0.00284395, dev acc 0.8922, dev avg loss 0.242742, throughput 9.44439K wps
[Epoch 61 Batch 30/162] avg loss 0.00296553, throughput 9.60712K wps
[Epoch 61 Batch 60/162] avg loss 0.00284937, throughput 9.33362K wps
[Epoch 61 Batch 90/162] avg loss 0.00274432, throughput 9.29038K wps
[Epoch 61 Batch 120/162] avg loss 0.00277312, throughput 9.50726K wps
[Epoch 61 Batch 150/162] avg loss 0.00257539, throughput 9.30509K wps
Begin Testing...
[Epoch 61] train avg loss 0.00280122, dev acc 0.8933, dev avg loss 0.241166, throughput 9.41444K wps
[Epoch 62 Batch 30/162] avg loss 0.00268774, throughput 9.63816K wps
[Epoch 62 Batch 60/162] avg loss 0.00267288, throughput 9.57096K wps
[Epoch 62 Batch 90/162] avg loss 0.0029004, throughput 9.44975K wps
[Epoch 62 Batch 120/162] avg loss 0.00280886, throughput 9.30267K wps
[Epoch 62 Batch 150/162] avg loss 0.00288593, throughput 9.31096K wps
Begin Testing...
[Epoch 62] train avg loss 0.00279368, dev acc 0.8956, dev avg loss 0.240419, throughput 9.43975K wps
[Epoch 63 Batch 30/162] avg loss 0.0027409, throughput 9.68967K wps
[Epoch 63 Batch 60/162] avg loss 0.0026557, throughput 9.29339K wps
[Epoch 63 Batch 90/162] avg loss 0.00274204, throughput 9.27821K wps
[Epoch 63 Batch 120/162] avg loss 0.00272929, throughput 9.3222K wps
[Epoch 63 Batch 150/162] avg loss 0.00277854, throughput 9.43655K wps
Begin Testing...
[Epoch 63] train avg loss 0.00273605, dev acc 0.8900, dev avg loss 0.242675, throughput 9.40261K wps
[Epoch 64 Batch 30/162] avg loss 0.0029107, throughput 9.58045K wps
[Epoch 64 Batch 60/162] avg loss 0.00239503, throughput 9.57582K wps
[Epoch 64 Batch 90/162] avg loss 0.00273153, throughput 9.54843K wps
[Epoch 64 Batch 120/162] avg loss 0.00264102, throughput 9.22509K wps
[Epoch 64 Batch 150/162] avg loss 0.00288633, throughput 9.39246K wps
Begin Testing...
[Epoch 64] train avg loss 0.00270589, dev acc 0.8922, dev avg loss 0.240346, throughput 9.452K wps
[Epoch 65 Batch 30/162] avg loss 0.00256487, throughput 9.57727K wps
[Epoch 65 Batch 60/162] avg loss 0.00265273, throughput 9.41609K wps
[Epoch 65 Batch 90/162] avg loss 0.00261227, throughput 9.50738K wps
[Epoch 65 Batch 120/162] avg loss 0.00275373, throughput 9.41399K wps
[Epoch 65 Batch 150/162] avg loss 0.00283105, throughput 9.46512K wps
Begin Testing...
[Epoch 65] train avg loss 0.00266069, dev acc 0.8956, dev avg loss 0.238772, throughput 9.47908K wps
[Epoch 66 Batch 30/162] avg loss 0.00257276, throughput 9.4911K wps
[Epoch 66 Batch 60/162] avg loss 0.00261596, throughput 9.52248K wps
[Epoch 66 Batch 90/162] avg loss 0.00264558, throughput 9.30811K wps
[Epoch 66 Batch 120/162] avg loss 0.00267441, throughput 9.47871K wps
[Epoch 66 Batch 150/162] avg loss 0.00256187, throughput 9.49969K wps
Begin Testing...
[Epoch 66] train avg loss 0.00260294, dev acc 0.8933, dev avg loss 0.239123, throughput 9.46664K wps
[Epoch 67 Batch 30/162] avg loss 0.00252531, throughput 9.49183K wps
[Epoch 67 Batch 60/162] avg loss 0.00273705, throughput 9.53064K wps
[Epoch 67 Batch 90/162] avg loss 0.00264121, throughput 9.43002K wps
[Epoch 67 Batch 120/162] avg loss 0.0025853, throughput 9.41602K wps
[Epoch 67 Batch 150/162] avg loss 0.00268825, throughput 9.45228K wps
Begin Testing...
[Epoch 67] train avg loss 0.0026075, dev acc 0.8967, dev avg loss 0.238198, throughput 9.46974K wps
Observed Improvement.
Begin Testing...
[Epoch 68 Batch 30/162] avg loss 0.00274637, throughput 9.80898K wps
[Epoch 68 Batch 60/162] avg loss 0.00243224, throughput 9.51634K wps
[Epoch 68 Batch 90/162] avg loss 0.00245602, throughput 9.38608K wps
[Epoch 68 Batch 120/162] avg loss 0.00265919, throughput 9.3773K wps
[Epoch 68 Batch 150/162] avg loss 0.00258693, throughput 9.37211K wps
Begin Testing...
[Epoch 68] train avg loss 0.00256318, dev acc 0.8911, dev avg loss 0.239043, throughput 9.4858K wps
[Epoch 69 Batch 30/162] avg loss 0.00212567, throughput 9.63847K wps
[Epoch 69 Batch 60/162] avg loss 0.00256415, throughput 9.51503K wps
[Epoch 69 Batch 90/162] avg loss 0.00251513, throughput 9.29159K wps
[Epoch 69 Batch 120/162] avg loss 0.00281181, throughput 9.3739K wps
[Epoch 69 Batch 150/162] avg loss 0.00243542, throughput 9.33214K wps
Begin Testing...
[Epoch 69] train avg loss 0.00250262, dev acc 0.8900, dev avg loss 0.238562, throughput 9.44172K wps
[Epoch 70 Batch 30/162] avg loss 0.00266931, throughput 9.62432K wps
[Epoch 70 Batch 60/162] avg loss 0.00260568, throughput 9.4409K wps
[Epoch 70 Batch 90/162] avg loss 0.00247521, throughput 9.54233K wps
[Epoch 70 Batch 120/162] avg loss 0.00274011, throughput 9.42976K wps
[Epoch 70 Batch 150/162] avg loss 0.00236558, throughput 9.58084K wps
Begin Testing...
[Epoch 70] train avg loss 0.00254315, dev acc 0.8967, dev avg loss 0.236535, throughput 9.51022K wps
Observed Improvement.
Begin Testing...
[Epoch 71 Batch 30/162] avg loss 0.00262354, throughput 9.59845K wps
[Epoch 71 Batch 60/162] avg loss 0.00245066, throughput 9.31755K wps
[Epoch 71 Batch 90/162] avg loss 0.00245554, throughput 9.64106K wps
[Epoch 71 Batch 120/162] avg loss 0.00244514, throughput 9.62366K wps
[Epoch 71 Batch 150/162] avg loss 0.00244192, throughput 9.58315K wps
Begin Testing...
[Epoch 71] train avg loss 0.00246346, dev acc 0.8967, dev avg loss 0.235927, throughput 9.53166K wps
Observed Improvement.
Begin Testing...
[Epoch 72 Batch 30/162] avg loss 0.00233429, throughput 9.47144K wps
[Epoch 72 Batch 60/162] avg loss 0.0025653, throughput 9.33496K wps
[Epoch 72 Batch 90/162] avg loss 0.00224535, throughput 9.30548K wps
[Epoch 72 Batch 120/162] avg loss 0.00240337, throughput 8.47566K wps
[Epoch 72 Batch 150/162] avg loss 0.00261886, throughput 9.58394K wps
Begin Testing...
[Epoch 72] train avg loss 0.00242096, dev acc 0.8889, dev avg loss 0.237147, throughput 9.22763K wps
[Epoch 73 Batch 30/162] avg loss 0.00239907, throughput 9.47191K wps
[Epoch 73 Batch 60/162] avg loss 0.00266038, throughput 9.36537K wps
[Epoch 73 Batch 90/162] avg loss 0.00218214, throughput 9.41795K wps
[Epoch 73 Batch 120/162] avg loss 0.0023022, throughput 9.30014K wps
[Epoch 73 Batch 150/162] avg loss 0.002166, throughput 9.53699K wps
Begin Testing...
[Epoch 73] train avg loss 0.00234894, dev acc 0.8978, dev avg loss 0.235217, throughput 9.39633K wps
Observed Improvement.
Begin Testing...
[Epoch 74 Batch 30/162] avg loss 0.00234546, throughput 9.39383K wps
[Epoch 74 Batch 60/162] avg loss 0.00238256, throughput 9.31181K wps
[Epoch 74 Batch 90/162] avg loss 0.00244491, throughput 9.31486K wps
[Epoch 74 Batch 120/162] avg loss 0.00217877, throughput 9.41073K wps
[Epoch 74 Batch 150/162] avg loss 0.00247799, throughput 9.2327K wps
Begin Testing...
[Epoch 74] train avg loss 0.00235234, dev acc 0.8956, dev avg loss 0.235349, throughput 9.33221K wps
[Epoch 75 Batch 30/162] avg loss 0.00219784, throughput 9.60646K wps
[Epoch 75 Batch 60/162] avg loss 0.0023521, throughput 9.34782K wps
[Epoch 75 Batch 90/162] avg loss 0.00247738, throughput 9.60435K wps
[Epoch 75 Batch 120/162] avg loss 0.00237879, throughput 9.42713K wps
[Epoch 75 Batch 150/162] avg loss 0.00243676, throughput 9.39055K wps
Begin Testing...
[Epoch 75] train avg loss 0.00237213, dev acc 0.8967, dev avg loss 0.234056, throughput 9.46101K wps
[Epoch 76 Batch 30/162] avg loss 0.00212232, throughput 9.61293K wps
[Epoch 76 Batch 60/162] avg loss 0.00233502, throughput 9.26979K wps
[Epoch 76 Batch 90/162] avg loss 0.00238647, throughput 9.46898K wps
[Epoch 76 Batch 120/162] avg loss 0.00213213, throughput 9.36334K wps
[Epoch 76 Batch 150/162] avg loss 0.00227248, throughput 9.38745K wps
Begin Testing...
[Epoch 76] train avg loss 0.00225029, dev acc 0.8989, dev avg loss 0.233727, throughput 9.41176K wps
Observed Improvement.
Begin Testing...
[Epoch 77 Batch 30/162] avg loss 0.00252519, throughput 9.48769K wps
[Epoch 77 Batch 60/162] avg loss 0.00240595, throughput 9.44432K wps
[Epoch 77 Batch 90/162] avg loss 0.00212033, throughput 9.4699K wps
[Epoch 77 Batch 120/162] avg loss 0.00222168, throughput 9.44982K wps
[Epoch 77 Batch 150/162] avg loss 0.00222852, throughput 9.28281K wps
Begin Testing...
[Epoch 77] train avg loss 0.00229487, dev acc 0.8989, dev avg loss 0.234221, throughput 9.40599K wps
Observed Improvement.
Begin Testing...
[Epoch 78 Batch 30/162] avg loss 0.00215713, throughput 9.57939K wps
[Epoch 78 Batch 60/162] avg loss 0.00238776, throughput 9.3332K wps
[Epoch 78 Batch 90/162] avg loss 0.00210219, throughput 9.33845K wps
[Epoch 78 Batch 120/162] avg loss 0.0022106, throughput 9.40562K wps
[Epoch 78 Batch 150/162] avg loss 0.00228915, throughput 9.55604K wps
Begin Testing...
[Epoch 78] train avg loss 0.00222765, dev acc 0.8978, dev avg loss 0.23389, throughput 9.42317K wps
[Epoch 79 Batch 30/162] avg loss 0.00229194, throughput 9.61324K wps
[Epoch 79 Batch 60/162] avg loss 0.00229443, throughput 9.31734K wps
[Epoch 79 Batch 90/162] avg loss 0.00220223, throughput 9.40574K wps
[Epoch 79 Batch 120/162] avg loss 0.00214529, throughput 9.25112K wps
[Epoch 79 Batch 150/162] avg loss 0.00208421, throughput 9.32203K wps
Begin Testing...
[Epoch 79] train avg loss 0.00218858, dev acc 0.8956, dev avg loss 0.233796, throughput 9.39618K wps
[Epoch 80 Batch 30/162] avg loss 0.00239574, throughput 9.63437K wps
[Epoch 80 Batch 60/162] avg loss 0.00199913, throughput 9.54536K wps
[Epoch 80 Batch 90/162] avg loss 0.00212439, throughput 9.42756K wps
[Epoch 80 Batch 120/162] avg loss 0.00225371, throughput 9.24205K wps
[Epoch 80 Batch 150/162] avg loss 0.00209646, throughput 9.29234K wps
Begin Testing...
[Epoch 80] train avg loss 0.00218313, dev acc 0.9000, dev avg loss 0.232771, throughput 9.43321K wps
Observed Improvement.
Begin Testing...
[Epoch 81 Batch 30/162] avg loss 0.00217784, throughput 9.54414K wps
[Epoch 81 Batch 60/162] avg loss 0.00206072, throughput 9.38167K wps
[Epoch 81 Batch 90/162] avg loss 0.00206004, throughput 9.53227K wps
[Epoch 81 Batch 120/162] avg loss 0.00205685, throughput 9.47595K wps
[Epoch 81 Batch 150/162] avg loss 0.00216127, throughput 9.2631K wps
Begin Testing...
[Epoch 81] train avg loss 0.00210981, dev acc 0.9011, dev avg loss 0.233498, throughput 9.43069K wps
Observed Improvement.
Begin Testing...
[Epoch 82 Batch 30/162] avg loss 0.00218932, throughput 9.57463K wps
[Epoch 82 Batch 60/162] avg loss 0.00220242, throughput 9.36176K wps
[Epoch 82 Batch 90/162] avg loss 0.00230341, throughput 9.51688K wps
[Epoch 82 Batch 120/162] avg loss 0.00214886, throughput 9.33343K wps
[Epoch 82 Batch 150/162] avg loss 0.0019563, throughput 9.4901K wps
Begin Testing...
[Epoch 82] train avg loss 0.00214788, dev acc 0.9000, dev avg loss 0.232568, throughput 9.45714K wps
[Epoch 83 Batch 30/162] avg loss 0.00218645, throughput 9.67593K wps
[Epoch 83 Batch 60/162] avg loss 0.00196928, throughput 9.51696K wps
[Epoch 83 Batch 90/162] avg loss 0.00194357, throughput 9.48172K wps
[Epoch 83 Batch 120/162] avg loss 0.00193722, throughput 9.45338K wps
[Epoch 83 Batch 150/162] avg loss 0.0022867, throughput 9.2621K wps
Begin Testing...
[Epoch 83] train avg loss 0.00208427, dev acc 0.9000, dev avg loss 0.23259, throughput 9.46649K wps
[Epoch 84 Batch 30/162] avg loss 0.00204165, throughput 9.48024K wps
[Epoch 84 Batch 60/162] avg loss 0.00207077, throughput 9.37478K wps
[Epoch 84 Batch 90/162] avg loss 0.0020035, throughput 9.51225K wps
[Epoch 84 Batch 120/162] avg loss 0.00188873, throughput 9.30255K wps
[Epoch 84 Batch 150/162] avg loss 0.00211149, throughput 9.36731K wps
Begin Testing...
[Epoch 84] train avg loss 0.00203938, dev acc 0.8967, dev avg loss 0.235494, throughput 9.40292K wps
[Epoch 85 Batch 30/162] avg loss 0.00182035, throughput 9.52859K wps
[Epoch 85 Batch 60/162] avg loss 0.00189461, throughput 9.36701K wps
[Epoch 85 Batch 90/162] avg loss 0.00204067, throughput 9.40316K wps
[Epoch 85 Batch 120/162] avg loss 0.00196696, throughput 9.56991K wps
[Epoch 85 Batch 150/162] avg loss 0.00197897, throughput 9.56518K wps
Begin Testing...
[Epoch 85] train avg loss 0.00194903, dev acc 0.8956, dev avg loss 0.234468, throughput 9.46731K wps
[Epoch 86 Batch 30/162] avg loss 0.00207339, throughput 9.47224K wps
[Epoch 86 Batch 60/162] avg loss 0.00204412, throughput 9.36418K wps
[Epoch 86 Batch 90/162] avg loss 0.0019791, throughput 9.42293K wps
[Epoch 86 Batch 120/162] avg loss 0.00198485, throughput 9.51775K wps
[Epoch 86 Batch 150/162] avg loss 0.00192776, throughput 9.59324K wps
Begin Testing...
[Epoch 86] train avg loss 0.001977, dev acc 0.8989, dev avg loss 0.231761, throughput 9.47567K wps
[Epoch 87 Batch 30/162] avg loss 0.00187238, throughput 9.76664K wps
[Epoch 87 Batch 60/162] avg loss 0.00190297, throughput 9.34194K wps
[Epoch 87 Batch 90/162] avg loss 0.00209677, throughput 9.50187K wps
[Epoch 87 Batch 120/162] avg loss 0.00192234, throughput 9.39384K wps
[Epoch 87 Batch 150/162] avg loss 0.00192222, throughput 9.3327K wps
Begin Testing...
[Epoch 87] train avg loss 0.00195394, dev acc 0.8989, dev avg loss 0.231547, throughput 9.45558K wps
[Epoch 88 Batch 30/162] avg loss 0.00217883, throughput 9.61858K wps
[Epoch 88 Batch 60/162] avg loss 0.00187452, throughput 9.39801K wps
[Epoch 88 Batch 90/162] avg loss 0.0018866, throughput 9.39553K wps
[Epoch 88 Batch 120/162] avg loss 0.00196614, throughput 9.28423K wps
[Epoch 88 Batch 150/162] avg loss 0.0019496, throughput 9.49323K wps
Begin Testing...
[Epoch 88] train avg loss 0.00196344, dev acc 0.9000, dev avg loss 0.231374, throughput 9.45285K wps
[Epoch 89 Batch 30/162] avg loss 0.00216425, throughput 9.63936K wps
[Epoch 89 Batch 60/162] avg loss 0.00190606, throughput 9.62812K wps
[Epoch 89 Batch 90/162] avg loss 0.00193641, throughput 9.43479K wps
[Epoch 89 Batch 120/162] avg loss 0.00177316, throughput 9.45497K wps
[Epoch 89 Batch 150/162] avg loss 0.0018594, throughput 9.31072K wps
Begin Testing...
[Epoch 89] train avg loss 0.00191093, dev acc 0.8989, dev avg loss 0.231078, throughput 9.49929K wps
[Epoch 90 Batch 30/162] avg loss 0.00195051, throughput 9.58267K wps
[Epoch 90 Batch 60/162] avg loss 0.00183411, throughput 9.3747K wps
[Epoch 90 Batch 90/162] avg loss 0.00191986, throughput 9.43845K wps
[Epoch 90 Batch 120/162] avg loss 0.00184949, throughput 9.60506K wps
[Epoch 90 Batch 150/162] avg loss 0.00174372, throughput 9.52422K wps
Begin Testing...
[Epoch 90] train avg loss 0.00185369, dev acc 0.9011, dev avg loss 0.232182, throughput 9.51356K wps
Observed Improvement.
Begin Testing...
[Epoch 91 Batch 30/162] avg loss 0.00169101, throughput 9.58509K wps
[Epoch 91 Batch 60/162] avg loss 0.00207246, throughput 9.2417K wps
[Epoch 91 Batch 90/162] avg loss 0.0018803, throughput 9.527K wps
[Epoch 91 Batch 120/162] avg loss 0.00207241, throughput 9.44761K wps
[Epoch 91 Batch 150/162] avg loss 0.00166296, throughput 9.44278K wps
Begin Testing...
[Epoch 91] train avg loss 0.00187821, dev acc 0.8989, dev avg loss 0.230533, throughput 9.43136K wps
[Epoch 92 Batch 30/162] avg loss 0.00192242, throughput 9.62969K wps
[Epoch 92 Batch 60/162] avg loss 0.001862, throughput 9.46559K wps
[Epoch 92 Batch 90/162] avg loss 0.00183308, throughput 9.60567K wps
[Epoch 92 Batch 120/162] avg loss 0.00203162, throughput 9.40288K wps
[Epoch 92 Batch 150/162] avg loss 0.00182059, throughput 9.39712K wps
Begin Testing...
[Epoch 92] train avg loss 0.00187145, dev acc 0.9000, dev avg loss 0.231033, throughput 9.50585K wps
[Epoch 93 Batch 30/162] avg loss 0.00173168, throughput 9.49854K wps
[Epoch 93 Batch 60/162] avg loss 0.002027, throughput 9.44338K wps
[Epoch 93 Batch 90/162] avg loss 0.00172144, throughput 9.26043K wps
[Epoch 93 Batch 120/162] avg loss 0.00187929, throughput 9.49932K wps
[Epoch 93 Batch 150/162] avg loss 0.00183148, throughput 9.40126K wps
Begin Testing...
[Epoch 93] train avg loss 0.00183018, dev acc 0.9000, dev avg loss 0.230326, throughput 9.43197K wps
[Epoch 94 Batch 30/162] avg loss 0.00172673, throughput 9.55472K wps
[Epoch 94 Batch 60/162] avg loss 0.00179149, throughput 9.48091K wps
[Epoch 94 Batch 90/162] avg loss 0.00184594, throughput 9.6197K wps
[Epoch 94 Batch 120/162] avg loss 0.00183689, throughput 9.25583K wps
[Epoch 94 Batch 150/162] avg loss 0.00180329, throughput 9.35194K wps
Begin Testing...
[Epoch 94] train avg loss 0.00181377, dev acc 0.8978, dev avg loss 0.230335, throughput 9.44622K wps
[Epoch 95 Batch 30/162] avg loss 0.00167832, throughput 9.6151K wps
[Epoch 95 Batch 60/162] avg loss 0.00189907, throughput 9.3832K wps
[Epoch 95 Batch 90/162] avg loss 0.00175009, throughput 9.52389K wps
[Epoch 95 Batch 120/162] avg loss 0.00169701, throughput 9.30671K wps
[Epoch 95 Batch 150/162] avg loss 0.00173792, throughput 9.36436K wps
Begin Testing...
[Epoch 95] train avg loss 0.00175745, dev acc 0.8989, dev avg loss 0.230295, throughput 9.43728K wps
[Epoch 96 Batch 30/162] avg loss 0.00177386, throughput 9.59088K wps
[Epoch 96 Batch 60/162] avg loss 0.00169207, throughput 9.37038K wps
[Epoch 96 Batch 90/162] avg loss 0.00175597, throughput 9.40461K wps
[Epoch 96 Batch 120/162] avg loss 0.00182263, throughput 9.52753K wps
[Epoch 96 Batch 150/162] avg loss 0.00182384, throughput 9.46752K wps
Begin Testing...
[Epoch 96] train avg loss 0.00178255, dev acc 0.8989, dev avg loss 0.229911, throughput 9.47701K wps
[Epoch 97 Batch 30/162] avg loss 0.00168648, throughput 9.59696K wps
[Epoch 97 Batch 60/162] avg loss 0.00176477, throughput 9.40069K wps
[Epoch 97 Batch 90/162] avg loss 0.00163539, throughput 9.3484K wps
[Epoch 97 Batch 120/162] avg loss 0.00143552, throughput 9.35651K wps
[Epoch 97 Batch 150/162] avg loss 0.0018318, throughput 9.50826K wps
Begin Testing...
[Epoch 97] train avg loss 0.00168356, dev acc 0.9022, dev avg loss 0.230964, throughput 9.4237K wps
Observed Improvement.
Begin Testing...
[Epoch 98 Batch 30/162] avg loss 0.00164796, throughput 9.66225K wps
[Epoch 98 Batch 60/162] avg loss 0.00171134, throughput 9.36997K wps
[Epoch 98 Batch 90/162] avg loss 0.00185597, throughput 9.62256K wps
[Epoch 98 Batch 120/162] avg loss 0.00176851, throughput 9.50883K wps
[Epoch 98 Batch 150/162] avg loss 0.00161523, throughput 9.32444K wps
Begin Testing...
[Epoch 98] train avg loss 0.00171488, dev acc 0.9022, dev avg loss 0.229782, throughput 9.49635K wps
Observed Improvement.
Begin Testing...
[Epoch 99 Batch 30/162] avg loss 0.00158175, throughput 9.62459K wps
[Epoch 99 Batch 60/162] avg loss 0.00155318, throughput 9.62118K wps
[Epoch 99 Batch 90/162] avg loss 0.00164221, throughput 9.43861K wps
[Epoch 99 Batch 120/162] avg loss 0.00155142, throughput 9.38274K wps
[Epoch 99 Batch 150/162] avg loss 0.00192595, throughput 9.48765K wps
Begin Testing...
[Epoch 99] train avg loss 0.00165123, dev acc 0.9022, dev avg loss 0.231718, throughput 9.47538K wps
Observed Improvement.
Begin Testing...
[Epoch 100 Batch 30/162] avg loss 0.00156471, throughput 9.68544K wps
[Epoch 100 Batch 60/162] avg loss 0.00169474, throughput 9.17916K wps
[Epoch 100 Batch 90/162] avg loss 0.0016847, throughput 9.42355K wps
[Epoch 100 Batch 120/162] avg loss 0.00178601, throughput 9.50791K wps
[Epoch 100 Batch 150/162] avg loss 0.00148046, throughput 9.29912K wps
Begin Testing...
[Epoch 100] train avg loss 0.00163424, dev acc 0.8978, dev avg loss 0.229707, throughput 9.42158K wps
[Epoch 101 Batch 30/162] avg loss 0.00166289, throughput 9.54044K wps
[Epoch 101 Batch 60/162] avg loss 0.00153689, throughput 9.39446K wps
[Epoch 101 Batch 90/162] avg loss 0.00157604, throughput 9.47864K wps
[Epoch 101 Batch 120/162] avg loss 0.00158887, throughput 9.4789K wps
[Epoch 101 Batch 150/162] avg loss 0.0015243, throughput 9.29068K wps
Begin Testing...
[Epoch 101] train avg loss 0.00157747, dev acc 0.9011, dev avg loss 0.229835, throughput 9.43894K wps
[Epoch 102 Batch 30/162] avg loss 0.00158943, throughput 9.56256K wps
[Epoch 102 Batch 60/162] avg loss 0.00146202, throughput 9.30371K wps
[Epoch 102 Batch 90/162] avg loss 0.00162326, throughput 9.41726K wps
[Epoch 102 Batch 120/162] avg loss 0.00163306, throughput 9.44411K wps
[Epoch 102 Batch 150/162] avg loss 0.00160103, throughput 9.56018K wps
Begin Testing...
[Epoch 102] train avg loss 0.0015772, dev acc 0.9000, dev avg loss 0.229626, throughput 9.46953K wps
[Epoch 103 Batch 30/162] avg loss 0.00155718, throughput 9.58428K wps
[Epoch 103 Batch 60/162] avg loss 0.00158105, throughput 9.35728K wps
[Epoch 103 Batch 90/162] avg loss 0.00148708, throughput 9.50147K wps
[Epoch 103 Batch 120/162] avg loss 0.00164184, throughput 9.25251K wps
[Epoch 103 Batch 150/162] avg loss 0.0016279, throughput 9.45684K wps
Begin Testing...
[Epoch 103] train avg loss 0.00157103, dev acc 0.9022, dev avg loss 0.229635, throughput 9.44856K wps
Observed Improvement.
Begin Testing...
[Epoch 104 Batch 30/162] avg loss 0.00154338, throughput 9.60337K wps
[Epoch 104 Batch 60/162] avg loss 0.00171871, throughput 9.41416K wps
[Epoch 104 Batch 90/162] avg loss 0.00143081, throughput 9.45156K wps
[Epoch 104 Batch 120/162] avg loss 0.00165203, throughput 9.32518K wps
[Epoch 104 Batch 150/162] avg loss 0.00152502, throughput 9.46844K wps
Begin Testing...
[Epoch 104] train avg loss 0.00156831, dev acc 0.9011, dev avg loss 0.229316, throughput 9.43615K wps
[Epoch 105 Batch 30/162] avg loss 0.00164707, throughput 9.34131K wps
[Epoch 105 Batch 60/162] avg loss 0.0016618, throughput 9.38364K wps
[Epoch 105 Batch 90/162] avg loss 0.00157718, throughput 9.52079K wps
[Epoch 105 Batch 120/162] avg loss 0.00149669, throughput 9.57522K wps
[Epoch 105 Batch 150/162] avg loss 0.00156261, throughput 9.51499K wps
Begin Testing...
[Epoch 105] train avg loss 0.00157545, dev acc 0.9011, dev avg loss 0.229393, throughput 9.45711K wps
[Epoch 106 Batch 30/162] avg loss 0.00152589, throughput 9.6395K wps
[Epoch 106 Batch 60/162] avg loss 0.00157519, throughput 9.40476K wps
[Epoch 106 Batch 90/162] avg loss 0.00139346, throughput 9.47849K wps
[Epoch 106 Batch 120/162] avg loss 0.00153026, throughput 9.33397K wps
[Epoch 106 Batch 150/162] avg loss 0.0017678, throughput 9.41224K wps
Begin Testing...
[Epoch 106] train avg loss 0.00155466, dev acc 0.9044, dev avg loss 0.228847, throughput 9.46071K wps
Observed Improvement.
Begin Testing...
[Epoch 107 Batch 30/162] avg loss 0.00153281, throughput 9.65459K wps
[Epoch 107 Batch 60/162] avg loss 0.00156307, throughput 9.5602K wps
[Epoch 107 Batch 90/162] avg loss 0.00150626, throughput 9.43648K wps
[Epoch 107 Batch 120/162] avg loss 0.0014957, throughput 9.52578K wps
[Epoch 107 Batch 150/162] avg loss 0.00150797, throughput 9.36892K wps
Begin Testing...
[Epoch 107] train avg loss 0.00151264, dev acc 0.9022, dev avg loss 0.22932, throughput 9.4704K wps
[Epoch 108 Batch 30/162] avg loss 0.00136826, throughput 9.63025K wps
[Epoch 108 Batch 60/162] avg loss 0.00158914, throughput 9.58123K wps
[Epoch 108 Batch 90/162] avg loss 0.00146122, throughput 9.2606K wps
[Epoch 108 Batch 120/162] avg loss 0.00157174, throughput 9.31088K wps
[Epoch 108 Batch 150/162] avg loss 0.00148317, throughput 9.51739K wps
Begin Testing...
[Epoch 108] train avg loss 0.00150865, dev acc 0.9000, dev avg loss 0.229187, throughput 9.44038K wps
[Epoch 109 Batch 30/162] avg loss 0.00148304, throughput 9.63174K wps
[Epoch 109 Batch 60/162] avg loss 0.00155601, throughput 9.61838K wps
[Epoch 109 Batch 90/162] avg loss 0.00151325, throughput 9.44688K wps
[Epoch 109 Batch 120/162] avg loss 0.00148817, throughput 9.40675K wps
[Epoch 109 Batch 150/162] avg loss 0.00160745, throughput 9.43799K wps
Begin Testing...
[Epoch 109] train avg loss 0.00152165, dev acc 0.9022, dev avg loss 0.228667, throughput 9.50047K wps
[Epoch 110 Batch 30/162] avg loss 0.0014725, throughput 9.57295K wps
[Epoch 110 Batch 60/162] avg loss 0.00150469, throughput 9.31726K wps
[Epoch 110 Batch 90/162] avg loss 0.00152596, throughput 9.34777K wps
[Epoch 110 Batch 120/162] avg loss 0.00137345, throughput 9.33396K wps
[Epoch 110 Batch 150/162] avg loss 0.00145334, throughput 9.21671K wps
Begin Testing...
[Epoch 110] train avg loss 0.0014674, dev acc 0.9033, dev avg loss 0.229478, throughput 9.34314K wps
[Epoch 111 Batch 30/162] avg loss 0.00142294, throughput 9.4969K wps
[Epoch 111 Batch 60/162] avg loss 0.00159521, throughput 9.47351K wps
[Epoch 111 Batch 90/162] avg loss 0.00168375, throughput 9.27858K wps
[Epoch 111 Batch 120/162] avg loss 0.00141751, throughput 9.35581K wps
[Epoch 111 Batch 150/162] avg loss 0.00130692, throughput 9.40991K wps
Begin Testing...
[Epoch 111] train avg loss 0.0014743, dev acc 0.9011, dev avg loss 0.228895, throughput 9.39375K wps
[Epoch 112 Batch 30/162] avg loss 0.00141152, throughput 9.51457K wps
[Epoch 112 Batch 60/162] avg loss 0.00140815, throughput 9.42316K wps
[Epoch 112 Batch 90/162] avg loss 0.00138166, throughput 9.46438K wps
[Epoch 112 Batch 120/162] avg loss 0.00144033, throughput 9.31244K wps
[Epoch 112 Batch 150/162] avg loss 0.00130052, throughput 9.43454K wps
Begin Testing...
[Epoch 112] train avg loss 0.00141306, dev acc 0.9000, dev avg loss 0.229016, throughput 9.42333K wps
[Epoch 113 Batch 30/162] avg loss 0.00136892, throughput 9.69759K wps
[Epoch 113 Batch 60/162] avg loss 0.00131096, throughput 9.43081K wps
[Epoch 113 Batch 90/162] avg loss 0.00139842, throughput 9.33384K wps
[Epoch 113 Batch 120/162] avg loss 0.00133373, throughput 9.4686K wps
[Epoch 113 Batch 150/162] avg loss 0.00127458, throughput 9.55425K wps
Begin Testing...
[Epoch 113] train avg loss 0.00133152, dev acc 0.9056, dev avg loss 0.228359, throughput 9.47869K wps
Observed Improvement.
Begin Testing...
[Epoch 114 Batch 30/162] avg loss 0.00138034, throughput 9.52433K wps
[Epoch 114 Batch 60/162] avg loss 0.0013205, throughput 9.51767K wps
[Epoch 114 Batch 90/162] avg loss 0.00143324, throughput 9.47674K wps
[Epoch 114 Batch 120/162] avg loss 0.00134947, throughput 9.53763K wps
[Epoch 114 Batch 150/162] avg loss 0.00136266, throughput 9.55874K wps
Begin Testing...
[Epoch 114] train avg loss 0.00137175, dev acc 0.9011, dev avg loss 0.228457, throughput 9.53494K wps
[Epoch 115 Batch 30/162] avg loss 0.0014088, throughput 9.4661K wps
[Epoch 115 Batch 60/162] avg loss 0.00133879, throughput 9.53234K wps
[Epoch 115 Batch 90/162] avg loss 0.00152828, throughput 9.50602K wps
[Epoch 115 Batch 120/162] avg loss 0.00149211, throughput 9.31118K wps
[Epoch 115 Batch 150/162] avg loss 0.00139059, throughput 9.33086K wps
Begin Testing...
[Epoch 115] train avg loss 0.0014235, dev acc 0.9033, dev avg loss 0.228247, throughput 9.41929K wps
[Epoch 116 Batch 30/162] avg loss 0.00131769, throughput 9.63791K wps
[Epoch 116 Batch 60/162] avg loss 0.0012596, throughput 9.11888K wps
[Epoch 116 Batch 90/162] avg loss 0.00133689, throughput 9.38587K wps
[Epoch 116 Batch 120/162] avg loss 0.00129099, throughput 9.29684K wps
[Epoch 116 Batch 150/162] avg loss 0.00138055, throughput 9.44778K wps
Begin Testing...
[Epoch 116] train avg loss 0.00132767, dev acc 0.9022, dev avg loss 0.228274, throughput 9.3958K wps
[Epoch 117 Batch 30/162] avg loss 0.00138043, throughput 9.48456K wps
[Epoch 117 Batch 60/162] avg loss 0.0013876, throughput 9.45629K wps
[Epoch 117 Batch 90/162] avg loss 0.00113684, throughput 9.40546K wps
[Epoch 117 Batch 120/162] avg loss 0.0012274, throughput 9.49604K wps
[Epoch 117 Batch 150/162] avg loss 0.00130974, throughput 9.52704K wps
Begin Testing...
[Epoch 117] train avg loss 0.00129223, dev acc 0.9033, dev avg loss 0.22859, throughput 9.47736K wps
[Epoch 118 Batch 30/162] avg loss 0.00123726, throughput 9.47951K wps
[Epoch 118 Batch 60/162] avg loss 0.00123882, throughput 9.31905K wps
[Epoch 118 Batch 90/162] avg loss 0.00133795, throughput 9.45299K wps
[Epoch 118 Batch 120/162] avg loss 0.00122764, throughput 9.44053K wps
[Epoch 118 Batch 150/162] avg loss 0.00141472, throughput 9.32922K wps
Begin Testing...
[Epoch 118] train avg loss 0.00128826, dev acc 0.9022, dev avg loss 0.228901, throughput 9.39565K wps
[Epoch 119 Batch 30/162] avg loss 0.00128634, throughput 9.57712K wps
[Epoch 119 Batch 60/162] avg loss 0.00136237, throughput 9.38513K wps
[Epoch 119 Batch 90/162] avg loss 0.0011895, throughput 9.45068K wps
[Epoch 119 Batch 120/162] avg loss 0.00135235, throughput 9.43218K wps
[Epoch 119 Batch 150/162] avg loss 0.00132737, throughput 9.3261K wps
Begin Testing...
[Epoch 119] train avg loss 0.00130789, dev acc 0.9056, dev avg loss 0.228601, throughput 9.42102K wps
Observed Improvement.
Begin Testing...
[Epoch 120 Batch 30/162] avg loss 0.00137005, throughput 9.60404K wps
[Epoch 120 Batch 60/162] avg loss 0.0012713, throughput 9.42281K wps
[Epoch 120 Batch 90/162] avg loss 0.00122234, throughput 9.48131K wps
[Epoch 120 Batch 120/162] avg loss 0.00120558, throughput 9.31326K wps
[Epoch 120 Batch 150/162] avg loss 0.00149161, throughput 9.53556K wps
Begin Testing...
[Epoch 120] train avg loss 0.0012872, dev acc 0.9044, dev avg loss 0.228555, throughput 9.45783K wps
[Epoch 121 Batch 30/162] avg loss 0.0012262, throughput 9.58231K wps
[Epoch 121 Batch 60/162] avg loss 0.00119325, throughput 9.41686K wps
[Epoch 121 Batch 90/162] avg loss 0.00137732, throughput 9.36033K wps
[Epoch 121 Batch 120/162] avg loss 0.00130451, throughput 9.22065K wps
[Epoch 121 Batch 150/162] avg loss 0.00132258, throughput 9.37125K wps
Begin Testing...
[Epoch 121] train avg loss 0.00127187, dev acc 0.9033, dev avg loss 0.228632, throughput 9.4175K wps
[Epoch 122 Batch 30/162] avg loss 0.00121094, throughput 9.84752K wps
[Epoch 122 Batch 60/162] avg loss 0.00122628, throughput 9.28725K wps
[Epoch 122 Batch 90/162] avg loss 0.00130944, throughput 9.4648K wps
[Epoch 122 Batch 120/162] avg loss 0.00130713, throughput 9.44772K wps
[Epoch 122 Batch 150/162] avg loss 0.00119844, throughput 9.39832K wps
Begin Testing...
[Epoch 122] train avg loss 0.00124569, dev acc 0.9044, dev avg loss 0.229701, throughput 9.47457K wps
[Epoch 123 Batch 30/162] avg loss 0.00116474, throughput 9.66057K wps
[Epoch 123 Batch 60/162] avg loss 0.00127263, throughput 9.51093K wps
[Epoch 123 Batch 90/162] avg loss 0.00128279, throughput 9.38929K wps
[Epoch 123 Batch 120/162] avg loss 0.00123709, throughput 9.30459K wps
[Epoch 123 Batch 150/162] avg loss 0.00125334, throughput 9.4466K wps
Begin Testing...
[Epoch 123] train avg loss 0.00123741, dev acc 0.9044, dev avg loss 0.228212, throughput 9.47283K wps
[Epoch 124 Batch 30/162] avg loss 0.00118146, throughput 9.4424K wps
[Epoch 124 Batch 60/162] avg loss 0.00118927, throughput 9.31098K wps
[Epoch 124 Batch 90/162] avg loss 0.00120613, throughput 9.29511K wps
[Epoch 124 Batch 120/162] avg loss 0.00135543, throughput 9.35203K wps
[Epoch 124 Batch 150/162] avg loss 0.00123358, throughput 9.32058K wps
Begin Testing...
[Epoch 124] train avg loss 0.00123251, dev acc 0.9033, dev avg loss 0.228666, throughput 9.33817K wps
[Epoch 125 Batch 30/162] avg loss 0.0012395, throughput 9.47283K wps
[Epoch 125 Batch 60/162] avg loss 0.00129361, throughput 9.2552K wps
[Epoch 125 Batch 90/162] avg loss 0.00122874, throughput 9.3266K wps
[Epoch 125 Batch 120/162] avg loss 0.00108636, throughput 9.42394K wps
[Epoch 125 Batch 150/162] avg loss 0.00122701, throughput 9.3133K wps
Begin Testing...
[Epoch 125] train avg loss 0.00121352, dev acc 0.9033, dev avg loss 0.228839, throughput 9.35947K wps
[Epoch 126 Batch 30/162] avg loss 0.00119009, throughput 9.78623K wps
[Epoch 126 Batch 60/162] avg loss 0.0011559, throughput 9.51151K wps
[Epoch 126 Batch 90/162] avg loss 0.0012021, throughput 9.29546K wps
[Epoch 126 Batch 120/162] avg loss 0.00121666, throughput 9.47087K wps
[Epoch 126 Batch 150/162] avg loss 0.00130179, throughput 9.38544K wps
Begin Testing...
[Epoch 126] train avg loss 0.00120505, dev acc 0.9033, dev avg loss 0.228586, throughput 9.47857K wps
[Epoch 127 Batch 30/162] avg loss 0.0011845, throughput 9.50964K wps
[Epoch 127 Batch 60/162] avg loss 0.00109873, throughput 9.2935K wps
[Epoch 127 Batch 90/162] avg loss 0.00119251, throughput 9.53522K wps
[Epoch 127 Batch 120/162] avg loss 0.00111289, throughput 9.42894K wps
[Epoch 127 Batch 150/162] avg loss 0.0011918, throughput 9.44186K wps
Begin Testing...
[Epoch 127] train avg loss 0.00114551, dev acc 0.9033, dev avg loss 0.228936, throughput 9.42316K wps
[Epoch 128 Batch 30/162] avg loss 0.00104349, throughput 9.57247K wps
[Epoch 128 Batch 60/162] avg loss 0.00124209, throughput 9.50232K wps
[Epoch 128 Batch 90/162] avg loss 0.00123014, throughput 9.46988K wps
[Epoch 128 Batch 120/162] avg loss 0.00119116, throughput 9.58464K wps
[Epoch 128 Batch 150/162] avg loss 0.00117218, throughput 9.27334K wps
Begin Testing...
[Epoch 128] train avg loss 0.00118989, dev acc 0.9044, dev avg loss 0.229064, throughput 9.46602K wps
[Epoch 129 Batch 30/162] avg loss 0.00108442, throughput 9.58023K wps
[Epoch 129 Batch 60/162] avg loss 0.00106755, throughput 9.61245K wps
[Epoch 129 Batch 90/162] avg loss 0.00134271, throughput 9.58557K wps
[Epoch 129 Batch 120/162] avg loss 0.00113374, throughput 9.60151K wps
[Epoch 129 Batch 150/162] avg loss 0.00110263, throughput 9.32304K wps
Begin Testing...
[Epoch 129] train avg loss 0.00113962, dev acc 0.9033, dev avg loss 0.229111, throughput 9.54066K wps
[Epoch 130 Batch 30/162] avg loss 0.00105649, throughput 9.61768K wps
[Epoch 130 Batch 60/162] avg loss 0.00123556, throughput 9.46508K wps
[Epoch 130 Batch 90/162] avg loss 0.00129504, throughput 9.31678K wps
[Epoch 130 Batch 120/162] avg loss 0.00114731, throughput 9.47414K wps
[Epoch 130 Batch 150/162] avg loss 0.00105521, throughput 9.50802K wps
Begin Testing...
[Epoch 130] train avg loss 0.00116039, dev acc 0.9033, dev avg loss 0.229271, throughput 9.45845K wps
[Epoch 131 Batch 30/162] avg loss 0.00115036, throughput 9.6991K wps
[Epoch 131 Batch 60/162] avg loss 0.00106929, throughput 9.63785K wps
[Epoch 131 Batch 90/162] avg loss 0.00107844, throughput 9.41099K wps
[Epoch 131 Batch 120/162] avg loss 0.00113583, throughput 9.37172K wps
[Epoch 131 Batch 150/162] avg loss 0.00114616, throughput 9.37083K wps
Begin Testing...
[Epoch 131] train avg loss 0.0011189, dev acc 0.9011, dev avg loss 0.228904, throughput 9.48262K wps
[Epoch 132 Batch 30/162] avg loss 0.000979309, throughput 9.53841K wps
[Epoch 132 Batch 60/162] avg loss 0.00106597, throughput 9.29468K wps
[Epoch 132 Batch 90/162] avg loss 0.0010564, throughput 9.4764K wps
[Epoch 132 Batch 120/162] avg loss 0.00116579, throughput 9.54616K wps
[Epoch 132 Batch 150/162] avg loss 0.00108202, throughput 9.35807K wps
Begin Testing...
[Epoch 132] train avg loss 0.00107284, dev acc 0.9022, dev avg loss 0.228526, throughput 9.45042K wps
[Epoch 133 Batch 30/162] avg loss 0.00108948, throughput 9.5827K wps
[Epoch 133 Batch 60/162] avg loss 0.00104176, throughput 9.49383K wps
[Epoch 133 Batch 90/162] avg loss 0.00116041, throughput 9.49868K wps
[Epoch 133 Batch 120/162] avg loss 0.0011457, throughput 9.49356K wps
[Epoch 133 Batch 150/162] avg loss 0.00099983, throughput 9.40878K wps
Begin Testing...
[Epoch 133] train avg loss 0.00109257, dev acc 0.9044, dev avg loss 0.228861, throughput 9.49567K wps
[Epoch 134 Batch 30/162] avg loss 0.00110642, throughput 9.60995K wps
[Epoch 134 Batch 60/162] avg loss 0.00116395, throughput 9.34744K wps
[Epoch 134 Batch 90/162] avg loss 0.0010594, throughput 9.38789K wps
[Epoch 134 Batch 120/162] avg loss 0.00108319, throughput 9.39656K wps
[Epoch 134 Batch 150/162] avg loss 0.000966747, throughput 9.46406K wps
Begin Testing...
[Epoch 134] train avg loss 0.00107372, dev acc 0.9011, dev avg loss 0.228501, throughput 9.42534K wps
[Epoch 135 Batch 30/162] avg loss 0.00107663, throughput 9.77979K wps
[Epoch 135 Batch 60/162] avg loss 0.00107761, throughput 9.40427K wps
[Epoch 135 Batch 90/162] avg loss 0.0010647, throughput 9.55121K wps
[Epoch 135 Batch 120/162] avg loss 0.00111942, throughput 9.21051K wps
[Epoch 135 Batch 150/162] avg loss 0.00113706, throughput 9.4508K wps
Begin Testing...
[Epoch 135] train avg loss 0.00108559, dev acc 0.9033, dev avg loss 0.22798, throughput 9.4589K wps
[Epoch 136 Batch 30/162] avg loss 0.00109232, throughput 9.72931K wps
[Epoch 136 Batch 60/162] avg loss 0.0010873, throughput 9.30654K wps
[Epoch 136 Batch 90/162] avg loss 0.00104022, throughput 9.32795K wps
[Epoch 136 Batch 120/162] avg loss 0.00110663, throughput 9.5753K wps
[Epoch 136 Batch 150/162] avg loss 0.00111288, throughput 9.25394K wps
Begin Testing...
[Epoch 136] train avg loss 0.00108823, dev acc 0.9011, dev avg loss 0.227837, throughput 9.42802K wps
[Epoch 137 Batch 30/162] avg loss 0.00104438, throughput 9.69426K wps
[Epoch 137 Batch 60/162] avg loss 0.00104891, throughput 9.32602K wps
[Epoch 137 Batch 90/162] avg loss 0.00111233, throughput 9.38392K wps
[Epoch 137 Batch 120/162] avg loss 0.00100036, throughput 9.40776K wps
[Epoch 137 Batch 150/162] avg loss 0.000993474, throughput 9.44144K wps
Begin Testing...
[Epoch 137] train avg loss 0.00105476, dev acc 0.9000, dev avg loss 0.22866, throughput 9.44667K wps
[Epoch 138 Batch 30/162] avg loss 0.00107143, throughput 9.61081K wps
[Epoch 138 Batch 60/162] avg loss 0.0010687, throughput 9.33078K wps
[Epoch 138 Batch 90/162] avg loss 0.00110234, throughput 9.49468K wps
[Epoch 138 Batch 120/162] avg loss 0.00108722, throughput 9.51631K wps
[Epoch 138 Batch 150/162] avg loss 0.000957869, throughput 9.36028K wps
Begin Testing...
[Epoch 138] train avg loss 0.00106298, dev acc 0.9022, dev avg loss 0.229251, throughput 9.45086K wps
[Epoch 139 Batch 30/162] avg loss 0.000957169, throughput 9.5904K wps
[Epoch 139 Batch 60/162] avg loss 0.00102036, throughput 9.58199K wps
[Epoch 139 Batch 90/162] avg loss 0.00113437, throughput 9.23476K wps
[Epoch 139 Batch 120/162] avg loss 0.000988312, throughput 9.51335K wps
[Epoch 139 Batch 150/162] avg loss 0.00109621, throughput 9.31113K wps
Begin Testing...
[Epoch 139] train avg loss 0.00103999, dev acc 0.9044, dev avg loss 0.227957, throughput 9.42724K wps
[Epoch 140 Batch 30/162] avg loss 0.00106384, throughput 9.57182K wps
[Epoch 140 Batch 60/162] avg loss 0.00107761, throughput 9.46286K wps
[Epoch 140 Batch 90/162] avg loss 0.000938521, throughput 9.39715K wps
[Epoch 140 Batch 120/162] avg loss 0.000966835, throughput 9.30758K wps
[Epoch 140 Batch 150/162] avg loss 0.00103668, throughput 9.46839K wps
Begin Testing...
[Epoch 140] train avg loss 0.0010259, dev acc 0.9056, dev avg loss 0.228838, throughput 9.44648K wps
Observed Improvement.
Begin Testing...
[Epoch 141 Batch 30/162] avg loss 0.00112114, throughput 9.68545K wps
[Epoch 141 Batch 60/162] avg loss 0.000958378, throughput 9.33722K wps
[Epoch 141 Batch 90/162] avg loss 0.000975968, throughput 9.2644K wps
[Epoch 141 Batch 120/162] avg loss 0.00090705, throughput 9.37425K wps
[Epoch 141 Batch 150/162] avg loss 0.00104356, throughput 9.52963K wps
Begin Testing...
[Epoch 141] train avg loss 0.00100559, dev acc 0.9067, dev avg loss 0.228548, throughput 9.4428K wps
Observed Improvement.
Begin Testing...
[Epoch 142 Batch 30/162] avg loss 0.000821265, throughput 9.55205K wps
[Epoch 142 Batch 60/162] avg loss 0.00101624, throughput 9.35026K wps
[Epoch 142 Batch 90/162] avg loss 0.00105354, throughput 9.5037K wps
[Epoch 142 Batch 120/162] avg loss 0.000924523, throughput 9.27311K wps
[Epoch 142 Batch 150/162] avg loss 0.000953305, throughput 9.30333K wps
Begin Testing...
[Epoch 142] train avg loss 0.000961808, dev acc 0.9022, dev avg loss 0.229952, throughput 9.38727K wps
[Epoch 143 Batch 30/162] avg loss 0.000942869, throughput 9.70502K wps
[Epoch 143 Batch 60/162] avg loss 0.000912932, throughput 9.33101K wps
[Epoch 143 Batch 90/162] avg loss 0.00104814, throughput 9.46734K wps
[Epoch 143 Batch 120/162] avg loss 0.00108221, throughput 9.37362K wps
[Epoch 143 Batch 150/162] avg loss 0.000942595, throughput 9.43418K wps
Begin Testing...
[Epoch 143] train avg loss 0.000986873, dev acc 0.9056, dev avg loss 0.228381, throughput 9.47072K wps
[Epoch 144 Batch 30/162] avg loss 0.00105801, throughput 9.70251K wps
[Epoch 144 Batch 60/162] avg loss 0.000963123, throughput 9.3213K wps
[Epoch 144 Batch 90/162] avg loss 0.000942768, throughput 9.54164K wps
[Epoch 144 Batch 120/162] avg loss 0.000861258, throughput 9.5905K wps
[Epoch 144 Batch 150/162] avg loss 0.00088067, throughput 9.5672K wps
Begin Testing...
[Epoch 144] train avg loss 0.000941106, dev acc 0.9056, dev avg loss 0.229054, throughput 9.52563K wps
[Epoch 145 Batch 30/162] avg loss 0.000965914, throughput 9.5871K wps
[Epoch 145 Batch 60/162] avg loss 0.000958447, throughput 9.47776K wps
[Epoch 145 Batch 90/162] avg loss 0.00084588, throughput 9.58608K wps
[Epoch 145 Batch 120/162] avg loss 0.000916399, throughput 9.33919K wps
[Epoch 145 Batch 150/162] avg loss 0.00104662, throughput 9.39961K wps
Begin Testing...
[Epoch 145] train avg loss 0.000954122, dev acc 0.9011, dev avg loss 0.22953, throughput 9.47543K wps
[Epoch 146 Batch 30/162] avg loss 0.00100339, throughput 9.73736K wps
[Epoch 146 Batch 60/162] avg loss 0.00100562, throughput 9.41824K wps
[Epoch 146 Batch 90/162] avg loss 0.000984535, throughput 9.46085K wps
[Epoch 146 Batch 120/162] avg loss 0.000998737, throughput 9.55867K wps
[Epoch 146 Batch 150/162] avg loss 0.000881692, throughput 9.27074K wps
Begin Testing...
[Epoch 146] train avg loss 0.000976323, dev acc 0.9011, dev avg loss 0.228762, throughput 9.47276K wps
[Epoch 147 Batch 30/162] avg loss 0.000865802, throughput 9.55779K wps
[Epoch 147 Batch 60/162] avg loss 0.000907602, throughput 9.46399K wps
[Epoch 147 Batch 90/162] avg loss 0.000915244, throughput 9.54827K wps
[Epoch 147 Batch 120/162] avg loss 0.000985189, throughput 9.54012K wps
[Epoch 147 Batch 150/162] avg loss 0.00100324, throughput 9.43135K wps
Begin Testing...
[Epoch 147] train avg loss 0.000932076, dev acc 0.9044, dev avg loss 0.228536, throughput 9.51859K wps
[Epoch 148 Batch 30/162] avg loss 0.000906716, throughput 9.81437K wps
[Epoch 148 Batch 60/162] avg loss 0.000977421, throughput 9.40124K wps
[Epoch 148 Batch 90/162] avg loss 0.00106852, throughput 9.45914K wps
[Epoch 148 Batch 120/162] avg loss 0.000846776, throughput 9.59332K wps
[Epoch 148 Batch 150/162] avg loss 0.000877037, throughput 9.37741K wps
Begin Testing...
[Epoch 148] train avg loss 0.000936031, dev acc 0.9067, dev avg loss 0.228616, throughput 9.53242K wps
Observed Improvement.
Begin Testing...
[Epoch 149 Batch 30/162] avg loss 0.00101079, throughput 9.66898K wps
[Epoch 149 Batch 60/162] avg loss 0.000872563, throughput 9.40694K wps
[Epoch 149 Batch 90/162] avg loss 0.000847489, throughput 9.57006K wps
[Epoch 149 Batch 120/162] avg loss 0.000891459, throughput 9.47464K wps
[Epoch 149 Batch 150/162] avg loss 0.000932588, throughput 9.40638K wps
Begin Testing...
[Epoch 149] train avg loss 0.000927093, dev acc 0.9056, dev avg loss 0.228736, throughput 9.50608K wps
[Epoch 150 Batch 30/162] avg loss 0.000835754, throughput 9.7475K wps
[Epoch 150 Batch 60/162] avg loss 0.00102018, throughput 9.29251K wps
[Epoch 150 Batch 90/162] avg loss 0.000868556, throughput 9.52752K wps
[Epoch 150 Batch 120/162] avg loss 0.00083498, throughput 9.32095K wps
[Epoch 150 Batch 150/162] avg loss 0.00101366, throughput 9.41958K wps
Begin Testing...
[Epoch 150] train avg loss 0.000915832, dev acc 0.9044, dev avg loss 0.229274, throughput 9.47412K wps
[Epoch 151 Batch 30/162] avg loss 0.000971494, throughput 9.71595K wps
[Epoch 151 Batch 60/162] avg loss 0.000955748, throughput 9.38522K wps
[Epoch 151 Batch 90/162] avg loss 0.000823767, throughput 9.43338K wps
[Epoch 151 Batch 120/162] avg loss 0.000827819, throughput 9.46515K wps
[Epoch 151 Batch 150/162] avg loss 0.000894189, throughput 9.2818K wps
Begin Testing...
[Epoch 151] train avg loss 0.000892665, dev acc 0.9078, dev avg loss 0.229456, throughput 9.44236K wps
Observed Improvement.
Begin Testing...
[Epoch 152 Batch 30/162] avg loss 0.000820069, throughput 9.44778K wps
[Epoch 152 Batch 60/162] avg loss 0.000929272, throughput 9.27564K wps
[Epoch 152 Batch 90/162] avg loss 0.0007992, throughput 9.27565K wps
[Epoch 152 Batch 120/162] avg loss 0.00102517, throughput 9.44982K wps
[Epoch 152 Batch 150/162] avg loss 0.00087949, throughput 9.36694K wps
Begin Testing...
[Epoch 152] train avg loss 0.000895597, dev acc 0.9089, dev avg loss 0.22949, throughput 9.37013K wps
Observed Improvement.
Begin Testing...
[Epoch 153 Batch 30/162] avg loss 0.000959217, throughput 9.58099K wps
[Epoch 153 Batch 60/162] avg loss 0.000789498, throughput 9.40938K wps
[Epoch 153 Batch 90/162] avg loss 0.000793005, throughput 9.52987K wps
[Epoch 153 Batch 120/162] avg loss 0.00080445, throughput 9.233K wps
[Epoch 153 Batch 150/162] avg loss 0.000867884, throughput 9.29468K wps
Begin Testing...
[Epoch 153] train avg loss 0.000850673, dev acc 0.9044, dev avg loss 0.229549, throughput 9.40224K wps
[Epoch 154 Batch 30/162] avg loss 0.00084199, throughput 9.65937K wps
[Epoch 154 Batch 60/162] avg loss 0.000888722, throughput 9.33859K wps
[Epoch 154 Batch 90/162] avg loss 0.000827623, throughput 9.4215K wps
[Epoch 154 Batch 120/162] avg loss 0.000788347, throughput 9.26585K wps
[Epoch 154 Batch 150/162] avg loss 0.000935032, throughput 9.38254K wps
Begin Testing...
[Epoch 154] train avg loss 0.000888951, dev acc 0.9078, dev avg loss 0.229658, throughput 9.41787K wps
[Epoch 155 Batch 30/162] avg loss 0.000774965, throughput 9.54991K wps
[Epoch 155 Batch 60/162] avg loss 0.00100777, throughput 9.55815K wps
[Epoch 155 Batch 90/162] avg loss 0.000769997, throughput 9.39305K wps
[Epoch 155 Batch 120/162] avg loss 0.000875706, throughput 9.32057K wps
[Epoch 155 Batch 150/162] avg loss 0.000774844, throughput 9.43401K wps
Begin Testing...
[Epoch 155] train avg loss 0.000850279, dev acc 0.9011, dev avg loss 0.229562, throughput 9.4298K wps
[Epoch 156 Batch 30/162] avg loss 0.000907131, throughput 9.63829K wps
[Epoch 156 Batch 60/162] avg loss 0.00085277, throughput 9.39863K wps
[Epoch 156 Batch 90/162] avg loss 0.000764008, throughput 9.53824K wps
[Epoch 156 Batch 120/162] avg loss 0.000800341, throughput 9.5356K wps
[Epoch 156 Batch 150/162] avg loss 0.000948413, throughput 9.45735K wps
Begin Testing...
[Epoch 156] train avg loss 0.00085469, dev acc 0.9044, dev avg loss 0.229496, throughput 9.50818K wps
[Epoch 157 Batch 30/162] avg loss 0.000879339, throughput 9.46572K wps
[Epoch 157 Batch 60/162] avg loss 0.000864642, throughput 9.36833K wps
[Epoch 157 Batch 90/162] avg loss 0.000871848, throughput 9.36277K wps
[Epoch 157 Batch 120/162] avg loss 0.000965739, throughput 9.37806K wps
[Epoch 157 Batch 150/162] avg loss 0.000831629, throughput 9.41366K wps
Begin Testing...
[Epoch 157] train avg loss 0.000882327, dev acc 0.9067, dev avg loss 0.22992, throughput 9.39843K wps
[Epoch 158 Batch 30/162] avg loss 0.000853763, throughput 9.77837K wps
[Epoch 158 Batch 60/162] avg loss 0.000805673, throughput 9.4564K wps
[Epoch 158 Batch 90/162] avg loss 0.000873056, throughput 9.40313K wps
[Epoch 158 Batch 120/162] avg loss 0.000694609, throughput 9.47377K wps
[Epoch 158 Batch 150/162] avg loss 0.000844271, throughput 9.32317K wps
Begin Testing...
[Epoch 158] train avg loss 0.000817966, dev acc 0.9067, dev avg loss 0.230092, throughput 9.48673K wps
[Epoch 159 Batch 30/162] avg loss 0.000878048, throughput 9.66775K wps
[Epoch 159 Batch 60/162] avg loss 0.000881621, throughput 9.25746K wps
[Epoch 159 Batch 90/162] avg loss 0.000822146, throughput 9.35219K wps
[Epoch 159 Batch 120/162] avg loss 0.000830385, throughput 9.44159K wps
[Epoch 159 Batch 150/162] avg loss 0.000816958, throughput 9.36659K wps
Begin Testing...
[Epoch 159] train avg loss 0.000840466, dev acc 0.9078, dev avg loss 0.229918, throughput 9.43437K wps
[Epoch 160 Batch 30/162] avg loss 0.000858648, throughput 9.4112K wps
[Epoch 160 Batch 60/162] avg loss 0.000867481, throughput 9.31422K wps
[Epoch 160 Batch 90/162] avg loss 0.000784092, throughput 9.47064K wps
[Epoch 160 Batch 120/162] avg loss 0.000820563, throughput 9.34253K wps
[Epoch 160 Batch 150/162] avg loss 0.000820562, throughput 9.28088K wps
Begin Testing...
[Epoch 160] train avg loss 0.000818556, dev acc 0.9033, dev avg loss 0.229933, throughput 9.35925K wps
[Epoch 161 Batch 30/162] avg loss 0.000799735, throughput 9.44699K wps
[Epoch 161 Batch 60/162] avg loss 0.000898449, throughput 9.34426K wps
[Epoch 161 Batch 90/162] avg loss 0.000808056, throughput 9.50412K wps
[Epoch 161 Batch 120/162] avg loss 0.000735638, throughput 9.43799K wps
[Epoch 161 Batch 150/162] avg loss 0.000837374, throughput 9.44092K wps
Begin Testing...
[Epoch 161] train avg loss 0.000821722, dev acc 0.9078, dev avg loss 0.229726, throughput 9.42505K wps
[Epoch 162 Batch 30/162] avg loss 0.000754611, throughput 9.50442K wps
[Epoch 162 Batch 60/162] avg loss 0.000849603, throughput 9.36478K wps
[Epoch 162 Batch 90/162] avg loss 0.000951688, throughput 9.29134K wps
[Epoch 162 Batch 120/162] avg loss 0.000720309, throughput 9.51358K wps
[Epoch 162 Batch 150/162] avg loss 0.000802889, throughput 9.34681K wps
Begin Testing...
[Epoch 162] train avg loss 0.000820604, dev acc 0.9044, dev avg loss 0.231043, throughput 9.39716K wps
[Epoch 163 Batch 30/162] avg loss 0.000715213, throughput 9.72384K wps
[Epoch 163 Batch 60/162] avg loss 0.000839963, throughput 9.29752K wps
[Epoch 163 Batch 90/162] avg loss 0.000832573, throughput 9.42436K wps
[Epoch 163 Batch 120/162] avg loss 0.000749261, throughput 9.34046K wps
[Epoch 163 Batch 150/162] avg loss 0.000819963, throughput 9.38097K wps
Begin Testing...
[Epoch 163] train avg loss 0.000796189, dev acc 0.9078, dev avg loss 0.23016, throughput 9.44234K wps
[Epoch 164 Batch 30/162] avg loss 0.000863869, throughput 9.6428K wps
[Epoch 164 Batch 60/162] avg loss 0.000761935, throughput 9.27K wps
[Epoch 164 Batch 90/162] avg loss 0.000706601, throughput 9.37906K wps
[Epoch 164 Batch 120/162] avg loss 0.000825802, throughput 9.34386K wps
[Epoch 164 Batch 150/162] avg loss 0.00075793, throughput 9.48249K wps
Begin Testing...
[Epoch 164] train avg loss 0.000780306, dev acc 0.9044, dev avg loss 0.229996, throughput 9.41374K wps
[Epoch 165 Batch 30/162] avg loss 0.000812061, throughput 9.45496K wps
[Epoch 165 Batch 60/162] avg loss 0.000837612, throughput 9.49217K wps
[Epoch 165 Batch 90/162] avg loss 0.000722365, throughput 9.31865K wps
[Epoch 165 Batch 120/162] avg loss 0.000884148, throughput 9.63036K wps
[Epoch 165 Batch 150/162] avg loss 0.000732633, throughput 9.31833K wps
Begin Testing...
[Epoch 165] train avg loss 0.000793756, dev acc 0.9067, dev avg loss 0.230583, throughput 9.45312K wps
[Epoch 166 Batch 30/162] avg loss 0.000841908, throughput 9.44649K wps
[Epoch 166 Batch 60/162] avg loss 0.000810591, throughput 9.48681K wps
[Epoch 166 Batch 90/162] avg loss 0.000711833, throughput 9.24485K wps
[Epoch 166 Batch 120/162] avg loss 0.000662113, throughput 9.36932K wps
[Epoch 166 Batch 150/162] avg loss 0.000713825, throughput 9.32987K wps
Begin Testing...
[Epoch 166] train avg loss 0.000746068, dev acc 0.9033, dev avg loss 0.23101, throughput 9.39924K wps
[Epoch 167 Batch 30/162] avg loss 0.000832787, throughput 9.5936K wps
[Epoch 167 Batch 60/162] avg loss 0.000845972, throughput 9.25392K wps
[Epoch 167 Batch 90/162] avg loss 0.000713637, throughput 9.41786K wps
[Epoch 167 Batch 120/162] avg loss 0.00078285, throughput 9.50236K wps
[Epoch 167 Batch 150/162] avg loss 0.000782128, throughput 9.4687K wps
Begin Testing...
[Epoch 167] train avg loss 0.000789285, dev acc 0.9078, dev avg loss 0.230329, throughput 9.42055K wps
[Epoch 168 Batch 30/162] avg loss 0.000795281, throughput 9.60908K wps
[Epoch 168 Batch 60/162] avg loss 0.00079207, throughput 9.50232K wps
[Epoch 168 Batch 90/162] avg loss 0.000852323, throughput 9.35483K wps
[Epoch 168 Batch 120/162] avg loss 0.000679751, throughput 9.39914K wps
[Epoch 168 Batch 150/162] avg loss 0.000854166, throughput 9.21979K wps
Begin Testing...
[Epoch 168] train avg loss 0.000781258, dev acc 0.9044, dev avg loss 0.230259, throughput 9.40941K wps
[Epoch 169 Batch 30/162] avg loss 0.000776614, throughput 9.62615K wps
[Epoch 169 Batch 60/162] avg loss 0.000875909, throughput 9.53065K wps
[Epoch 169 Batch 90/162] avg loss 0.000671984, throughput 9.21335K wps
[Epoch 169 Batch 120/162] avg loss 0.000689701, throughput 9.58552K wps
[Epoch 169 Batch 150/162] avg loss 0.000816773, throughput 9.46769K wps
Begin Testing...
[Epoch 169] train avg loss 0.000764849, dev acc 0.9067, dev avg loss 0.229655, throughput 9.45089K wps
[Epoch 170 Batch 30/162] avg loss 0.000808136, throughput 9.46165K wps
[Epoch 170 Batch 60/162] avg loss 0.000802675, throughput 9.35008K wps
[Epoch 170 Batch 90/162] avg loss 0.0007414, throughput 9.44824K wps
[Epoch 170 Batch 120/162] avg loss 0.000648442, throughput 9.55437K wps
[Epoch 170 Batch 150/162] avg loss 0.000769905, throughput 9.26401K wps
Begin Testing...
[Epoch 170] train avg loss 0.000747549, dev acc 0.9067, dev avg loss 0.230819, throughput 9.40775K wps
[Epoch 171 Batch 30/162] avg loss 0.000678461, throughput 9.64248K wps
[Epoch 171 Batch 60/162] avg loss 0.000666961, throughput 9.62485K wps
[Epoch 171 Batch 90/162] avg loss 0.000742727, throughput 9.4155K wps
[Epoch 171 Batch 120/162] avg loss 0.000899328, throughput 9.35346K wps
[Epoch 171 Batch 150/162] avg loss 0.000731472, throughput 9.60137K wps
Begin Testing...
[Epoch 171] train avg loss 0.000742684, dev acc 0.9089, dev avg loss 0.231045, throughput 9.5071K wps
Observed Improvement.
Begin Testing...
[Epoch 172 Batch 30/162] avg loss 0.000746772, throughput 9.7299K wps
[Epoch 172 Batch 60/162] avg loss 0.000589927, throughput 9.62839K wps
[Epoch 172 Batch 90/162] avg loss 0.00076548, throughput 9.37396K wps
[Epoch 172 Batch 120/162] avg loss 0.000742776, throughput 9.45993K wps
[Epoch 172 Batch 150/162] avg loss 0.000746811, throughput 9.47454K wps
Begin Testing...
[Epoch 172] train avg loss 0.000723035, dev acc 0.9044, dev avg loss 0.23121, throughput 9.50807K wps
[Epoch 173 Batch 30/162] avg loss 0.000712162, throughput 9.59228K wps
[Epoch 173 Batch 60/162] avg loss 0.000737581, throughput 9.25274K wps
[Epoch 173 Batch 90/162] avg loss 0.000687732, throughput 9.38709K wps
[Epoch 173 Batch 120/162] avg loss 0.000663602, throughput 9.41305K wps
[Epoch 173 Batch 150/162] avg loss 0.000692658, throughput 9.26446K wps
Begin Testing...
[Epoch 173] train avg loss 0.000694701, dev acc 0.9000, dev avg loss 0.232497, throughput 9.38134K wps
[Epoch 174 Batch 30/162] avg loss 0.000731221, throughput 9.50091K wps
[Epoch 174 Batch 60/162] avg loss 0.000719155, throughput 9.23878K wps
[Epoch 174 Batch 90/162] avg loss 0.000663538, throughput 9.44652K wps
[Epoch 174 Batch 120/162] avg loss 0.000675716, throughput 9.42645K wps
[Epoch 174 Batch 150/162] avg loss 0.000711513, throughput 9.25142K wps
Begin Testing...
[Epoch 174] train avg loss 0.000706911, dev acc 0.9056, dev avg loss 0.231444, throughput 9.39655K wps
[Epoch 175 Batch 30/162] avg loss 0.000767158, throughput 9.5946K wps
[Epoch 175 Batch 60/162] avg loss 0.000733519, throughput 9.50083K wps
[Epoch 175 Batch 90/162] avg loss 0.000713494, throughput 9.56775K wps
[Epoch 175 Batch 120/162] avg loss 0.000733291, throughput 9.39904K wps
[Epoch 175 Batch 150/162] avg loss 0.000801811, throughput 9.4328K wps
Begin Testing...
[Epoch 175] train avg loss 0.000747899, dev acc 0.9067, dev avg loss 0.233094, throughput 9.48053K wps
[Epoch 176 Batch 30/162] avg loss 0.00066666, throughput 9.66202K wps
[Epoch 176 Batch 60/162] avg loss 0.000775135, throughput 9.4966K wps
[Epoch 176 Batch 90/162] avg loss 0.000802391, throughput 9.53696K wps
[Epoch 176 Batch 120/162] avg loss 0.000703745, throughput 9.39399K wps
[Epoch 176 Batch 150/162] avg loss 0.000696512, throughput 9.22724K wps
Begin Testing...
[Epoch 176] train avg loss 0.000724909, dev acc 0.9044, dev avg loss 0.231194, throughput 9.45115K wps
[Epoch 177 Batch 30/162] avg loss 0.000684545, throughput 9.54557K wps
[Epoch 177 Batch 60/162] avg loss 0.000701918, throughput 9.24962K wps
[Epoch 177 Batch 90/162] avg loss 0.000681608, throughput 9.34436K wps
[Epoch 177 Batch 120/162] avg loss 0.000714228, throughput 9.33164K wps
[Epoch 177 Batch 150/162] avg loss 0.000675673, throughput 9.42821K wps
Begin Testing...
[Epoch 177] train avg loss 0.000694616, dev acc 0.9056, dev avg loss 0.231488, throughput 9.36644K wps
[Epoch 178 Batch 30/162] avg loss 0.000760418, throughput 9.46903K wps
[Epoch 178 Batch 60/162] avg loss 0.000600558, throughput 9.38315K wps
[Epoch 178 Batch 90/162] avg loss 0.000710526, throughput 9.45494K wps
[Epoch 178 Batch 120/162] avg loss 0.00067557, throughput 9.56664K wps
[Epoch 178 Batch 150/162] avg loss 0.000675279, throughput 9.4K wps
Begin Testing...
[Epoch 178] train avg loss 0.000692846, dev acc 0.9011, dev avg loss 0.231741, throughput 9.45435K wps
[Epoch 179 Batch 30/162] avg loss 0.000677313, throughput 9.69935K wps
[Epoch 179 Batch 60/162] avg loss 0.000821246, throughput 9.5741K wps
[Epoch 179 Batch 90/162] avg loss 0.000710216, throughput 9.30071K wps
[Epoch 179 Batch 120/162] avg loss 0.00066397, throughput 9.34982K wps
[Epoch 179 Batch 150/162] avg loss 0.000686103, throughput 9.55045K wps
Begin Testing...
[Epoch 179] train avg loss 0.000713245, dev acc 0.9022, dev avg loss 0.231742, throughput 9.49532K wps
[Epoch 180 Batch 30/162] avg loss 0.000609103, throughput 9.61356K wps
[Epoch 180 Batch 60/162] avg loss 0.000760034, throughput 9.26746K wps
[Epoch 180 Batch 90/162] avg loss 0.000703281, throughput 9.68243K wps
[Epoch 180 Batch 120/162] avg loss 0.000801139, throughput 9.5124K wps
[Epoch 180 Batch 150/162] avg loss 0.000632746, throughput 9.35042K wps
Begin Testing...
[Epoch 180] train avg loss 0.000711532, dev acc 0.9067, dev avg loss 0.231945, throughput 9.47659K wps
[Epoch 181 Batch 30/162] avg loss 0.000711511, throughput 9.45099K wps
[Epoch 181 Batch 60/162] avg loss 0.000623781, throughput 9.29366K wps
[Epoch 181 Batch 90/162] avg loss 0.000667808, throughput 9.33848K wps
[Epoch 181 Batch 120/162] avg loss 0.000736698, throughput 9.50643K wps
[Epoch 181 Batch 150/162] avg loss 0.000714884, throughput 9.31541K wps
Begin Testing...
[Epoch 181] train avg loss 0.000696284, dev acc 0.9033, dev avg loss 0.231061, throughput 9.38084K wps
[Epoch 182 Batch 30/162] avg loss 0.000620332, throughput 9.45209K wps
[Epoch 182 Batch 60/162] avg loss 0.000749508, throughput 9.39628K wps
[Epoch 182 Batch 90/162] avg loss 0.000647486, throughput 9.55043K wps
[Epoch 182 Batch 120/162] avg loss 0.000692495, throughput 9.51556K wps
[Epoch 182 Batch 150/162] avg loss 0.000643369, throughput 9.38851K wps
Begin Testing...
[Epoch 182] train avg loss 0.000669018, dev acc 0.9044, dev avg loss 0.23122, throughput 9.46599K wps
[Epoch 183 Batch 30/162] avg loss 0.000610701, throughput 9.67372K wps
[Epoch 183 Batch 60/162] avg loss 0.000651257, throughput 9.45617K wps
[Epoch 183 Batch 90/162] avg loss 0.00066551, throughput 9.51067K wps
[Epoch 183 Batch 120/162] avg loss 0.000624377, throughput 9.40304K wps
[Epoch 183 Batch 150/162] avg loss 0.000776867, throughput 9.38168K wps
Begin Testing...
[Epoch 183] train avg loss 0.000664391, dev acc 0.9089, dev avg loss 0.231548, throughput 9.46634K wps
Observed Improvement.
Begin Testing...
[Epoch 184 Batch 30/162] avg loss 0.000651494, throughput 9.66006K wps
[Epoch 184 Batch 60/162] avg loss 0.000654592, throughput 9.58594K wps
[Epoch 184 Batch 90/162] avg loss 0.00059263, throughput 9.31936K wps
[Epoch 184 Batch 120/162] avg loss 0.000674682, throughput 9.45697K wps
[Epoch 184 Batch 150/162] avg loss 0.00064754, throughput 9.31186K wps
Begin Testing...
[Epoch 184] train avg loss 0.000652255, dev acc 0.9089, dev avg loss 0.231569, throughput 9.45592K wps
Observed Improvement.
Begin Testing...
[Epoch 185 Batch 30/162] avg loss 0.000589779, throughput 9.58115K wps
[Epoch 185 Batch 60/162] avg loss 0.000659999, throughput 9.57001K wps
[Epoch 185 Batch 90/162] avg loss 0.000671599, throughput 9.32366K wps
[Epoch 185 Batch 120/162] avg loss 0.00063428, throughput 9.46326K wps
[Epoch 185 Batch 150/162] avg loss 0.000640916, throughput 9.35336K wps
Begin Testing...
[Epoch 185] train avg loss 0.000642688, dev acc 0.9078, dev avg loss 0.231801, throughput 9.4437K wps
[Epoch 186 Batch 30/162] avg loss 0.000693161, throughput 9.59676K wps
[Epoch 186 Batch 60/162] avg loss 0.000585871, throughput 9.41755K wps
[Epoch 186 Batch 90/162] avg loss 0.000683373, throughput 9.45009K wps
[Epoch 186 Batch 120/162] avg loss 0.000725935, throughput 9.44703K wps
[Epoch 186 Batch 150/162] avg loss 0.000664306, throughput 9.47017K wps
Begin Testing...
[Epoch 186] train avg loss 0.000682463, dev acc 0.9078, dev avg loss 0.231797, throughput 9.4603K wps
[Epoch 187 Batch 30/162] avg loss 0.000708157, throughput 9.69177K wps
[Epoch 187 Batch 60/162] avg loss 0.00070559, throughput 9.56463K wps
[Epoch 187 Batch 90/162] avg loss 0.000675956, throughput 9.29964K wps
[Epoch 187 Batch 120/162] avg loss 0.000683264, throughput 9.56996K wps
[Epoch 187 Batch 150/162] avg loss 0.000634854, throughput 9.41172K wps
Begin Testing...
[Epoch 187] train avg loss 0.000676475, dev acc 0.9089, dev avg loss 0.232516, throughput 9.51082K wps
Observed Improvement.
Begin Testing...
[Epoch 188 Batch 30/162] avg loss 0.000718387, throughput 9.53652K wps
[Epoch 188 Batch 60/162] avg loss 0.000691777, throughput 9.29067K wps
[Epoch 188 Batch 90/162] avg loss 0.000667958, throughput 9.4475K wps
[Epoch 188 Batch 120/162] avg loss 0.000596838, throughput 9.55967K wps
[Epoch 188 Batch 150/162] avg loss 0.000655422, throughput 9.42294K wps
Begin Testing...
[Epoch 188] train avg loss 0.000670589, dev acc 0.9044, dev avg loss 0.232048, throughput 9.45512K wps
[Epoch 189 Batch 30/162] avg loss 0.000606674, throughput 9.58065K wps
[Epoch 189 Batch 60/162] avg loss 0.000585304, throughput 9.49274K wps
[Epoch 189 Batch 90/162] avg loss 0.000700372, throughput 9.58643K wps
[Epoch 189 Batch 120/162] avg loss 0.000648338, throughput 9.49963K wps
[Epoch 189 Batch 150/162] avg loss 0.000619126, throughput 9.33775K wps
Begin Testing...
[Epoch 189] train avg loss 0.000634894, dev acc 0.9089, dev avg loss 0.232148, throughput 9.47615K wps
Observed Improvement.
Begin Testing...
[Epoch 190 Batch 30/162] avg loss 0.000653221, throughput 9.70547K wps
[Epoch 190 Batch 60/162] avg loss 0.000585672, throughput 9.45556K wps
[Epoch 190 Batch 90/162] avg loss 0.000643041, throughput 9.41671K wps
[Epoch 190 Batch 120/162] avg loss 0.00061853, throughput 9.37481K wps
[Epoch 190 Batch 150/162] avg loss 0.000703003, throughput 9.46322K wps
Begin Testing...
[Epoch 190] train avg loss 0.000634917, dev acc 0.9078, dev avg loss 0.232279, throughput 9.47532K wps
[Epoch 191 Batch 30/162] avg loss 0.000533198, throughput 9.65354K wps
[Epoch 191 Batch 60/162] avg loss 0.000533803, throughput 9.28003K wps
[Epoch 191 Batch 90/162] avg loss 0.000612632, throughput 9.4319K wps
[Epoch 191 Batch 120/162] avg loss 0.000613616, throughput 9.43613K wps
[Epoch 191 Batch 150/162] avg loss 0.000676123, throughput 9.35067K wps
Begin Testing...
[Epoch 191] train avg loss 0.000598002, dev acc 0.9067, dev avg loss 0.23351, throughput 9.43996K wps
[Epoch 192 Batch 30/162] avg loss 0.000660867, throughput 9.83276K wps
[Epoch 192 Batch 60/162] avg loss 0.00063091, throughput 9.26189K wps
[Epoch 192 Batch 90/162] avg loss 0.000639797, throughput 9.37645K wps
[Epoch 192 Batch 120/162] avg loss 0.0006449, throughput 9.35541K wps
[Epoch 192 Batch 150/162] avg loss 0.0006095, throughput 9.61931K wps
Begin Testing...
[Epoch 192] train avg loss 0.0006376, dev acc 0.9044, dev avg loss 0.232837, throughput 9.47745K wps
[Epoch 193 Batch 30/162] avg loss 0.000715039, throughput 9.46107K wps
[Epoch 193 Batch 60/162] avg loss 0.000565941, throughput 9.54389K wps
[Epoch 193 Batch 90/162] avg loss 0.000559493, throughput 9.58075K wps
[Epoch 193 Batch 120/162] avg loss 0.000697819, throughput 9.42614K wps
[Epoch 193 Batch 150/162] avg loss 0.000622458, throughput 9.56297K wps
Begin Testing...
[Epoch 193] train avg loss 0.000640433, dev acc 0.9089, dev avg loss 0.232485, throughput 9.49158K wps
Observed Improvement.
Begin Testing...
[Epoch 194 Batch 30/162] avg loss 0.000648807, throughput 9.46559K wps
[Epoch 194 Batch 60/162] avg loss 0.000577163, throughput 9.29604K wps
[Epoch 194 Batch 90/162] avg loss 0.00059807, throughput 9.37325K wps
[Epoch 194 Batch 120/162] avg loss 0.000585821, throughput 9.33261K wps
[Epoch 194 Batch 150/162] avg loss 0.000582434, throughput 9.35057K wps
Begin Testing...
[Epoch 194] train avg loss 0.000592434, dev acc 0.9033, dev avg loss 0.233035, throughput 9.35428K wps
[Epoch 195 Batch 30/162] avg loss 0.000619268, throughput 9.68162K wps
[Epoch 195 Batch 60/162] avg loss 0.000611557, throughput 9.40345K wps
[Epoch 195 Batch 90/162] avg loss 0.000628886, throughput 9.57965K wps
[Epoch 195 Batch 120/162] avg loss 0.000636334, throughput 9.33654K wps
[Epoch 195 Batch 150/162] avg loss 0.000759035, throughput 9.5401K wps
Begin Testing...
[Epoch 195] train avg loss 0.000643481, dev acc 0.9078, dev avg loss 0.232617, throughput 9.48433K wps
[Epoch 196 Batch 30/162] avg loss 0.000608315, throughput 9.65797K wps
[Epoch 196 Batch 60/162] avg loss 0.00065306, throughput 9.42814K wps
[Epoch 196 Batch 90/162] avg loss 0.000534581, throughput 9.35086K wps
[Epoch 196 Batch 120/162] avg loss 0.000602572, throughput 9.48966K wps
[Epoch 196 Batch 150/162] avg loss 0.000586622, throughput 9.37339K wps
Begin Testing...
[Epoch 196] train avg loss 0.000594923, dev acc 0.9044, dev avg loss 0.232647, throughput 9.45368K wps
[Epoch 197 Batch 30/162] avg loss 0.000619987, throughput 9.49041K wps
[Epoch 197 Batch 60/162] avg loss 0.00061732, throughput 9.58454K wps
[Epoch 197 Batch 90/162] avg loss 0.000504403, throughput 9.37569K wps
[Epoch 197 Batch 120/162] avg loss 0.000583386, throughput 9.45838K wps
[Epoch 197 Batch 150/162] avg loss 0.000640873, throughput 9.41211K wps
Begin Testing...
[Epoch 197] train avg loss 0.000589755, dev acc 0.9056, dev avg loss 0.232476, throughput 9.47273K wps
[Epoch 198 Batch 30/162] avg loss 0.000535952, throughput 9.45223K wps
[Epoch 198 Batch 60/162] avg loss 0.000636545, throughput 9.6114K wps
[Epoch 198 Batch 90/162] avg loss 0.00060018, throughput 9.46101K wps
[Epoch 198 Batch 120/162] avg loss 0.000630706, throughput 9.35626K wps
[Epoch 198 Batch 150/162] avg loss 0.00051896, throughput 9.45684K wps
Begin Testing...
[Epoch 198] train avg loss 0.000594385, dev acc 0.9056, dev avg loss 0.232749, throughput 9.44698K wps
[Epoch 199 Batch 30/162] avg loss 0.000544314, throughput 9.59941K wps
[Epoch 199 Batch 60/162] avg loss 0.000654516, throughput 9.45536K wps
[Epoch 199 Batch 90/162] avg loss 0.000550571, throughput 9.26929K wps
[Epoch 199 Batch 120/162] avg loss 0.000626803, throughput 9.32386K wps
[Epoch 199 Batch 150/162] avg loss 0.00055966, throughput 9.26306K wps
Begin Testing...
[Epoch 199] train avg loss 0.000594761, dev acc 0.9067, dev avg loss 0.232584, throughput 9.39105K wps
Test loss 0.175218, test acc 0.9400
Total time cost 443.44s
[Epoch 0 Batch 30/162] avg loss 0.0140443, throughput 7.26772K wps
[Epoch 0 Batch 60/162] avg loss 0.0137886, throughput 9.52007K wps
[Epoch 0 Batch 90/162] avg loss 0.0136215, throughput 9.35669K wps
[Epoch 0 Batch 120/162] avg loss 0.0136137, throughput 9.36988K wps
[Epoch 0 Batch 150/162] avg loss 0.0134178, throughput 9.49737K wps
Begin Testing...
[Epoch 0] train avg loss 0.0136745, dev acc 0.7544, dev avg loss 0.661316, throughput 8.93774K wps
Observed Improvement.
Begin Testing...
[Epoch 1 Batch 30/162] avg loss 0.0132877, throughput 9.66706K wps
[Epoch 1 Batch 60/162] avg loss 0.0131951, throughput 9.3169K wps
[Epoch 1 Batch 90/162] avg loss 0.0130651, throughput 9.42214K wps
[Epoch 1 Batch 120/162] avg loss 0.0129382, throughput 9.39166K wps
[Epoch 1 Batch 150/162] avg loss 0.0127771, throughput 9.27715K wps
Begin Testing...
[Epoch 1] train avg loss 0.0130127, dev acc 0.8333, dev avg loss 0.62987, throughput 9.40816K wps
Observed Improvement.
Begin Testing...
[Epoch 2 Batch 30/162] avg loss 0.012624, throughput 9.8106K wps
[Epoch 2 Batch 60/162] avg loss 0.0125739, throughput 9.31797K wps
[Epoch 2 Batch 90/162] avg loss 0.0123118, throughput 9.52048K wps
[Epoch 2 Batch 120/162] avg loss 0.0124806, throughput 9.28829K wps
[Epoch 2 Batch 150/162] avg loss 0.0121855, throughput 9.3999K wps
Begin Testing...
[Epoch 2] train avg loss 0.0124096, dev acc 0.8567, dev avg loss 0.59557, throughput 9.46992K wps
Observed Improvement.
Begin Testing...
[Epoch 3 Batch 30/162] avg loss 0.0119209, throughput 9.63568K wps
[Epoch 3 Batch 60/162] avg loss 0.0116867, throughput 9.20546K wps
[Epoch 3 Batch 90/162] avg loss 0.0115212, throughput 9.5107K wps
[Epoch 3 Batch 120/162] avg loss 0.0115861, throughput 9.42782K wps
[Epoch 3 Batch 150/162] avg loss 0.0114687, throughput 9.48212K wps
Begin Testing...
[Epoch 3] train avg loss 0.0116283, dev acc 0.8567, dev avg loss 0.557987, throughput 9.45081K wps
Observed Improvement.
Begin Testing...
[Epoch 4 Batch 30/162] avg loss 0.0109711, throughput 9.68876K wps
[Epoch 4 Batch 60/162] avg loss 0.0112254, throughput 9.369K wps
[Epoch 4 Batch 90/162] avg loss 0.0107575, throughput 9.33767K wps
[Epoch 4 Batch 120/162] avg loss 0.0106583, throughput 9.51882K wps
[Epoch 4 Batch 150/162] avg loss 0.0106647, throughput 9.38567K wps
Begin Testing...
[Epoch 4] train avg loss 0.0108324, dev acc 0.8589, dev avg loss 0.517315, throughput 9.4451K wps
Observed Improvement.
Begin Testing...
[Epoch 5 Batch 30/162] avg loss 0.0104338, throughput 9.59142K wps
[Epoch 5 Batch 60/162] avg loss 0.0101572, throughput 9.33178K wps
[Epoch 5 Batch 90/162] avg loss 0.0100831, throughput 9.36474K wps
[Epoch 5 Batch 120/162] avg loss 0.00980301, throughput 9.34837K wps
[Epoch 5 Batch 150/162] avg loss 0.00967527, throughput 9.27428K wps
Begin Testing...
[Epoch 5] train avg loss 0.00999851, dev acc 0.8589, dev avg loss 0.477792, throughput 9.38176K wps
Observed Improvement.
Begin Testing...
[Epoch 6 Batch 30/162] avg loss 0.00972869, throughput 9.52634K wps
[Epoch 6 Batch 60/162] avg loss 0.00934932, throughput 9.48655K wps
[Epoch 6 Batch 90/162] avg loss 0.00943156, throughput 9.30064K wps
[Epoch 6 Batch 120/162] avg loss 0.00896844, throughput 9.42494K wps
[Epoch 6 Batch 150/162] avg loss 0.00920985, throughput 9.39836K wps
Begin Testing...
[Epoch 6] train avg loss 0.00930666, dev acc 0.8567, dev avg loss 0.442584, throughput 9.42414K wps
[Epoch 7 Batch 30/162] avg loss 0.00885472, throughput 9.53572K wps
[Epoch 7 Batch 60/162] avg loss 0.0086763, throughput 9.32788K wps
[Epoch 7 Batch 90/162] avg loss 0.00870802, throughput 9.34875K wps
[Epoch 7 Batch 120/162] avg loss 0.00836684, throughput 9.42209K wps
[Epoch 7 Batch 150/162] avg loss 0.00843406, throughput 9.3475K wps
Begin Testing...
[Epoch 7] train avg loss 0.00859618, dev acc 0.8656, dev avg loss 0.412908, throughput 9.41978K wps
Observed Improvement.
Begin Testing...
[Epoch 8 Batch 30/162] avg loss 0.00806912, throughput 9.7301K wps
[Epoch 8 Batch 60/162] avg loss 0.00820234, throughput 9.43989K wps
[Epoch 8 Batch 90/162] avg loss 0.00805145, throughput 9.39106K wps
[Epoch 8 Batch 120/162] avg loss 0.0078793, throughput 9.52033K wps
[Epoch 8 Batch 150/162] avg loss 0.0080455, throughput 9.53167K wps
Begin Testing...
[Epoch 8] train avg loss 0.00806016, dev acc 0.8667, dev avg loss 0.3887, throughput 9.50221K wps
Observed Improvement.
Begin Testing...
[Epoch 9 Batch 30/162] avg loss 0.00750047, throughput 9.5567K wps
[Epoch 9 Batch 60/162] avg loss 0.00799271, throughput 9.44845K wps
[Epoch 9 Batch 90/162] avg loss 0.00784349, throughput 9.34739K wps
[Epoch 9 Batch 120/162] avg loss 0.00745752, throughput 9.56061K wps
[Epoch 9 Batch 150/162] avg loss 0.00766892, throughput 9.51101K wps
Begin Testing...
[Epoch 9] train avg loss 0.00765048, dev acc 0.8656, dev avg loss 0.369554, throughput 9.48571K wps
[Epoch 10 Batch 30/162] avg loss 0.00740377, throughput 9.75305K wps
[Epoch 10 Batch 60/162] avg loss 0.00729179, throughput 9.5592K wps
[Epoch 10 Batch 90/162] avg loss 0.00712125, throughput 9.60119K wps
[Epoch 10 Batch 120/162] avg loss 0.00734595, throughput 9.60966K wps
[Epoch 10 Batch 150/162] avg loss 0.00719343, throughput 9.61205K wps
Begin Testing...
[Epoch 10] train avg loss 0.00726185, dev acc 0.8667, dev avg loss 0.353704, throughput 9.61808K wps
Observed Improvement.
Begin Testing...
[Epoch 11 Batch 30/162] avg loss 0.00710261, throughput 9.70969K wps
[Epoch 11 Batch 60/162] avg loss 0.00693902, throughput 9.31352K wps
[Epoch 11 Batch 90/162] avg loss 0.00707425, throughput 9.47858K wps
[Epoch 11 Batch 120/162] avg loss 0.00649506, throughput 9.33234K wps
[Epoch 11 Batch 150/162] avg loss 0.00697072, throughput 9.34748K wps
Begin Testing...
[Epoch 11] train avg loss 0.00692122, dev acc 0.8733, dev avg loss 0.341066, throughput 9.4301K wps
Observed Improvement.
Begin Testing...
[Epoch 12 Batch 30/162] avg loss 0.00631853, throughput 9.55049K wps
[Epoch 12 Batch 60/162] avg loss 0.00663218, throughput 9.61023K wps
[Epoch 12 Batch 90/162] avg loss 0.00673604, throughput 9.55756K wps
[Epoch 12 Batch 120/162] avg loss 0.00693172, throughput 9.41589K wps
[Epoch 12 Batch 150/162] avg loss 0.00654904, throughput 9.38467K wps
Begin Testing...
[Epoch 12] train avg loss 0.00664445, dev acc 0.8767, dev avg loss 0.328612, throughput 9.50984K wps
Observed Improvement.
Begin Testing...
[Epoch 13 Batch 30/162] avg loss 0.00628318, throughput 9.61178K wps
[Epoch 13 Batch 60/162] avg loss 0.00640106, throughput 9.36653K wps
[Epoch 13 Batch 90/162] avg loss 0.00680906, throughput 9.3197K wps
[Epoch 13 Batch 120/162] avg loss 0.00651375, throughput 9.47155K wps
[Epoch 13 Batch 150/162] avg loss 0.00650468, throughput 9.58766K wps
Begin Testing...
[Epoch 13] train avg loss 0.00648943, dev acc 0.8767, dev avg loss 0.319403, throughput 9.47163K wps
Observed Improvement.
Begin Testing...
[Epoch 14 Batch 30/162] avg loss 0.00660427, throughput 9.84856K wps
[Epoch 14 Batch 60/162] avg loss 0.0063452, throughput 9.27139K wps
[Epoch 14 Batch 90/162] avg loss 0.00654229, throughput 9.25659K wps
[Epoch 14 Batch 120/162] avg loss 0.00607842, throughput 9.54302K wps
[Epoch 14 Batch 150/162] avg loss 0.00609628, throughput 9.4878K wps
Begin Testing...
[Epoch 14] train avg loss 0.00633399, dev acc 0.8800, dev avg loss 0.311135, throughput 9.45432K wps
Observed Improvement.
Begin Testing...
[Epoch 15 Batch 30/162] avg loss 0.00646639, throughput 9.46686K wps
[Epoch 15 Batch 60/162] avg loss 0.00598127, throughput 9.36905K wps
[Epoch 15 Batch 90/162] avg loss 0.00601278, throughput 9.31124K wps
[Epoch 15 Batch 120/162] avg loss 0.00617649, throughput 9.48315K wps
[Epoch 15 Batch 150/162] avg loss 0.00598403, throughput 9.56752K wps
Begin Testing...
[Epoch 15] train avg loss 0.00614973, dev acc 0.8844, dev avg loss 0.304547, throughput 9.44711K wps
Observed Improvement.
Begin Testing...
[Epoch 16 Batch 30/162] avg loss 0.00594706, throughput 9.56068K wps
[Epoch 16 Batch 60/162] avg loss 0.00607288, throughput 9.53689K wps
[Epoch 16 Batch 90/162] avg loss 0.00594977, throughput 9.31959K wps
[Epoch 16 Batch 120/162] avg loss 0.00607013, throughput 9.29825K wps
[Epoch 16 Batch 150/162] avg loss 0.00581879, throughput 9.3948K wps
Begin Testing...
[Epoch 16] train avg loss 0.00592857, dev acc 0.8867, dev avg loss 0.297664, throughput 9.43123K wps
Observed Improvement.
Begin Testing...
[Epoch 17 Batch 30/162] avg loss 0.00583465, throughput 9.48135K wps
[Epoch 17 Batch 60/162] avg loss 0.0059846, throughput 9.41098K wps
[Epoch 17 Batch 90/162] avg loss 0.00585893, throughput 9.31835K wps
[Epoch 17 Batch 120/162] avg loss 0.00563629, throughput 9.44213K wps
[Epoch 17 Batch 150/162] avg loss 0.00562804, throughput 9.52757K wps
Begin Testing...
[Epoch 17] train avg loss 0.00580811, dev acc 0.8900, dev avg loss 0.292328, throughput 9.45288K wps
Observed Improvement.
Begin Testing...
[Epoch 18 Batch 30/162] avg loss 0.00535752, throughput 9.79914K wps
[Epoch 18 Batch 60/162] avg loss 0.00574516, throughput 9.46184K wps
[Epoch 18 Batch 90/162] avg loss 0.00575284, throughput 9.28375K wps
[Epoch 18 Batch 120/162] avg loss 0.00559057, throughput 9.46167K wps
[Epoch 18 Batch 150/162] avg loss 0.00578278, throughput 9.47966K wps
Begin Testing...
[Epoch 18] train avg loss 0.00565703, dev acc 0.8922, dev avg loss 0.286797, throughput 9.48595K wps
Observed Improvement.
Begin Testing...
[Epoch 19 Batch 30/162] avg loss 0.00560813, throughput 9.4919K wps
[Epoch 19 Batch 60/162] avg loss 0.00540952, throughput 9.51449K wps
[Epoch 19 Batch 90/162] avg loss 0.00582129, throughput 9.29493K wps
[Epoch 19 Batch 120/162] avg loss 0.00516214, throughput 9.38459K wps
[Epoch 19 Batch 150/162] avg loss 0.00563269, throughput 9.19882K wps
Begin Testing...
[Epoch 19] train avg loss 0.00556424, dev acc 0.8889, dev avg loss 0.282806, throughput 9.3864K wps
[Epoch 20 Batch 30/162] avg loss 0.00564318, throughput 9.55158K wps
[Epoch 20 Batch 60/162] avg loss 0.00553172, throughput 9.53462K wps
[Epoch 20 Batch 90/162] avg loss 0.00538131, throughput 9.46273K wps
[Epoch 20 Batch 120/162] avg loss 0.00516098, throughput 9.49521K wps
[Epoch 20 Batch 150/162] avg loss 0.00542954, throughput 9.54365K wps
Begin Testing...
[Epoch 20] train avg loss 0.00541477, dev acc 0.8978, dev avg loss 0.277972, throughput 9.52291K wps
Observed Improvement.
Begin Testing...
[Epoch 21 Batch 30/162] avg loss 0.00523994, throughput 9.55739K wps
[Epoch 21 Batch 60/162] avg loss 0.00566302, throughput 9.35851K wps
[Epoch 21 Batch 90/162] avg loss 0.00514172, throughput 9.40032K wps
[Epoch 21 Batch 120/162] avg loss 0.00552506, throughput 9.26519K wps
[Epoch 21 Batch 150/162] avg loss 0.00506583, throughput 9.50668K wps
Begin Testing...
[Epoch 21] train avg loss 0.00530618, dev acc 0.8933, dev avg loss 0.273645, throughput 9.42566K wps
[Epoch 22 Batch 30/162] avg loss 0.00521912, throughput 9.6366K wps
[Epoch 22 Batch 60/162] avg loss 0.00529461, throughput 9.32882K wps
[Epoch 22 Batch 90/162] avg loss 0.00510741, throughput 9.33636K wps
[Epoch 22 Batch 120/162] avg loss 0.00494408, throughput 9.43341K wps
[Epoch 22 Batch 150/162] avg loss 0.005402, throughput 9.27279K wps
Begin Testing...
[Epoch 22] train avg loss 0.00522381, dev acc 0.9000, dev avg loss 0.27019, throughput 9.39647K wps
Observed Improvement.
Begin Testing...
[Epoch 23 Batch 30/162] avg loss 0.00495843, throughput 9.53871K wps
[Epoch 23 Batch 60/162] avg loss 0.00495238, throughput 9.34393K wps
[Epoch 23 Batch 90/162] avg loss 0.00528676, throughput 9.54701K wps
[Epoch 23 Batch 120/162] avg loss 0.00472634, throughput 9.34672K wps
[Epoch 23 Batch 150/162] avg loss 0.00524473, throughput 9.45264K wps
Begin Testing...
[Epoch 23] train avg loss 0.00504655, dev acc 0.9022, dev avg loss 0.266702, throughput 9.46174K wps
Observed Improvement.
Begin Testing...
[Epoch 24 Batch 30/162] avg loss 0.0049871, throughput 9.4778K wps
[Epoch 24 Batch 60/162] avg loss 0.00519308, throughput 9.48348K wps
[Epoch 24 Batch 90/162] avg loss 0.00511727, throughput 9.28619K wps
[Epoch 24 Batch 120/162] avg loss 0.00472395, throughput 9.49498K wps
[Epoch 24 Batch 150/162] avg loss 0.00479156, throughput 9.40056K wps
Begin Testing...
[Epoch 24] train avg loss 0.00497652, dev acc 0.9022, dev avg loss 0.26372, throughput 9.41525K wps
Observed Improvement.
Begin Testing...
[Epoch 25 Batch 30/162] avg loss 0.00483612, throughput 9.59675K wps
[Epoch 25 Batch 60/162] avg loss 0.00464176, throughput 9.41563K wps
[Epoch 25 Batch 90/162] avg loss 0.00511961, throughput 9.34795K wps
[Epoch 25 Batch 120/162] avg loss 0.0045783, throughput 9.38304K wps
[Epoch 25 Batch 150/162] avg loss 0.00507207, throughput 9.44556K wps
Begin Testing...
[Epoch 25] train avg loss 0.00488533, dev acc 0.9067, dev avg loss 0.260396, throughput 9.43253K wps
Observed Improvement.
Begin Testing...
[Epoch 26 Batch 30/162] avg loss 0.00478406, throughput 9.65418K wps
[Epoch 26 Batch 60/162] avg loss 0.00457549, throughput 9.26003K wps
[Epoch 26 Batch 90/162] avg loss 0.0048799, throughput 9.33453K wps
[Epoch 26 Batch 120/162] avg loss 0.00492129, throughput 9.56185K wps
[Epoch 26 Batch 150/162] avg loss 0.00504633, throughput 9.40066K wps
Begin Testing...
[Epoch 26] train avg loss 0.00484426, dev acc 0.9056, dev avg loss 0.257635, throughput 9.43176K wps
[Epoch 27 Batch 30/162] avg loss 0.00487778, throughput 9.67809K wps
[Epoch 27 Batch 60/162] avg loss 0.00476072, throughput 9.42554K wps
[Epoch 27 Batch 90/162] avg loss 0.00472078, throughput 9.40577K wps
[Epoch 27 Batch 120/162] avg loss 0.00500905, throughput 9.46239K wps
[Epoch 27 Batch 150/162] avg loss 0.00450428, throughput 9.62653K wps
Begin Testing...
[Epoch 27] train avg loss 0.00480198, dev acc 0.9056, dev avg loss 0.255742, throughput 9.51887K wps
[Epoch 28 Batch 30/162] avg loss 0.00452125, throughput 9.75428K wps
[Epoch 28 Batch 60/162] avg loss 0.00434574, throughput 9.65742K wps
[Epoch 28 Batch 90/162] avg loss 0.0044311, throughput 9.34693K wps
[Epoch 28 Batch 120/162] avg loss 0.00468256, throughput 9.37244K wps
[Epoch 28 Batch 150/162] avg loss 0.00527594, throughput 9.34153K wps
Begin Testing...
[Epoch 28] train avg loss 0.00466196, dev acc 0.9067, dev avg loss 0.252901, throughput 9.50613K wps
Observed Improvement.
Begin Testing...
[Epoch 29 Batch 30/162] avg loss 0.00477465, throughput 9.64574K wps
[Epoch 29 Batch 60/162] avg loss 0.00460124, throughput 9.43298K wps
[Epoch 29 Batch 90/162] avg loss 0.00468383, throughput 9.63657K wps
[Epoch 29 Batch 120/162] avg loss 0.0046104, throughput 9.44176K wps
[Epoch 29 Batch 150/162] avg loss 0.00476819, throughput 9.53812K wps
Begin Testing...
[Epoch 29] train avg loss 0.00464942, dev acc 0.9078, dev avg loss 0.251143, throughput 9.54012K wps
Observed Improvement.
Begin Testing...
[Epoch 30 Batch 30/162] avg loss 0.00436063, throughput 9.63484K wps
[Epoch 30 Batch 60/162] avg loss 0.00436115, throughput 9.37346K wps
[Epoch 30 Batch 90/162] avg loss 0.00460125, throughput 9.31886K wps
[Epoch 30 Batch 120/162] avg loss 0.004433, throughput 9.43734K wps
[Epoch 30 Batch 150/162] avg loss 0.00486167, throughput 9.39925K wps
Begin Testing...
[Epoch 30] train avg loss 0.0045216, dev acc 0.9056, dev avg loss 0.247955, throughput 9.42865K wps
[Epoch 31 Batch 30/162] avg loss 0.00460965, throughput 9.66807K wps
[Epoch 31 Batch 60/162] avg loss 0.00445438, throughput 9.48706K wps
[Epoch 31 Batch 90/162] avg loss 0.00451678, throughput 9.52231K wps
[Epoch 31 Batch 120/162] avg loss 0.00418388, throughput 9.56985K wps
[Epoch 31 Batch 150/162] avg loss 0.00454303, throughput 9.62875K wps
Begin Testing...
[Epoch 31] train avg loss 0.00443703, dev acc 0.9078, dev avg loss 0.247396, throughput 9.54812K wps
Observed Improvement.
Begin Testing...
[Epoch 32 Batch 30/162] avg loss 0.00420622, throughput 9.52443K wps
[Epoch 32 Batch 60/162] avg loss 0.00446417, throughput 9.40759K wps
[Epoch 32 Batch 90/162] avg loss 0.00415023, throughput 9.52854K wps
[Epoch 32 Batch 120/162] avg loss 0.00421479, throughput 9.44247K wps
[Epoch 32 Batch 150/162] avg loss 0.00463659, throughput 9.4787K wps
Begin Testing...
[Epoch 32] train avg loss 0.00435941, dev acc 0.9089, dev avg loss 0.244622, throughput 9.48057K wps
Observed Improvement.
Begin Testing...
[Epoch 33 Batch 30/162] avg loss 0.00425542, throughput 9.57699K wps
[Epoch 33 Batch 60/162] avg loss 0.00434641, throughput 9.31524K wps
[Epoch 33 Batch 90/162] avg loss 0.00426613, throughput 9.45086K wps
[Epoch 33 Batch 120/162] avg loss 0.00408631, throughput 9.33918K wps
[Epoch 33 Batch 150/162] avg loss 0.00442509, throughput 9.44528K wps
Begin Testing...
[Epoch 33] train avg loss 0.00425713, dev acc 0.9089, dev avg loss 0.242447, throughput 9.41047K wps
Observed Improvement.
Begin Testing...
[Epoch 34 Batch 30/162] avg loss 0.00406545, throughput 9.55801K wps
[Epoch 34 Batch 60/162] avg loss 0.00418443, throughput 9.42699K wps
[Epoch 34 Batch 90/162] avg loss 0.00417431, throughput 9.32451K wps
[Epoch 34 Batch 120/162] avg loss 0.00440464, throughput 9.41508K wps
[Epoch 34 Batch 150/162] avg loss 0.00441537, throughput 9.34383K wps
Begin Testing...
[Epoch 34] train avg loss 0.00423775, dev acc 0.9067, dev avg loss 0.240167, throughput 9.40727K wps
[Epoch 35 Batch 30/162] avg loss 0.00383354, throughput 9.66794K wps
[Epoch 35 Batch 60/162] avg loss 0.00416397, throughput 9.40851K wps
[Epoch 35 Batch 90/162] avg loss 0.00425117, throughput 9.48345K wps
[Epoch 35 Batch 120/162] avg loss 0.00414007, throughput 9.33598K wps
[Epoch 35 Batch 150/162] avg loss 0.00415459, throughput 9.37042K wps
Begin Testing...
[Epoch 35] train avg loss 0.00411082, dev acc 0.9089, dev avg loss 0.238537, throughput 9.46262K wps
Observed Improvement.
Begin Testing...
[Epoch 36 Batch 30/162] avg loss 0.00408504, throughput 9.7111K wps
[Epoch 36 Batch 60/162] avg loss 0.00435667, throughput 9.28503K wps
[Epoch 36 Batch 90/162] avg loss 0.00380566, throughput 9.47346K wps
[Epoch 36 Batch 120/162] avg loss 0.00420147, throughput 9.35228K wps
[Epoch 36 Batch 150/162] avg loss 0.00392749, throughput 9.2543K wps
Begin Testing...
[Epoch 36] train avg loss 0.00405599, dev acc 0.9100, dev avg loss 0.237322, throughput 9.41131K wps
Observed Improvement.
Begin Testing...
[Epoch 37 Batch 30/162] avg loss 0.0037669, throughput 9.66519K wps
[Epoch 37 Batch 60/162] avg loss 0.00398405, throughput 9.4164K wps
[Epoch 37 Batch 90/162] avg loss 0.00386321, throughput 9.33987K wps
[Epoch 37 Batch 120/162] avg loss 0.00408181, throughput 9.63412K wps
[Epoch 37 Batch 150/162] avg loss 0.00438054, throughput 9.40096K wps
Begin Testing...
[Epoch 37] train avg loss 0.00405731, dev acc 0.9111, dev avg loss 0.235672, throughput 9.50046K wps
Observed Improvement.
Begin Testing...
[Epoch 38 Batch 30/162] avg loss 0.00406346, throughput 9.497K wps
[Epoch 38 Batch 60/162] avg loss 0.00389415, throughput 9.53711K wps
[Epoch 38 Batch 90/162] avg loss 0.00383885, throughput 9.64241K wps
[Epoch 38 Batch 120/162] avg loss 0.00405133, throughput 9.53045K wps
[Epoch 38 Batch 150/162] avg loss 0.00368905, throughput 9.33284K wps
Begin Testing...
[Epoch 38] train avg loss 0.00391384, dev acc 0.9100, dev avg loss 0.234983, throughput 9.49323K wps
[Epoch 39 Batch 30/162] avg loss 0.00393512, throughput 9.86066K wps
[Epoch 39 Batch 60/162] avg loss 0.00360985, throughput 9.37953K wps
[Epoch 39 Batch 90/162] avg loss 0.00377205, throughput 9.3411K wps
[Epoch 39 Batch 120/162] avg loss 0.00412724, throughput 9.39593K wps
[Epoch 39 Batch 150/162] avg loss 0.00390451, throughput 9.48308K wps
Begin Testing...
[Epoch 39] train avg loss 0.00384671, dev acc 0.9100, dev avg loss 0.233743, throughput 9.48727K wps
[Epoch 40 Batch 30/162] avg loss 0.00388092, throughput 9.57072K wps
[Epoch 40 Batch 60/162] avg loss 0.00385099, throughput 9.32482K wps
[Epoch 40 Batch 90/162] avg loss 0.00385664, throughput 9.35352K wps
[Epoch 40 Batch 120/162] avg loss 0.00390976, throughput 9.44745K wps
[Epoch 40 Batch 150/162] avg loss 0.00373293, throughput 9.45809K wps
Begin Testing...
[Epoch 40] train avg loss 0.0038746, dev acc 0.9122, dev avg loss 0.231287, throughput 9.41868K wps
Observed Improvement.
Begin Testing...
[Epoch 41 Batch 30/162] avg loss 0.00426183, throughput 9.57025K wps
[Epoch 41 Batch 60/162] avg loss 0.00385923, throughput 9.58151K wps
[Epoch 41 Batch 90/162] avg loss 0.00379537, throughput 9.50705K wps
[Epoch 41 Batch 120/162] avg loss 0.00364216, throughput 9.44985K wps
[Epoch 41 Batch 150/162] avg loss 0.00356505, throughput 9.51729K wps
Begin Testing...
[Epoch 41] train avg loss 0.00380987, dev acc 0.9133, dev avg loss 0.229897, throughput 9.51478K wps
Observed Improvement.
Begin Testing...
[Epoch 42 Batch 30/162] avg loss 0.00369556, throughput 9.56096K wps
[Epoch 42 Batch 60/162] avg loss 0.00333394, throughput 9.59764K wps
[Epoch 42 Batch 90/162] avg loss 0.00374504, throughput 9.47066K wps
[Epoch 42 Batch 120/162] avg loss 0.00403983, throughput 9.41211K wps
[Epoch 42 Batch 150/162] avg loss 0.00387656, throughput 9.28724K wps
Begin Testing...
[Epoch 42] train avg loss 0.00373613, dev acc 0.9144, dev avg loss 0.228566, throughput 9.4504K wps
Observed Improvement.
Begin Testing...
[Epoch 43 Batch 30/162] avg loss 0.00315349, throughput 9.63995K wps
[Epoch 43 Batch 60/162] avg loss 0.00345343, throughput 9.39673K wps
[Epoch 43 Batch 90/162] avg loss 0.00361769, throughput 9.56593K wps
[Epoch 43 Batch 120/162] avg loss 0.00396842, throughput 9.41416K wps
[Epoch 43 Batch 150/162] avg loss 0.00362562, throughput 9.35257K wps
Begin Testing...
[Epoch 43] train avg loss 0.00360532, dev acc 0.9122, dev avg loss 0.227892, throughput 9.47263K wps
[Epoch 44 Batch 30/162] avg loss 0.00366727, throughput 9.65239K wps
[Epoch 44 Batch 60/162] avg loss 0.00348665, throughput 9.49594K wps
[Epoch 44 Batch 90/162] avg loss 0.00342233, throughput 9.52823K wps
[Epoch 44 Batch 120/162] avg loss 0.00356927, throughput 9.61137K wps
[Epoch 44 Batch 150/162] avg loss 0.00368072, throughput 9.53701K wps
Begin Testing...
[Epoch 44] train avg loss 0.00358512, dev acc 0.9156, dev avg loss 0.226473, throughput 9.56279K wps
Observed Improvement.
Begin Testing...
[Epoch 45 Batch 30/162] avg loss 0.00362152, throughput 9.69376K wps
[Epoch 45 Batch 60/162] avg loss 0.00356702, throughput 9.42077K wps
[Epoch 45 Batch 90/162] avg loss 0.00374377, throughput 9.27785K wps
[Epoch 45 Batch 120/162] avg loss 0.00335657, throughput 9.61865K wps
[Epoch 45 Batch 150/162] avg loss 0.00377884, throughput 9.26487K wps
Begin Testing...
[Epoch 45] train avg loss 0.00357415, dev acc 0.9156, dev avg loss 0.225669, throughput 9.4462K wps
Observed Improvement.
Begin Testing...
[Epoch 46 Batch 30/162] avg loss 0.00334084, throughput 9.53136K wps
[Epoch 46 Batch 60/162] avg loss 0.00357465, throughput 9.34733K wps
[Epoch 46 Batch 90/162] avg loss 0.00348136, throughput 9.50936K wps
[Epoch 46 Batch 120/162] avg loss 0.00372269, throughput 9.31877K wps
[Epoch 46 Batch 150/162] avg loss 0.00332941, throughput 9.41405K wps
Begin Testing...
[Epoch 46] train avg loss 0.00345628, dev acc 0.9156, dev avg loss 0.224458, throughput 9.41732K wps
Observed Improvement.
Begin Testing...
[Epoch 47 Batch 30/162] avg loss 0.00352207, throughput 9.50129K wps
[Epoch 47 Batch 60/162] avg loss 0.00329303, throughput 9.53134K wps
[Epoch 47 Batch 90/162] avg loss 0.00344771, throughput 9.31125K wps
[Epoch 47 Batch 120/162] avg loss 0.00342248, throughput 9.4594K wps
[Epoch 47 Batch 150/162] avg loss 0.00320117, throughput 9.43958K wps
Begin Testing...
[Epoch 47] train avg loss 0.00346401, dev acc 0.9178, dev avg loss 0.225761, throughput 9.45105K wps
Observed Improvement.
Begin Testing...
[Epoch 48 Batch 30/162] avg loss 0.00354354, throughput 9.51538K wps
[Epoch 48 Batch 60/162] avg loss 0.00341876, throughput 9.49926K wps
[Epoch 48 Batch 90/162] avg loss 0.00324349, throughput 9.49185K wps
[Epoch 48 Batch 120/162] avg loss 0.00359092, throughput 9.37376K wps
[Epoch 48 Batch 150/162] avg loss 0.00345554, throughput 9.30192K wps
Begin Testing...
[Epoch 48] train avg loss 0.00345582, dev acc 0.9167, dev avg loss 0.223216, throughput 9.43185K wps
[Epoch 49 Batch 30/162] avg loss 0.00350184, throughput 9.74943K wps
[Epoch 49 Batch 60/162] avg loss 0.00332023, throughput 9.46507K wps
[Epoch 49 Batch 90/162] avg loss 0.0034021, throughput 9.51085K wps
[Epoch 49 Batch 120/162] avg loss 0.0035685, throughput 9.5042K wps
[Epoch 49 Batch 150/162] avg loss 0.00315494, throughput 9.49529K wps
Begin Testing...
[Epoch 49] train avg loss 0.00337652, dev acc 0.9144, dev avg loss 0.222067, throughput 9.5286K wps
[Epoch 50 Batch 30/162] avg loss 0.00308501, throughput 9.59108K wps
[Epoch 50 Batch 60/162] avg loss 0.00344647, throughput 9.54811K wps
[Epoch 50 Batch 90/162] avg loss 0.00327992, throughput 9.3721K wps
[Epoch 50 Batch 120/162] avg loss 0.00337598, throughput 9.50468K wps
[Epoch 50 Batch 150/162] avg loss 0.00354227, throughput 9.45515K wps
Begin Testing...
[Epoch 50] train avg loss 0.00333527, dev acc 0.9156, dev avg loss 0.221123, throughput 9.48983K wps
[Epoch 51 Batch 30/162] avg loss 0.00315521, throughput 9.56668K wps
[Epoch 51 Batch 60/162] avg loss 0.00327059, throughput 9.35536K wps
[Epoch 51 Batch 90/162] avg loss 0.00342324, throughput 9.46106K wps
[Epoch 51 Batch 120/162] avg loss 0.00327808, throughput 9.33101K wps
[Epoch 51 Batch 150/162] avg loss 0.00334738, throughput 9.32542K wps
Begin Testing...
[Epoch 51] train avg loss 0.00332145, dev acc 0.9167, dev avg loss 0.220512, throughput 9.41257K wps
[Epoch 52 Batch 30/162] avg loss 0.00319075, throughput 9.59843K wps
[Epoch 52 Batch 60/162] avg loss 0.00295505, throughput 9.47073K wps
[Epoch 52 Batch 90/162] avg loss 0.00344457, throughput 9.43088K wps
[Epoch 52 Batch 120/162] avg loss 0.00305463, throughput 9.42518K wps
[Epoch 52 Batch 150/162] avg loss 0.00341097, throughput 9.35764K wps
Begin Testing...
[Epoch 52] train avg loss 0.00321385, dev acc 0.9133, dev avg loss 0.220126, throughput 9.44812K wps
[Epoch 53 Batch 30/162] avg loss 0.00297898, throughput 9.6734K wps
[Epoch 53 Batch 60/162] avg loss 0.00316718, throughput 9.4229K wps
[Epoch 53 Batch 90/162] avg loss 0.00328465, throughput 9.50635K wps
[Epoch 53 Batch 120/162] avg loss 0.00339371, throughput 9.47033K wps
[Epoch 53 Batch 150/162] avg loss 0.00299492, throughput 9.62237K wps
Begin Testing...
[Epoch 53] train avg loss 0.00317905, dev acc 0.9189, dev avg loss 0.219618, throughput 9.52147K wps
Observed Improvement.
Begin Testing...
[Epoch 54 Batch 30/162] avg loss 0.00304264, throughput 9.55647K wps
[Epoch 54 Batch 60/162] avg loss 0.00335492, throughput 9.38461K wps
[Epoch 54 Batch 90/162] avg loss 0.00343523, throughput 9.47602K wps
[Epoch 54 Batch 120/162] avg loss 0.0029398, throughput 9.45281K wps
[Epoch 54 Batch 150/162] avg loss 0.00305618, throughput 9.3994K wps
Begin Testing...
[Epoch 54] train avg loss 0.00317582, dev acc 0.9189, dev avg loss 0.219422, throughput 9.44758K wps
Observed Improvement.
Begin Testing...
[Epoch 55 Batch 30/162] avg loss 0.00297708, throughput 9.65878K wps
[Epoch 55 Batch 60/162] avg loss 0.00320414, throughput 9.43412K wps
[Epoch 55 Batch 90/162] avg loss 0.00285005, throughput 9.48062K wps
[Epoch 55 Batch 120/162] avg loss 0.00316178, throughput 9.43701K wps
[Epoch 55 Batch 150/162] avg loss 0.0034096, throughput 9.28857K wps
Begin Testing...
[Epoch 55] train avg loss 0.00308901, dev acc 0.9189, dev avg loss 0.217401, throughput 9.46269K wps
Observed Improvement.
Begin Testing...
[Epoch 56 Batch 30/162] avg loss 0.00312059, throughput 9.75452K wps
[Epoch 56 Batch 60/162] avg loss 0.00305829, throughput 9.41905K wps
[Epoch 56 Batch 90/162] avg loss 0.00266312, throughput 9.44953K wps
[Epoch 56 Batch 120/162] avg loss 0.00298384, throughput 9.31158K wps
[Epoch 56 Batch 150/162] avg loss 0.00309729, throughput 9.32276K wps
Begin Testing...
[Epoch 56] train avg loss 0.00299268, dev acc 0.9200, dev avg loss 0.216728, throughput 9.43511K wps
Observed Improvement.
Begin Testing...
[Epoch 57 Batch 30/162] avg loss 0.00269389, throughput 9.59015K wps
[Epoch 57 Batch 60/162] avg loss 0.00285155, throughput 9.33546K wps
[Epoch 57 Batch 90/162] avg loss 0.00311853, throughput 9.36452K wps
[Epoch 57 Batch 120/162] avg loss 0.00305225, throughput 9.49742K wps
[Epoch 57 Batch 150/162] avg loss 0.00297802, throughput 9.38308K wps
Begin Testing...
[Epoch 57] train avg loss 0.00295189, dev acc 0.9167, dev avg loss 0.215674, throughput 9.42676K wps
[Epoch 58 Batch 30/162] avg loss 0.00282621, throughput 9.49604K wps
[Epoch 58 Batch 60/162] avg loss 0.0032306, throughput 9.22635K wps
[Epoch 58 Batch 90/162] avg loss 0.00292797, throughput 9.31087K wps
[Epoch 58 Batch 120/162] avg loss 0.0027582, throughput 9.42997K wps
[Epoch 58 Batch 150/162] avg loss 0.00284288, throughput 9.4841K wps
Begin Testing...
[Epoch 58] train avg loss 0.00294232, dev acc 0.9211, dev avg loss 0.216652, throughput 9.37845K wps
Observed Improvement.
Begin Testing...
[Epoch 59 Batch 30/162] avg loss 0.00277569, throughput 9.69999K wps
[Epoch 59 Batch 60/162] avg loss 0.00308478, throughput 9.47894K wps
[Epoch 59 Batch 90/162] avg loss 0.00268679, throughput 9.44815K wps
[Epoch 59 Batch 120/162] avg loss 0.00309849, throughput 9.20161K wps
[Epoch 59 Batch 150/162] avg loss 0.00293884, throughput 9.31193K wps
Begin Testing...
[Epoch 59] train avg loss 0.00293761, dev acc 0.9222, dev avg loss 0.216959, throughput 9.41598K wps
Observed Improvement.
Begin Testing...
[Epoch 60 Batch 30/162] avg loss 0.00309496, throughput 9.42156K wps
[Epoch 60 Batch 60/162] avg loss 0.00298088, throughput 9.57968K wps
[Epoch 60 Batch 90/162] avg loss 0.00284989, throughput 9.2436K wps
[Epoch 60 Batch 120/162] avg loss 0.00254365, throughput 9.49312K wps
[Epoch 60 Batch 150/162] avg loss 0.00275094, throughput 9.48603K wps
Begin Testing...
[Epoch 60] train avg loss 0.00281464, dev acc 0.9233, dev avg loss 0.214249, throughput 9.4533K wps
Observed Improvement.
Begin Testing...
[Epoch 61 Batch 30/162] avg loss 0.00265659, throughput 9.60677K wps
[Epoch 61 Batch 60/162] avg loss 0.00304692, throughput 9.29835K wps
[Epoch 61 Batch 90/162] avg loss 0.00283466, throughput 9.35066K wps
[Epoch 61 Batch 120/162] avg loss 0.0026997, throughput 9.35587K wps
[Epoch 61 Batch 150/162] avg loss 0.00295456, throughput 9.40125K wps
Begin Testing...
[Epoch 61] train avg loss 0.00282413, dev acc 0.9244, dev avg loss 0.213766, throughput 9.4197K wps
Observed Improvement.
Begin Testing...
[Epoch 62 Batch 30/162] avg loss 0.00286028, throughput 9.57983K wps
[Epoch 62 Batch 60/162] avg loss 0.00267104, throughput 9.46429K wps
[Epoch 62 Batch 90/162] avg loss 0.00289336, throughput 9.47721K wps
[Epoch 62 Batch 120/162] avg loss 0.0027304, throughput 9.59819K wps
[Epoch 62 Batch 150/162] avg loss 0.00282833, throughput 9.42406K wps
Begin Testing...
[Epoch 62] train avg loss 0.00279192, dev acc 0.9178, dev avg loss 0.212517, throughput 9.49895K wps
[Epoch 63 Batch 30/162] avg loss 0.00254448, throughput 9.67151K wps
[Epoch 63 Batch 60/162] avg loss 0.00277218, throughput 9.45498K wps
[Epoch 63 Batch 90/162] avg loss 0.00273721, throughput 9.57808K wps
[Epoch 63 Batch 120/162] avg loss 0.00270732, throughput 9.45007K wps
[Epoch 63 Batch 150/162] avg loss 0.00274281, throughput 9.59422K wps
Begin Testing...
[Epoch 63] train avg loss 0.00270058, dev acc 0.9133, dev avg loss 0.212324, throughput 9.55274K wps
[Epoch 64 Batch 30/162] avg loss 0.00248418, throughput 9.65464K wps
[Epoch 64 Batch 60/162] avg loss 0.00286207, throughput 9.39722K wps
[Epoch 64 Batch 90/162] avg loss 0.00282825, throughput 9.41728K wps
[Epoch 64 Batch 120/162] avg loss 0.00278738, throughput 9.43599K wps
[Epoch 64 Batch 150/162] avg loss 0.00271865, throughput 9.46671K wps
Begin Testing...
[Epoch 64] train avg loss 0.00271333, dev acc 0.9156, dev avg loss 0.211601, throughput 9.45424K wps
[Epoch 65 Batch 30/162] avg loss 0.00248159, throughput 9.63847K wps
[Epoch 65 Batch 60/162] avg loss 0.00290095, throughput 9.33732K wps
[Epoch 65 Batch 90/162] avg loss 0.0024169, throughput 9.59213K wps
[Epoch 65 Batch 120/162] avg loss 0.00292065, throughput 9.39201K wps
[Epoch 65 Batch 150/162] avg loss 0.00232786, throughput 9.58522K wps
Begin Testing...
[Epoch 65] train avg loss 0.00260124, dev acc 0.9167, dev avg loss 0.211127, throughput 9.49524K wps
[Epoch 66 Batch 30/162] avg loss 0.00267339, throughput 9.73666K wps
[Epoch 66 Batch 60/162] avg loss 0.00263615, throughput 9.57565K wps
[Epoch 66 Batch 90/162] avg loss 0.00256139, throughput 9.55463K wps
[Epoch 66 Batch 120/162] avg loss 0.00249245, throughput 9.53232K wps
[Epoch 66 Batch 150/162] avg loss 0.00266076, throughput 9.46135K wps
Begin Testing...
[Epoch 66] train avg loss 0.0025875, dev acc 0.9222, dev avg loss 0.213142, throughput 9.54993K wps
[Epoch 67 Batch 30/162] avg loss 0.00272001, throughput 9.51908K wps
[Epoch 67 Batch 60/162] avg loss 0.0022673, throughput 9.31795K wps
[Epoch 67 Batch 90/162] avg loss 0.0026388, throughput 9.56817K wps
[Epoch 67 Batch 120/162] avg loss 0.00271669, throughput 9.45039K wps
[Epoch 67 Batch 150/162] avg loss 0.00270597, throughput 9.43765K wps
Begin Testing...
[Epoch 67] train avg loss 0.00262128, dev acc 0.9200, dev avg loss 0.210452, throughput 9.44748K wps
[Epoch 68 Batch 30/162] avg loss 0.00251807, throughput 9.66559K wps
[Epoch 68 Batch 60/162] avg loss 0.00270323, throughput 9.46845K wps
[Epoch 68 Batch 90/162] avg loss 0.00262588, throughput 9.29565K wps
[Epoch 68 Batch 120/162] avg loss 0.00255562, throughput 9.51342K wps
[Epoch 68 Batch 150/162] avg loss 0.0024727, throughput 9.35565K wps
Begin Testing...
[Epoch 68] train avg loss 0.00254703, dev acc 0.9211, dev avg loss 0.210071, throughput 9.45892K wps
[Epoch 69 Batch 30/162] avg loss 0.00251997, throughput 9.53591K wps
[Epoch 69 Batch 60/162] avg loss 0.00258998, throughput 9.3957K wps
[Epoch 69 Batch 90/162] avg loss 0.00232841, throughput 9.42548K wps
[Epoch 69 Batch 120/162] avg loss 0.00248578, throughput 9.37496K wps
[Epoch 69 Batch 150/162] avg loss 0.00250468, throughput 9.37617K wps
Begin Testing...
[Epoch 69] train avg loss 0.00248453, dev acc 0.9233, dev avg loss 0.211074, throughput 9.42949K wps
[Epoch 70 Batch 30/162] avg loss 0.00239474, throughput 9.50195K wps
[Epoch 70 Batch 60/162] avg loss 0.00217355, throughput 9.42338K wps
[Epoch 70 Batch 90/162] avg loss 0.0027454, throughput 9.44293K wps
[Epoch 70 Batch 120/162] avg loss 0.00238378, throughput 9.44654K wps
[Epoch 70 Batch 150/162] avg loss 0.00235287, throughput 9.35185K wps
Begin Testing...
[Epoch 70] train avg loss 0.00243013, dev acc 0.9200, dev avg loss 0.20896, throughput 9.45598K wps
[Epoch 71 Batch 30/162] avg loss 0.00255012, throughput 9.69867K wps
[Epoch 71 Batch 60/162] avg loss 0.00249323, throughput 9.4884K wps
[Epoch 71 Batch 90/162] avg loss 0.00245921, throughput 9.43063K wps
[Epoch 71 Batch 120/162] avg loss 0.00259206, throughput 9.27647K wps
[Epoch 71 Batch 150/162] avg loss 0.00238921, throughput 9.37805K wps
Begin Testing...
[Epoch 71] train avg loss 0.00246777, dev acc 0.9211, dev avg loss 0.209015, throughput 9.44245K wps
[Epoch 72 Batch 30/162] avg loss 0.00231949, throughput 9.64685K wps
[Epoch 72 Batch 60/162] avg loss 0.00244343, throughput 9.43977K wps
[Epoch 72 Batch 90/162] avg loss 0.00256313, throughput 9.39777K wps
[Epoch 72 Batch 120/162] avg loss 0.00237135, throughput 9.3312K wps
[Epoch 72 Batch 150/162] avg loss 0.00242063, throughput 9.46528K wps
Begin Testing...
[Epoch 72] train avg loss 0.00242478, dev acc 0.9244, dev avg loss 0.209655, throughput 9.4614K wps
Observed Improvement.
Begin Testing...
[Epoch 73 Batch 30/162] avg loss 0.0021945, throughput 9.63781K wps
[Epoch 73 Batch 60/162] avg loss 0.0022784, throughput 9.27314K wps
[Epoch 73 Batch 90/162] avg loss 0.00243337, throughput 9.50823K wps
[Epoch 73 Batch 120/162] avg loss 0.0023927, throughput 9.59945K wps
[Epoch 73 Batch 150/162] avg loss 0.00245518, throughput 9.54293K wps
Begin Testing...
[Epoch 73] train avg loss 0.00235562, dev acc 0.9211, dev avg loss 0.208078, throughput 9.4933K wps
[Epoch 74 Batch 30/162] avg loss 0.00242713, throughput 9.75882K wps
[Epoch 74 Batch 60/162] avg loss 0.00235984, throughput 9.54676K wps
[Epoch 74 Batch 90/162] avg loss 0.00232318, throughput 9.3858K wps
[Epoch 74 Batch 120/162] avg loss 0.00206459, throughput 9.37696K wps
[Epoch 74 Batch 150/162] avg loss 0.00229806, throughput 9.27311K wps
Begin Testing...
[Epoch 74] train avg loss 0.00228787, dev acc 0.9211, dev avg loss 0.207808, throughput 9.46267K wps
[Epoch 75 Batch 30/162] avg loss 0.00207518, throughput 9.707K wps
[Epoch 75 Batch 60/162] avg loss 0.00212959, throughput 9.45883K wps
[Epoch 75 Batch 90/162] avg loss 0.0023397, throughput 9.42135K wps
[Epoch 75 Batch 120/162] avg loss 0.00211707, throughput 9.38586K wps
[Epoch 75 Batch 150/162] avg loss 0.00236358, throughput 9.38092K wps
Begin Testing...
[Epoch 75] train avg loss 0.00221594, dev acc 0.9211, dev avg loss 0.207528, throughput 9.4754K wps
[Epoch 76 Batch 30/162] avg loss 0.00226175, throughput 9.60442K wps
[Epoch 76 Batch 60/162] avg loss 0.00217826, throughput 9.35046K wps
[Epoch 76 Batch 90/162] avg loss 0.00221076, throughput 9.32535K wps
[Epoch 76 Batch 120/162] avg loss 0.00242047, throughput 9.45956K wps
[Epoch 76 Batch 150/162] avg loss 0.00231846, throughput 9.42842K wps
Begin Testing...
[Epoch 76] train avg loss 0.00226195, dev acc 0.9211, dev avg loss 0.207098, throughput 9.42967K wps
[Epoch 77 Batch 30/162] avg loss 0.00231982, throughput 9.62737K wps
[Epoch 77 Batch 60/162] avg loss 0.00229532, throughput 9.43391K wps
[Epoch 77 Batch 90/162] avg loss 0.0024013, throughput 9.61799K wps
[Epoch 77 Batch 120/162] avg loss 0.00232639, throughput 9.36088K wps
[Epoch 77 Batch 150/162] avg loss 0.00232753, throughput 9.58087K wps
Begin Testing...
[Epoch 77] train avg loss 0.00230577, dev acc 0.9211, dev avg loss 0.206782, throughput 9.52954K wps
[Epoch 78 Batch 30/162] avg loss 0.00211787, throughput 9.66874K wps
[Epoch 78 Batch 60/162] avg loss 0.00224852, throughput 9.49414K wps
[Epoch 78 Batch 90/162] avg loss 0.00209251, throughput 9.59282K wps
[Epoch 78 Batch 120/162] avg loss 0.00195535, throughput 9.56179K wps
[Epoch 78 Batch 150/162] avg loss 0.00268989, throughput 9.4231K wps
Begin Testing...
[Epoch 78] train avg loss 0.00221114, dev acc 0.9222, dev avg loss 0.206682, throughput 9.54082K wps
[Epoch 79 Batch 30/162] avg loss 0.00210097, throughput 9.57887K wps
[Epoch 79 Batch 60/162] avg loss 0.00233019, throughput 9.39033K wps
[Epoch 79 Batch 90/162] avg loss 0.0020848, throughput 9.29084K wps
[Epoch 79 Batch 120/162] avg loss 0.00229396, throughput 9.43846K wps
[Epoch 79 Batch 150/162] avg loss 0.00230666, throughput 9.52034K wps
Begin Testing...
[Epoch 79] train avg loss 0.00219974, dev acc 0.9211, dev avg loss 0.206102, throughput 9.43797K wps
[Epoch 80 Batch 30/162] avg loss 0.00187974, throughput 9.44764K wps
[Epoch 80 Batch 60/162] avg loss 0.00219116, throughput 9.52109K wps
[Epoch 80 Batch 90/162] avg loss 0.00223691, throughput 9.39427K wps
[Epoch 80 Batch 120/162] avg loss 0.00202455, throughput 9.36286K wps
[Epoch 80 Batch 150/162] avg loss 0.00223439, throughput 9.21074K wps
Begin Testing...
[Epoch 80] train avg loss 0.00212718, dev acc 0.9211, dev avg loss 0.206076, throughput 9.38124K wps
[Epoch 81 Batch 30/162] avg loss 0.00205889, throughput 9.57821K wps
[Epoch 81 Batch 60/162] avg loss 0.00213062, throughput 9.56151K wps
[Epoch 81 Batch 90/162] avg loss 0.00222759, throughput 9.59452K wps
[Epoch 81 Batch 120/162] avg loss 0.00211883, throughput 9.41213K wps
[Epoch 81 Batch 150/162] avg loss 0.00209074, throughput 9.5143K wps
Begin Testing...
[Epoch 81] train avg loss 0.00210659, dev acc 0.9233, dev avg loss 0.206509, throughput 9.51398K wps
[Epoch 82 Batch 30/162] avg loss 0.00229056, throughput 9.49822K wps
[Epoch 82 Batch 60/162] avg loss 0.00203222, throughput 9.49515K wps
[Epoch 82 Batch 90/162] avg loss 0.00209236, throughput 9.42488K wps
[Epoch 82 Batch 120/162] avg loss 0.00207488, throughput 9.46791K wps
[Epoch 82 Batch 150/162] avg loss 0.0020838, throughput 9.47313K wps
Begin Testing...
[Epoch 82] train avg loss 0.00209507, dev acc 0.9211, dev avg loss 0.205499, throughput 9.46425K wps
[Epoch 83 Batch 30/162] avg loss 0.00196087, throughput 9.6164K wps
[Epoch 83 Batch 60/162] avg loss 0.00213653, throughput 9.28213K wps
[Epoch 83 Batch 90/162] avg loss 0.00199187, throughput 9.39584K wps
[Epoch 83 Batch 120/162] avg loss 0.00215475, throughput 9.35673K wps
[Epoch 83 Batch 150/162] avg loss 0.00187753, throughput 9.46498K wps
Begin Testing...
[Epoch 83] train avg loss 0.00206155, dev acc 0.9211, dev avg loss 0.205562, throughput 9.42071K wps
[Epoch 84 Batch 30/162] avg loss 0.00212762, throughput 9.58606K wps
[Epoch 84 Batch 60/162] avg loss 0.00201353, throughput 9.39946K wps
[Epoch 84 Batch 90/162] avg loss 0.002025, throughput 9.30209K wps
[Epoch 84 Batch 120/162] avg loss 0.00191519, throughput 9.37319K wps
[Epoch 84 Batch 150/162] avg loss 0.00194526, throughput 9.35304K wps
Begin Testing...
[Epoch 84] train avg loss 0.00199982, dev acc 0.9222, dev avg loss 0.204985, throughput 9.40264K wps
[Epoch 85 Batch 30/162] avg loss 0.00209207, throughput 9.50075K wps
[Epoch 85 Batch 60/162] avg loss 0.00199798, throughput 9.61042K wps
[Epoch 85 Batch 90/162] avg loss 0.00218432, throughput 9.43156K wps
[Epoch 85 Batch 120/162] avg loss 0.0018818, throughput 9.49036K wps
[Epoch 85 Batch 150/162] avg loss 0.00189478, throughput 9.26383K wps
Begin Testing...
[Epoch 85] train avg loss 0.00199071, dev acc 0.9222, dev avg loss 0.205979, throughput 9.45861K wps
[Epoch 86 Batch 30/162] avg loss 0.00204731, throughput 9.686K wps
[Epoch 86 Batch 60/162] avg loss 0.00208833, throughput 9.47065K wps
[Epoch 86 Batch 90/162] avg loss 0.00194284, throughput 9.35596K wps
[Epoch 86 Batch 120/162] avg loss 0.00185654, throughput 9.49427K wps
[Epoch 86 Batch 150/162] avg loss 0.0019407, throughput 9.493K wps
Begin Testing...
[Epoch 86] train avg loss 0.00196651, dev acc 0.9233, dev avg loss 0.204131, throughput 9.49683K wps
[Epoch 87 Batch 30/162] avg loss 0.00195459, throughput 9.54267K wps
[Epoch 87 Batch 60/162] avg loss 0.00198784, throughput 9.33784K wps
[Epoch 87 Batch 90/162] avg loss 0.00208091, throughput 9.34242K wps
[Epoch 87 Batch 120/162] avg loss 0.00184575, throughput 9.43135K wps
[Epoch 87 Batch 150/162] avg loss 0.00205951, throughput 9.5619K wps
Begin Testing...
[Epoch 87] train avg loss 0.00196039, dev acc 0.9244, dev avg loss 0.210054, throughput 9.4538K wps
Observed Improvement.
Begin Testing...
[Epoch 88 Batch 30/162] avg loss 0.00194644, throughput 9.5449K wps
[Epoch 88 Batch 60/162] avg loss 0.00159468, throughput 9.30649K wps
[Epoch 88 Batch 90/162] avg loss 0.00199066, throughput 9.3253K wps
[Epoch 88 Batch 120/162] avg loss 0.00203797, throughput 9.54555K wps
[Epoch 88 Batch 150/162] avg loss 0.00206979, throughput 9.39294K wps
Begin Testing...
[Epoch 88] train avg loss 0.00193645, dev acc 0.9211, dev avg loss 0.204577, throughput 9.42811K wps
[Epoch 89 Batch 30/162] avg loss 0.00196916, throughput 9.59751K wps
[Epoch 89 Batch 60/162] avg loss 0.00192561, throughput 9.32896K wps
[Epoch 89 Batch 90/162] avg loss 0.00187884, throughput 9.33669K wps
[Epoch 89 Batch 120/162] avg loss 0.00169219, throughput 9.38627K wps
[Epoch 89 Batch 150/162] avg loss 0.00199551, throughput 9.47098K wps
Begin Testing...
[Epoch 89] train avg loss 0.00189481, dev acc 0.9233, dev avg loss 0.205053, throughput 9.42794K wps
[Epoch 90 Batch 30/162] avg loss 0.00161669, throughput 9.74899K wps
[Epoch 90 Batch 60/162] avg loss 0.00177536, throughput 9.48583K wps
[Epoch 90 Batch 90/162] avg loss 0.00181728, throughput 9.35256K wps
[Epoch 90 Batch 120/162] avg loss 0.00187514, throughput 9.52305K wps
[Epoch 90 Batch 150/162] avg loss 0.00208418, throughput 9.39069K wps
Begin Testing...
[Epoch 90] train avg loss 0.00183349, dev acc 0.9256, dev avg loss 0.207747, throughput 9.50371K wps
Observed Improvement.
Begin Testing...
[Epoch 91 Batch 30/162] avg loss 0.00157796, throughput 9.64219K wps
[Epoch 91 Batch 60/162] avg loss 0.00220249, throughput 9.22907K wps
[Epoch 91 Batch 90/162] avg loss 0.00163015, throughput 9.47745K wps
[Epoch 91 Batch 120/162] avg loss 0.00188249, throughput 9.24638K wps
[Epoch 91 Batch 150/162] avg loss 0.00183925, throughput 9.508K wps
Begin Testing...
[Epoch 91] train avg loss 0.00183595, dev acc 0.9233, dev avg loss 0.203474, throughput 9.3991K wps
[Epoch 92 Batch 30/162] avg loss 0.00193043, throughput 9.53729K wps
[Epoch 92 Batch 60/162] avg loss 0.00190057, throughput 9.46335K wps
[Epoch 92 Batch 90/162] avg loss 0.00175855, throughput 9.48714K wps
[Epoch 92 Batch 120/162] avg loss 0.00169433, throughput 9.39459K wps
[Epoch 92 Batch 150/162] avg loss 0.0018763, throughput 9.53678K wps
Begin Testing...
[Epoch 92] train avg loss 0.00183129, dev acc 0.9233, dev avg loss 0.203252, throughput 9.47458K wps
[Epoch 93 Batch 30/162] avg loss 0.00192441, throughput 9.73058K wps
[Epoch 93 Batch 60/162] avg loss 0.00175015, throughput 9.53051K wps
[Epoch 93 Batch 90/162] avg loss 0.00179377, throughput 9.5084K wps
[Epoch 93 Batch 120/162] avg loss 0.00181925, throughput 9.56534K wps
[Epoch 93 Batch 150/162] avg loss 0.00162491, throughput 9.36392K wps
Begin Testing...
[Epoch 93] train avg loss 0.00178648, dev acc 0.9222, dev avg loss 0.203213, throughput 9.53883K wps
[Epoch 94 Batch 30/162] avg loss 0.00163098, throughput 9.48267K wps
[Epoch 94 Batch 60/162] avg loss 0.0018116, throughput 9.41889K wps
[Epoch 94 Batch 90/162] avg loss 0.00181541, throughput 9.37826K wps
[Epoch 94 Batch 120/162] avg loss 0.00171804, throughput 9.32408K wps
[Epoch 94 Batch 150/162] avg loss 0.0019551, throughput 9.38121K wps
Begin Testing...
[Epoch 94] train avg loss 0.00178261, dev acc 0.9233, dev avg loss 0.203386, throughput 9.39766K wps
[Epoch 95 Batch 30/162] avg loss 0.00163902, throughput 9.67681K wps
[Epoch 95 Batch 60/162] avg loss 0.00193656, throughput 9.5868K wps
[Epoch 95 Batch 90/162] avg loss 0.00163474, throughput 9.64256K wps
[Epoch 95 Batch 120/162] avg loss 0.00174039, throughput 9.27915K wps
[Epoch 95 Batch 150/162] avg loss 0.00192247, throughput 9.45406K wps
Begin Testing...
[Epoch 95] train avg loss 0.00176633, dev acc 0.9222, dev avg loss 0.202631, throughput 9.53186K wps
[Epoch 96 Batch 30/162] avg loss 0.00160915, throughput 9.78138K wps
[Epoch 96 Batch 60/162] avg loss 0.00198227, throughput 9.5085K wps
[Epoch 96 Batch 90/162] avg loss 0.00176379, throughput 9.54236K wps
[Epoch 96 Batch 120/162] avg loss 0.00173519, throughput 9.69673K wps
[Epoch 96 Batch 150/162] avg loss 0.00160583, throughput 9.45488K wps
Begin Testing...
[Epoch 96] train avg loss 0.00175066, dev acc 0.9244, dev avg loss 0.203479, throughput 9.56775K wps
[Epoch 97 Batch 30/162] avg loss 0.00166704, throughput 9.5232K wps
[Epoch 97 Batch 60/162] avg loss 0.00176722, throughput 9.62551K wps
[Epoch 97 Batch 90/162] avg loss 0.0017331, throughput 9.48113K wps
[Epoch 97 Batch 120/162] avg loss 0.00162017, throughput 9.52966K wps
[Epoch 97 Batch 150/162] avg loss 0.00172142, throughput 9.36582K wps
Begin Testing...
[Epoch 97] train avg loss 0.00173055, dev acc 0.9233, dev avg loss 0.202337, throughput 9.49002K wps
[Epoch 98 Batch 30/162] avg loss 0.00162093, throughput 9.51604K wps
[Epoch 98 Batch 60/162] avg loss 0.00155901, throughput 9.46407K wps
[Epoch 98 Batch 90/162] avg loss 0.00155725, throughput 9.58779K wps
[Epoch 98 Batch 120/162] avg loss 0.00183394, throughput 9.42528K wps
[Epoch 98 Batch 150/162] avg loss 0.00194257, throughput 9.43682K wps
Begin Testing...
[Epoch 98] train avg loss 0.00171083, dev acc 0.9233, dev avg loss 0.205171, throughput 9.47128K wps
[Epoch 99 Batch 30/162] avg loss 0.0015791, throughput 9.58426K wps
[Epoch 99 Batch 60/162] avg loss 0.00161487, throughput 9.57849K wps
[Epoch 99 Batch 90/162] avg loss 0.00158925, throughput 9.43597K wps
[Epoch 99 Batch 120/162] avg loss 0.0017338, throughput 9.41892K wps
[Epoch 99 Batch 150/162] avg loss 0.00171889, throughput 9.68755K wps
Begin Testing...
[Epoch 99] train avg loss 0.00165643, dev acc 0.9244, dev avg loss 0.203012, throughput 9.54241K wps
[Epoch 100 Batch 30/162] avg loss 0.00154153, throughput 9.52784K wps
[Epoch 100 Batch 60/162] avg loss 0.00164454, throughput 9.30129K wps
[Epoch 100 Batch 90/162] avg loss 0.00148648, throughput 9.56667K wps
[Epoch 100 Batch 120/162] avg loss 0.00148182, throughput 9.4091K wps
[Epoch 100 Batch 150/162] avg loss 0.00182493, throughput 9.34583K wps
Begin Testing...
[Epoch 100] train avg loss 0.00159165, dev acc 0.9256, dev avg loss 0.203292, throughput 9.43368K wps
Observed Improvement.
Begin Testing...
[Epoch 101 Batch 30/162] avg loss 0.00173673, throughput 9.69293K wps
[Epoch 101 Batch 60/162] avg loss 0.00160724, throughput 9.42899K wps
[Epoch 101 Batch 90/162] avg loss 0.00152343, throughput 9.32942K wps
[Epoch 101 Batch 120/162] avg loss 0.00147202, throughput 9.55056K wps
[Epoch 101 Batch 150/162] avg loss 0.00161212, throughput 9.43802K wps
Begin Testing...
[Epoch 101] train avg loss 0.00159293, dev acc 0.9244, dev avg loss 0.202404, throughput 9.487K wps
[Epoch 102 Batch 30/162] avg loss 0.0015004, throughput 9.70803K wps
[Epoch 102 Batch 60/162] avg loss 0.00148417, throughput 9.356K wps
[Epoch 102 Batch 90/162] avg loss 0.00158121, throughput 9.34482K wps
[Epoch 102 Batch 120/162] avg loss 0.00163136, throughput 9.49188K wps
[Epoch 102 Batch 150/162] avg loss 0.00163222, throughput 9.19789K wps
Begin Testing...
[Epoch 102] train avg loss 0.00156816, dev acc 0.9189, dev avg loss 0.201942, throughput 9.42656K wps
[Epoch 103 Batch 30/162] avg loss 0.00158239, throughput 9.68196K wps
[Epoch 103 Batch 60/162] avg loss 0.00186189, throughput 9.41597K wps
[Epoch 103 Batch 90/162] avg loss 0.00139846, throughput 9.4326K wps
[Epoch 103 Batch 120/162] avg loss 0.00140018, throughput 9.38977K wps
[Epoch 103 Batch 150/162] avg loss 0.00156127, throughput 9.37384K wps
Begin Testing...
[Epoch 103] train avg loss 0.0015708, dev acc 0.9244, dev avg loss 0.203126, throughput 9.46034K wps
[Epoch 104 Batch 30/162] avg loss 0.00165652, throughput 9.49102K wps
[Epoch 104 Batch 60/162] avg loss 0.00146861, throughput 9.36673K wps
[Epoch 104 Batch 90/162] avg loss 0.00148715, throughput 9.41715K wps
[Epoch 104 Batch 120/162] avg loss 0.00182564, throughput 9.37813K wps
[Epoch 104 Batch 150/162] avg loss 0.00149063, throughput 9.47642K wps
Begin Testing...
[Epoch 104] train avg loss 0.00157329, dev acc 0.9233, dev avg loss 0.202671, throughput 9.40877K wps
[Epoch 105 Batch 30/162] avg loss 0.0014982, throughput 9.54739K wps
[Epoch 105 Batch 60/162] avg loss 0.00165306, throughput 9.31869K wps
[Epoch 105 Batch 90/162] avg loss 0.00160842, throughput 9.3981K wps
[Epoch 105 Batch 120/162] avg loss 0.00152189, throughput 9.4413K wps
[Epoch 105 Batch 150/162] avg loss 0.00147934, throughput 9.52417K wps
Begin Testing...
[Epoch 105] train avg loss 0.00154117, dev acc 0.9244, dev avg loss 0.202726, throughput 9.44722K wps
[Epoch 106 Batch 30/162] avg loss 0.00148933, throughput 9.67491K wps
[Epoch 106 Batch 60/162] avg loss 0.00131232, throughput 9.4745K wps
[Epoch 106 Batch 90/162] avg loss 0.00158677, throughput 9.32856K wps
[Epoch 106 Batch 120/162] avg loss 0.00166374, throughput 9.5213K wps
[Epoch 106 Batch 150/162] avg loss 0.00156726, throughput 9.36798K wps
Begin Testing...
[Epoch 106] train avg loss 0.00152566, dev acc 0.9244, dev avg loss 0.202523, throughput 9.46511K wps
[Epoch 107 Batch 30/162] avg loss 0.00159739, throughput 9.62798K wps
[Epoch 107 Batch 60/162] avg loss 0.00141866, throughput 9.45446K wps
[Epoch 107 Batch 90/162] avg loss 0.00128639, throughput 9.58404K wps
[Epoch 107 Batch 120/162] avg loss 0.00146877, throughput 9.41635K wps
[Epoch 107 Batch 150/162] avg loss 0.0016545, throughput 9.39155K wps
Begin Testing...
[Epoch 107] train avg loss 0.0014948, dev acc 0.9233, dev avg loss 0.202025, throughput 9.49951K wps
[Epoch 108 Batch 30/162] avg loss 0.00157514, throughput 9.7641K wps
[Epoch 108 Batch 60/162] avg loss 0.00139654, throughput 9.46406K wps
[Epoch 108 Batch 90/162] avg loss 0.00145591, throughput 9.42733K wps
[Epoch 108 Batch 120/162] avg loss 0.00155112, throughput 9.60231K wps
[Epoch 108 Batch 150/162] avg loss 0.00143859, throughput 9.53069K wps
Begin Testing...
[Epoch 108] train avg loss 0.00149942, dev acc 0.9222, dev avg loss 0.201292, throughput 9.55174K wps
[Epoch 109 Batch 30/162] avg loss 0.00136831, throughput 9.58585K wps
[Epoch 109 Batch 60/162] avg loss 0.00152803, throughput 9.4934K wps
[Epoch 109 Batch 90/162] avg loss 0.00144156, throughput 9.3304K wps
[Epoch 109 Batch 120/162] avg loss 0.00147141, throughput 9.37241K wps
[Epoch 109 Batch 150/162] avg loss 0.00151994, throughput 9.38876K wps
Begin Testing...
[Epoch 109] train avg loss 0.00145506, dev acc 0.9256, dev avg loss 0.203265, throughput 9.45359K wps
Observed Improvement.
Begin Testing...
[Epoch 110 Batch 30/162] avg loss 0.00142826, throughput 9.50384K wps
[Epoch 110 Batch 60/162] avg loss 0.00130633, throughput 9.30507K wps
[Epoch 110 Batch 90/162] avg loss 0.00147079, throughput 9.55272K wps
[Epoch 110 Batch 120/162] avg loss 0.00134407, throughput 9.41137K wps
[Epoch 110 Batch 150/162] avg loss 0.00144229, throughput 9.56275K wps
Begin Testing...
[Epoch 110] train avg loss 0.00141053, dev acc 0.9244, dev avg loss 0.202747, throughput 9.47814K wps
[Epoch 111 Batch 30/162] avg loss 0.00138646, throughput 9.73205K wps
[Epoch 111 Batch 60/162] avg loss 0.0012837, throughput 9.4877K wps
[Epoch 111 Batch 90/162] avg loss 0.00148902, throughput 9.3959K wps
[Epoch 111 Batch 120/162] avg loss 0.00139464, throughput 9.46351K wps
[Epoch 111 Batch 150/162] avg loss 0.00159772, throughput 9.40222K wps
Begin Testing...
[Epoch 111] train avg loss 0.00142948, dev acc 0.9256, dev avg loss 0.202043, throughput 9.4957K wps
Observed Improvement.
Begin Testing...
[Epoch 112 Batch 30/162] avg loss 0.00136284, throughput 9.49932K wps
[Epoch 112 Batch 60/162] avg loss 0.00138415, throughput 9.33251K wps
[Epoch 112 Batch 90/162] avg loss 0.00137007, throughput 9.47857K wps
[Epoch 112 Batch 120/162] avg loss 0.00116675, throughput 9.51485K wps
[Epoch 112 Batch 150/162] avg loss 0.00148085, throughput 9.68629K wps
Begin Testing...
[Epoch 112] train avg loss 0.00135181, dev acc 0.9233, dev avg loss 0.201636, throughput 9.50442K wps
[Epoch 113 Batch 30/162] avg loss 0.00138991, throughput 9.5549K wps
[Epoch 113 Batch 60/162] avg loss 0.00150429, throughput 9.52891K wps
[Epoch 113 Batch 90/162] avg loss 0.00142094, throughput 9.39357K wps
[Epoch 113 Batch 120/162] avg loss 0.00134333, throughput 9.56007K wps
[Epoch 113 Batch 150/162] avg loss 0.00140256, throughput 9.47744K wps
Begin Testing...
[Epoch 113] train avg loss 0.00142483, dev acc 0.9233, dev avg loss 0.202052, throughput 9.48656K wps
[Epoch 114 Batch 30/162] avg loss 0.00134613, throughput 9.53696K wps
[Epoch 114 Batch 60/162] avg loss 0.0014622, throughput 9.39273K wps
[Epoch 114 Batch 90/162] avg loss 0.00154707, throughput 9.52565K wps
[Epoch 114 Batch 120/162] avg loss 0.00134085, throughput 9.47565K wps
[Epoch 114 Batch 150/162] avg loss 0.00133178, throughput 9.49749K wps
Begin Testing...
[Epoch 114] train avg loss 0.00141222, dev acc 0.9233, dev avg loss 0.201715, throughput 9.48438K wps
[Epoch 115 Batch 30/162] avg loss 0.00139111, throughput 9.50251K wps
[Epoch 115 Batch 60/162] avg loss 0.00132505, throughput 9.39712K wps
[Epoch 115 Batch 90/162] avg loss 0.00133145, throughput 9.29362K wps
[Epoch 115 Batch 120/162] avg loss 0.00127533, throughput 9.40789K wps
[Epoch 115 Batch 150/162] avg loss 0.0013472, throughput 9.44696K wps
Begin Testing...
[Epoch 115] train avg loss 0.00135158, dev acc 0.9211, dev avg loss 0.201567, throughput 9.39874K wps
[Epoch 116 Batch 30/162] avg loss 0.00144725, throughput 9.39229K wps
[Epoch 116 Batch 60/162] avg loss 0.00124801, throughput 9.43908K wps
[Epoch 116 Batch 90/162] avg loss 0.00132381, throughput 9.27777K wps
[Epoch 116 Batch 120/162] avg loss 0.00133681, throughput 9.47822K wps
[Epoch 116 Batch 150/162] avg loss 0.00140991, throughput 9.60068K wps
Begin Testing...
[Epoch 116] train avg loss 0.00134321, dev acc 0.9233, dev avg loss 0.201816, throughput 9.42134K wps
[Epoch 117 Batch 30/162] avg loss 0.0013316, throughput 9.57048K wps
[Epoch 117 Batch 60/162] avg loss 0.00117027, throughput 9.4322K wps
[Epoch 117 Batch 90/162] avg loss 0.00135528, throughput 9.52696K wps
[Epoch 117 Batch 120/162] avg loss 0.00136536, throughput 9.63291K wps
[Epoch 117 Batch 150/162] avg loss 0.00136781, throughput 9.31009K wps
Begin Testing...
[Epoch 117] train avg loss 0.00131376, dev acc 0.9233, dev avg loss 0.201148, throughput 9.48992K wps
[Epoch 118 Batch 30/162] avg loss 0.00137772, throughput 9.76145K wps
[Epoch 118 Batch 60/162] avg loss 0.00136151, throughput 9.53857K wps
[Epoch 118 Batch 90/162] avg loss 0.00131223, throughput 9.46297K wps
[Epoch 118 Batch 120/162] avg loss 0.00133381, throughput 9.28842K wps
[Epoch 118 Batch 150/162] avg loss 0.00130572, throughput 9.46586K wps
Begin Testing...
[Epoch 118] train avg loss 0.00133745, dev acc 0.9256, dev avg loss 0.203307, throughput 9.48172K wps
Observed Improvement.
Begin Testing...
[Epoch 119 Batch 30/162] avg loss 0.00133455, throughput 9.65096K wps
[Epoch 119 Batch 60/162] avg loss 0.00124589, throughput 9.43646K wps
[Epoch 119 Batch 90/162] avg loss 0.00120769, throughput 9.47165K wps
[Epoch 119 Batch 120/162] avg loss 0.00119017, throughput 9.34112K wps
[Epoch 119 Batch 150/162] avg loss 0.00126192, throughput 9.59037K wps
Begin Testing...
[Epoch 119] train avg loss 0.0012815, dev acc 0.9244, dev avg loss 0.20177, throughput 9.48061K wps
[Epoch 120 Batch 30/162] avg loss 0.0011659, throughput 9.4866K wps
[Epoch 120 Batch 60/162] avg loss 0.00119495, throughput 9.45884K wps
[Epoch 120 Batch 90/162] avg loss 0.00129187, throughput 9.3668K wps
[Epoch 120 Batch 120/162] avg loss 0.00117628, throughput 9.53073K wps
[Epoch 120 Batch 150/162] avg loss 0.00123714, throughput 9.37875K wps
Begin Testing...
[Epoch 120] train avg loss 0.00123315, dev acc 0.9222, dev avg loss 0.201217, throughput 9.44329K wps
[Epoch 121 Batch 30/162] avg loss 0.00120378, throughput 9.422K wps
[Epoch 121 Batch 60/162] avg loss 0.00132358, throughput 9.48897K wps
[Epoch 121 Batch 90/162] avg loss 0.00126254, throughput 9.5235K wps
[Epoch 121 Batch 120/162] avg loss 0.00120156, throughput 9.4996K wps
[Epoch 121 Batch 150/162] avg loss 0.00122317, throughput 9.4154K wps
Begin Testing...
[Epoch 121] train avg loss 0.00124695, dev acc 0.9244, dev avg loss 0.201591, throughput 9.44885K wps
[Epoch 122 Batch 30/162] avg loss 0.00130813, throughput 9.48197K wps
[Epoch 122 Batch 60/162] avg loss 0.00121134, throughput 9.6225K wps
[Epoch 122 Batch 90/162] avg loss 0.00122764, throughput 9.3918K wps
[Epoch 122 Batch 120/162] avg loss 0.00114518, throughput 9.43468K wps
[Epoch 122 Batch 150/162] avg loss 0.00124312, throughput 9.39861K wps
Begin Testing...
[Epoch 122] train avg loss 0.00122577, dev acc 0.9244, dev avg loss 0.202091, throughput 9.44506K wps
[Epoch 123 Batch 30/162] avg loss 0.0011434, throughput 9.57233K wps
[Epoch 123 Batch 60/162] avg loss 0.0013607, throughput 9.53198K wps
[Epoch 123 Batch 90/162] avg loss 0.00111576, throughput 9.43893K wps
[Epoch 123 Batch 120/162] avg loss 0.00116078, throughput 9.3721K wps
[Epoch 123 Batch 150/162] avg loss 0.0012121, throughput 9.3446K wps
Begin Testing...
[Epoch 123] train avg loss 0.00119516, dev acc 0.9256, dev avg loss 0.203904, throughput 9.44307K wps
Observed Improvement.
Begin Testing...
[Epoch 124 Batch 30/162] avg loss 0.00129984, throughput 9.77199K wps
[Epoch 124 Batch 60/162] avg loss 0.00122427, throughput 9.57315K wps
[Epoch 124 Batch 90/162] avg loss 0.00117892, throughput 9.41831K wps
[Epoch 124 Batch 120/162] avg loss 0.00118133, throughput 9.45582K wps
[Epoch 124 Batch 150/162] avg loss 0.00129336, throughput 9.28679K wps
Begin Testing...
[Epoch 124] train avg loss 0.00124152, dev acc 0.9256, dev avg loss 0.202672, throughput 9.49257K wps
Observed Improvement.
Begin Testing...
[Epoch 125 Batch 30/162] avg loss 0.00114131, throughput 9.56867K wps
[Epoch 125 Batch 60/162] avg loss 0.00106929, throughput 9.44898K wps
[Epoch 125 Batch 90/162] avg loss 0.00119182, throughput 9.31955K wps
[Epoch 125 Batch 120/162] avg loss 0.00126974, throughput 9.28261K wps
[Epoch 125 Batch 150/162] avg loss 0.001226, throughput 9.41216K wps
Begin Testing...
[Epoch 125] train avg loss 0.00117885, dev acc 0.9233, dev avg loss 0.201348, throughput 9.38788K wps
[Epoch 126 Batch 30/162] avg loss 0.00118871, throughput 9.54571K wps
[Epoch 126 Batch 60/162] avg loss 0.00127944, throughput 9.46064K wps
[Epoch 126 Batch 90/162] avg loss 0.00109242, throughput 9.39131K wps
[Epoch 126 Batch 120/162] avg loss 0.00103809, throughput 9.63535K wps
[Epoch 126 Batch 150/162] avg loss 0.00128093, throughput 9.39749K wps
Begin Testing...
[Epoch 126] train avg loss 0.00117723, dev acc 0.9267, dev avg loss 0.202524, throughput 9.47511K wps
Observed Improvement.
Begin Testing...
[Epoch 127 Batch 30/162] avg loss 0.00111277, throughput 9.4141K wps
[Epoch 127 Batch 60/162] avg loss 0.00111203, throughput 9.36078K wps
[Epoch 127 Batch 90/162] avg loss 0.00112989, throughput 9.40037K wps
[Epoch 127 Batch 120/162] avg loss 0.00110085, throughput 9.43931K wps
[Epoch 127 Batch 150/162] avg loss 0.00118058, throughput 9.37973K wps
Begin Testing...
[Epoch 127] train avg loss 0.00112331, dev acc 0.9278, dev avg loss 0.202582, throughput 9.41643K wps
Observed Improvement.
Begin Testing...
[Epoch 128 Batch 30/162] avg loss 0.00121207, throughput 9.55457K wps
[Epoch 128 Batch 60/162] avg loss 0.00114587, throughput 9.43051K wps
[Epoch 128 Batch 90/162] avg loss 0.00109304, throughput 9.39541K wps
[Epoch 128 Batch 120/162] avg loss 0.00107088, throughput 9.36972K wps
[Epoch 128 Batch 150/162] avg loss 0.00108113, throughput 9.49094K wps
Begin Testing...
[Epoch 128] train avg loss 0.00112527, dev acc 0.9256, dev avg loss 0.202324, throughput 9.4518K wps
[Epoch 129 Batch 30/162] avg loss 0.00112237, throughput 9.46627K wps
[Epoch 129 Batch 60/162] avg loss 0.00111399, throughput 9.56862K wps
[Epoch 129 Batch 90/162] avg loss 0.00105073, throughput 9.28154K wps
[Epoch 129 Batch 120/162] avg loss 0.00105877, throughput 9.43297K wps
[Epoch 129 Batch 150/162] avg loss 0.00125758, throughput 9.34436K wps
Begin Testing...
[Epoch 129] train avg loss 0.00111369, dev acc 0.9267, dev avg loss 0.203502, throughput 9.40944K wps
[Epoch 130 Batch 30/162] avg loss 0.00131624, throughput 9.48061K wps
[Epoch 130 Batch 60/162] avg loss 0.00096734, throughput 9.56546K wps
[Epoch 130 Batch 90/162] avg loss 0.00120519, throughput 9.32858K wps
[Epoch 130 Batch 120/162] avg loss 0.00115919, throughput 9.53085K wps
[Epoch 130 Batch 150/162] avg loss 0.00115578, throughput 9.2901K wps
Begin Testing...
[Epoch 130] train avg loss 0.00115063, dev acc 0.9267, dev avg loss 0.202923, throughput 9.43502K wps
[Epoch 131 Batch 30/162] avg loss 0.00108966, throughput 9.49829K wps
[Epoch 131 Batch 60/162] avg loss 0.00114699, throughput 9.6667K wps
[Epoch 131 Batch 90/162] avg loss 0.00119001, throughput 9.46293K wps
[Epoch 131 Batch 120/162] avg loss 0.00101985, throughput 9.23708K wps
[Epoch 131 Batch 150/162] avg loss 0.00118284, throughput 9.54844K wps
Begin Testing...
[Epoch 131] train avg loss 0.00111725, dev acc 0.9278, dev avg loss 0.203388, throughput 9.49293K wps
Observed Improvement.
Begin Testing...
[Epoch 132 Batch 30/162] avg loss 0.00106158, throughput 9.54645K wps
[Epoch 132 Batch 60/162] avg loss 0.00114195, throughput 9.45928K wps
[Epoch 132 Batch 90/162] avg loss 0.00121273, throughput 9.38548K wps
[Epoch 132 Batch 120/162] avg loss 0.00108378, throughput 9.54265K wps
[Epoch 132 Batch 150/162] avg loss 0.00101242, throughput 9.58481K wps
Begin Testing...
[Epoch 132] train avg loss 0.00111066, dev acc 0.9267, dev avg loss 0.203608, throughput 9.49698K wps
[Epoch 133 Batch 30/162] avg loss 0.00113497, throughput 9.60586K wps
[Epoch 133 Batch 60/162] avg loss 0.00104964, throughput 9.36288K wps
[Epoch 133 Batch 90/162] avg loss 0.00114399, throughput 9.45453K wps
[Epoch 133 Batch 120/162] avg loss 0.00105395, throughput 9.28536K wps
[Epoch 133 Batch 150/162] avg loss 0.00108168, throughput 9.4808K wps
Begin Testing...
[Epoch 133] train avg loss 0.00110277, dev acc 0.9267, dev avg loss 0.203768, throughput 9.42154K wps
[Epoch 134 Batch 30/162] avg loss 0.00108438, throughput 9.60764K wps
[Epoch 134 Batch 60/162] avg loss 0.00107826, throughput 9.42851K wps
[Epoch 134 Batch 90/162] avg loss 0.00109744, throughput 9.31158K wps
[Epoch 134 Batch 120/162] avg loss 0.00101538, throughput 9.68436K wps
[Epoch 134 Batch 150/162] avg loss 0.00106937, throughput 9.57327K wps
Begin Testing...
[Epoch 134] train avg loss 0.00106756, dev acc 0.9256, dev avg loss 0.207279, throughput 9.51761K wps
[Epoch 135 Batch 30/162] avg loss 0.00109209, throughput 9.45768K wps
[Epoch 135 Batch 60/162] avg loss 0.00107971, throughput 9.37839K wps
[Epoch 135 Batch 90/162] avg loss 0.00108532, throughput 9.49587K wps
[Epoch 135 Batch 120/162] avg loss 0.00103952, throughput 9.43799K wps
[Epoch 135 Batch 150/162] avg loss 0.0010297, throughput 9.29128K wps
Begin Testing...
[Epoch 135] train avg loss 0.00105985, dev acc 0.9256, dev avg loss 0.20469, throughput 9.42619K wps
[Epoch 136 Batch 30/162] avg loss 0.00111074, throughput 9.67735K wps
[Epoch 136 Batch 60/162] avg loss 0.00106132, throughput 9.38566K wps
[Epoch 136 Batch 90/162] avg loss 0.00108201, throughput 9.44294K wps
[Epoch 136 Batch 120/162] avg loss 0.00102899, throughput 9.41082K wps
[Epoch 136 Batch 150/162] avg loss 0.00112478, throughput 9.50992K wps
Begin Testing...
[Epoch 136] train avg loss 0.00107307, dev acc 0.9256, dev avg loss 0.203269, throughput 9.46732K wps
[Epoch 137 Batch 30/162] avg loss 0.00105559, throughput 9.61426K wps
[Epoch 137 Batch 60/162] avg loss 0.00103249, throughput 9.38535K wps
[Epoch 137 Batch 90/162] avg loss 0.00102809, throughput 9.27646K wps
[Epoch 137 Batch 120/162] avg loss 0.00102299, throughput 9.44331K wps
[Epoch 137 Batch 150/162] avg loss 0.00116073, throughput 9.50685K wps
Begin Testing...
[Epoch 137] train avg loss 0.00106216, dev acc 0.9267, dev avg loss 0.203235, throughput 9.45722K wps
[Epoch 138 Batch 30/162] avg loss 0.000917798, throughput 9.47688K wps
[Epoch 138 Batch 60/162] avg loss 0.00102939, throughput 9.43418K wps
[Epoch 138 Batch 90/162] avg loss 0.000908528, throughput 9.46926K wps
[Epoch 138 Batch 120/162] avg loss 0.00107867, throughput 9.55795K wps
[Epoch 138 Batch 150/162] avg loss 0.00107744, throughput 9.40711K wps
Begin Testing...
[Epoch 138] train avg loss 0.000991635, dev acc 0.9233, dev avg loss 0.202862, throughput 9.46168K wps
[Epoch 139 Batch 30/162] avg loss 0.00107028, throughput 9.62678K wps
[Epoch 139 Batch 60/162] avg loss 0.000926172, throughput 9.57414K wps
[Epoch 139 Batch 90/162] avg loss 0.0010532, throughput 9.57593K wps
[Epoch 139 Batch 120/162] avg loss 0.00109862, throughput 9.64634K wps
[Epoch 139 Batch 150/162] avg loss 0.00103609, throughput 9.36899K wps
Begin Testing...
[Epoch 139] train avg loss 0.00103818, dev acc 0.9256, dev avg loss 0.202762, throughput 9.54452K wps
[Epoch 140 Batch 30/162] avg loss 0.000930853, throughput 9.57739K wps
[Epoch 140 Batch 60/162] avg loss 0.00102872, throughput 9.47128K wps
[Epoch 140 Batch 90/162] avg loss 0.000921021, throughput 9.36923K wps
[Epoch 140 Batch 120/162] avg loss 0.00108652, throughput 9.42406K wps
[Epoch 140 Batch 150/162] avg loss 0.000996188, throughput 9.30833K wps
Begin Testing...
[Epoch 140] train avg loss 0.000997271, dev acc 0.9233, dev avg loss 0.202411, throughput 9.43887K wps
[Epoch 141 Batch 30/162] avg loss 0.000964566, throughput 9.8883K wps
[Epoch 141 Batch 60/162] avg loss 0.000880184, throughput 9.46809K wps
[Epoch 141 Batch 90/162] avg loss 0.0010303, throughput 9.39743K wps
[Epoch 141 Batch 120/162] avg loss 0.001165, throughput 9.38798K wps
[Epoch 141 Batch 150/162] avg loss 0.000927791, throughput 9.45876K wps
Begin Testing...
[Epoch 141] train avg loss 0.000989959, dev acc 0.9244, dev avg loss 0.202422, throughput 9.5146K wps
[Epoch 142 Batch 30/162] avg loss 0.000906651, throughput 9.82023K wps
[Epoch 142 Batch 60/162] avg loss 0.000983949, throughput 9.2961K wps
[Epoch 142 Batch 90/162] avg loss 0.00102882, throughput 9.47855K wps
[Epoch 142 Batch 120/162] avg loss 0.000933412, throughput 9.41882K wps
[Epoch 142 Batch 150/162] avg loss 0.000993972, throughput 9.278K wps
Begin Testing...
[Epoch 142] train avg loss 0.000982133, dev acc 0.9222, dev avg loss 0.202601, throughput 9.47232K wps
[Epoch 143 Batch 30/162] avg loss 0.00120479, throughput 9.57817K wps
[Epoch 143 Batch 60/162] avg loss 0.00103543, throughput 9.4905K wps
[Epoch 143 Batch 90/162] avg loss 0.000899953, throughput 9.36722K wps
[Epoch 143 Batch 120/162] avg loss 0.000882736, throughput 9.49204K wps
[Epoch 143 Batch 150/162] avg loss 0.00100171, throughput 9.36629K wps
Begin Testing...
[Epoch 143] train avg loss 0.00101484, dev acc 0.9278, dev avg loss 0.202786, throughput 9.47225K wps
Observed Improvement.
Begin Testing...
[Epoch 144 Batch 30/162] avg loss 0.000891639, throughput 9.46972K wps
[Epoch 144 Batch 60/162] avg loss 0.00109701, throughput 9.42032K wps
[Epoch 144 Batch 90/162] avg loss 0.000895984, throughput 9.31537K wps
[Epoch 144 Batch 120/162] avg loss 0.00101531, throughput 9.48617K wps
[Epoch 144 Batch 150/162] avg loss 0.000897554, throughput 9.31783K wps
Begin Testing...
[Epoch 144] train avg loss 0.000961219, dev acc 0.9256, dev avg loss 0.203646, throughput 9.40505K wps
[Epoch 145 Batch 30/162] avg loss 0.00100365, throughput 9.44684K wps
[Epoch 145 Batch 60/162] avg loss 0.000897284, throughput 9.44523K wps
[Epoch 145 Batch 90/162] avg loss 0.000963555, throughput 9.38626K wps
[Epoch 145 Batch 120/162] avg loss 0.000928011, throughput 9.38018K wps
[Epoch 145 Batch 150/162] avg loss 0.000982328, throughput 9.5432K wps
Begin Testing...
[Epoch 145] train avg loss 0.000944619, dev acc 0.9233, dev avg loss 0.202809, throughput 9.42609K wps
[Epoch 146 Batch 30/162] avg loss 0.000926932, throughput 9.64478K wps
[Epoch 146 Batch 60/162] avg loss 0.000983522, throughput 9.23487K wps
[Epoch 146 Batch 90/162] avg loss 0.000948354, throughput 9.5534K wps
[Epoch 146 Batch 120/162] avg loss 0.000880022, throughput 9.34004K wps
[Epoch 146 Batch 150/162] avg loss 0.000996027, throughput 9.51583K wps
Begin Testing...
[Epoch 146] train avg loss 0.000952216, dev acc 0.9267, dev avg loss 0.203745, throughput 9.44281K wps
[Epoch 147 Batch 30/162] avg loss 0.00094269, throughput 9.67563K wps
[Epoch 147 Batch 60/162] avg loss 0.000800981, throughput 9.5112K wps
[Epoch 147 Batch 90/162] avg loss 0.000927402, throughput 9.41384K wps
[Epoch 147 Batch 120/162] avg loss 0.000951098, throughput 9.50736K wps
[Epoch 147 Batch 150/162] avg loss 0.000892814, throughput 9.39901K wps
Begin Testing...
[Epoch 147] train avg loss 0.000907699, dev acc 0.9267, dev avg loss 0.203591, throughput 9.49957K wps
[Epoch 148 Batch 30/162] avg loss 0.000945546, throughput 8.55626K wps
[Epoch 148 Batch 60/162] avg loss 0.000902696, throughput 9.37089K wps
[Epoch 148 Batch 90/162] avg loss 0.000834359, throughput 9.51153K wps
[Epoch 148 Batch 120/162] avg loss 0.000934841, throughput 9.57265K wps
[Epoch 148 Batch 150/162] avg loss 0.000959502, throughput 9.47085K wps
Begin Testing...
[Epoch 148] train avg loss 0.000904833, dev acc 0.9289, dev avg loss 0.204209, throughput 9.29154K wps
Observed Improvement.
Begin Testing...
[Epoch 149 Batch 30/162] avg loss 0.000891047, throughput 9.79885K wps
[Epoch 149 Batch 60/162] avg loss 0.000928144, throughput 9.59666K wps
[Epoch 149 Batch 90/162] avg loss 0.000943175, throughput 9.37845K wps
[Epoch 149 Batch 120/162] avg loss 0.000837829, throughput 9.52998K wps
[Epoch 149 Batch 150/162] avg loss 0.000991119, throughput 9.55263K wps
Begin Testing...
[Epoch 149] train avg loss 0.000915379, dev acc 0.9256, dev avg loss 0.203768, throughput 9.55377K wps
[Epoch 150 Batch 30/162] avg loss 0.000894931, throughput 9.61947K wps
[Epoch 150 Batch 60/162] avg loss 0.00105173, throughput 9.43664K wps
[Epoch 150 Batch 90/162] avg loss 0.000853448, throughput 9.52377K wps
[Epoch 150 Batch 120/162] avg loss 0.000874729, throughput 9.5251K wps
[Epoch 150 Batch 150/162] avg loss 0.000868195, throughput 9.35125K wps
Begin Testing...
[Epoch 150] train avg loss 0.000907602, dev acc 0.9278, dev avg loss 0.204367, throughput 9.47348K wps
[Epoch 151 Batch 30/162] avg loss 0.000864301, throughput 9.77018K wps
[Epoch 151 Batch 60/162] avg loss 0.000927168, throughput 9.48831K wps
[Epoch 151 Batch 90/162] avg loss 0.000851447, throughput 9.36074K wps
[Epoch 151 Batch 120/162] avg loss 0.000918872, throughput 9.56251K wps
[Epoch 151 Batch 150/162] avg loss 0.000964523, throughput 9.5985K wps
Begin Testing...
[Epoch 151] train avg loss 0.000912747, dev acc 0.9289, dev avg loss 0.204411, throughput 9.55276K wps
Observed Improvement.
Begin Testing...
[Epoch 152 Batch 30/162] avg loss 0.000940982, throughput 9.53386K wps
[Epoch 152 Batch 60/162] avg loss 0.000829598, throughput 9.35297K wps
[Epoch 152 Batch 90/162] avg loss 0.00092636, throughput 9.49659K wps
[Epoch 152 Batch 120/162] avg loss 0.000869774, throughput 9.52851K wps
[Epoch 152 Batch 150/162] avg loss 0.000896474, throughput 9.37856K wps
Begin Testing...
[Epoch 152] train avg loss 0.000894508, dev acc 0.9267, dev avg loss 0.205323, throughput 9.4539K wps
[Epoch 153 Batch 30/162] avg loss 0.000902664, throughput 9.62486K wps
[Epoch 153 Batch 60/162] avg loss 0.000847587, throughput 9.42663K wps
[Epoch 153 Batch 90/162] avg loss 0.000875152, throughput 9.52133K wps
[Epoch 153 Batch 120/162] avg loss 0.00087854, throughput 9.31816K wps
[Epoch 153 Batch 150/162] avg loss 0.000790581, throughput 9.44202K wps
Begin Testing...
[Epoch 153] train avg loss 0.000875249, dev acc 0.9278, dev avg loss 0.204772, throughput 9.45928K wps
[Epoch 154 Batch 30/162] avg loss 0.000869225, throughput 9.83318K wps
[Epoch 154 Batch 60/162] avg loss 0.000850324, throughput 9.4484K wps
[Epoch 154 Batch 90/162] avg loss 0.000780996, throughput 9.56357K wps
[Epoch 154 Batch 120/162] avg loss 0.000813656, throughput 9.3718K wps
[Epoch 154 Batch 150/162] avg loss 0.000888413, throughput 9.54742K wps
Begin Testing...
[Epoch 154] train avg loss 0.000840819, dev acc 0.9233, dev avg loss 0.204, throughput 9.56401K wps
[Epoch 155 Batch 30/162] avg loss 0.000847852, throughput 9.50497K wps
[Epoch 155 Batch 60/162] avg loss 0.000978852, throughput 9.44806K wps
[Epoch 155 Batch 90/162] avg loss 0.000870127, throughput 9.46767K wps
[Epoch 155 Batch 120/162] avg loss 0.000932536, throughput 9.47596K wps
[Epoch 155 Batch 150/162] avg loss 0.000789717, throughput 9.30306K wps
Begin Testing...
[Epoch 155] train avg loss 0.000869254, dev acc 0.9278, dev avg loss 0.205028, throughput 9.43357K wps
[Epoch 156 Batch 30/162] avg loss 0.00085063, throughput 9.79148K wps
[Epoch 156 Batch 60/162] avg loss 0.000864336, throughput 9.56835K wps
[Epoch 156 Batch 90/162] avg loss 0.000915429, throughput 9.47221K wps
[Epoch 156 Batch 120/162] avg loss 0.000821389, throughput 9.53076K wps
[Epoch 156 Batch 150/162] avg loss 0.000915725, throughput 9.43371K wps
Begin Testing...
[Epoch 156] train avg loss 0.000864975, dev acc 0.9267, dev avg loss 0.205699, throughput 9.55141K wps
[Epoch 157 Batch 30/162] avg loss 0.000857543, throughput 9.74924K wps
[Epoch 157 Batch 60/162] avg loss 0.000747053, throughput 9.26738K wps
[Epoch 157 Batch 90/162] avg loss 0.000845284, throughput 9.50649K wps
[Epoch 157 Batch 120/162] avg loss 0.00086119, throughput 9.36998K wps
[Epoch 157 Batch 150/162] avg loss 0.000832582, throughput 9.47985K wps
Begin Testing...
[Epoch 157] train avg loss 0.000837415, dev acc 0.9278, dev avg loss 0.20522, throughput 9.45538K wps
[Epoch 158 Batch 30/162] avg loss 0.000895428, throughput 9.67385K wps
[Epoch 158 Batch 60/162] avg loss 0.000799986, throughput 9.27792K wps
[Epoch 158 Batch 90/162] avg loss 0.000847414, throughput 9.47604K wps
[Epoch 158 Batch 120/162] avg loss 0.000849242, throughput 9.53841K wps
[Epoch 158 Batch 150/162] avg loss 0.000834863, throughput 9.31505K wps
Begin Testing...
[Epoch 158] train avg loss 0.0008499, dev acc 0.9289, dev avg loss 0.2048, throughput 9.45449K wps
Observed Improvement.
Begin Testing...
[Epoch 159 Batch 30/162] avg loss 0.000814007, throughput 9.63001K wps
[Epoch 159 Batch 60/162] avg loss 0.000694777, throughput 9.59699K wps
[Epoch 159 Batch 90/162] avg loss 0.000820888, throughput 9.44046K wps
[Epoch 159 Batch 120/162] avg loss 0.00083197, throughput 9.46681K wps
[Epoch 159 Batch 150/162] avg loss 0.000836932, throughput 9.38033K wps
Begin Testing...
[Epoch 159] train avg loss 0.000822698, dev acc 0.9289, dev avg loss 0.205739, throughput 9.49432K wps
Observed Improvement.
Begin Testing...
[Epoch 160 Batch 30/162] avg loss 0.000769963, throughput 9.60296K wps
[Epoch 160 Batch 60/162] avg loss 0.000732542, throughput 9.42826K wps
[Epoch 160 Batch 90/162] avg loss 0.000819944, throughput 9.45847K wps
[Epoch 160 Batch 120/162] avg loss 0.000815471, throughput 9.28406K wps
[Epoch 160 Batch 150/162] avg loss 0.000753557, throughput 9.50907K wps
Begin Testing...
[Epoch 160] train avg loss 0.000780258, dev acc 0.9256, dev avg loss 0.207033, throughput 9.4642K wps
[Epoch 161 Batch 30/162] avg loss 0.000911137, throughput 9.69094K wps
[Epoch 161 Batch 60/162] avg loss 0.000889091, throughput 9.47811K wps
[Epoch 161 Batch 90/162] avg loss 0.00072924, throughput 9.46165K wps
[Epoch 161 Batch 120/162] avg loss 0.000818159, throughput 9.33961K wps
[Epoch 161 Batch 150/162] avg loss 0.000799714, throughput 9.49461K wps
Begin Testing...
[Epoch 161] train avg loss 0.000820947, dev acc 0.9267, dev avg loss 0.208371, throughput 9.50117K wps
[Epoch 162 Batch 30/162] avg loss 0.000799662, throughput 9.49325K wps
[Epoch 162 Batch 60/162] avg loss 0.000744536, throughput 9.45819K wps
[Epoch 162 Batch 90/162] avg loss 0.000893315, throughput 9.44467K wps
[Epoch 162 Batch 120/162] avg loss 0.000737231, throughput 9.41677K wps
[Epoch 162 Batch 150/162] avg loss 0.000836251, throughput 9.35945K wps
Begin Testing...
[Epoch 162] train avg loss 0.000800293, dev acc 0.9267, dev avg loss 0.20508, throughput 9.42508K wps
[Epoch 163 Batch 30/162] avg loss 0.000788458, throughput 9.76169K wps
[Epoch 163 Batch 60/162] avg loss 0.000737328, throughput 9.55207K wps
[Epoch 163 Batch 90/162] avg loss 0.000760845, throughput 9.44537K wps
[Epoch 163 Batch 120/162] avg loss 0.000756234, throughput 9.60798K wps
[Epoch 163 Batch 150/162] avg loss 0.000768407, throughput 9.34309K wps
Begin Testing...
[Epoch 163] train avg loss 0.000769118, dev acc 0.9289, dev avg loss 0.205973, throughput 9.53523K wps
Observed Improvement.
Begin Testing...
[Epoch 164 Batch 30/162] avg loss 0.000753967, throughput 9.52309K wps
[Epoch 164 Batch 60/162] avg loss 0.000786314, throughput 9.55037K wps
[Epoch 164 Batch 90/162] avg loss 0.000790451, throughput 9.3162K wps
[Epoch 164 Batch 120/162] avg loss 0.000768117, throughput 9.48788K wps
[Epoch 164 Batch 150/162] avg loss 0.0007085, throughput 9.29212K wps
Begin Testing...
[Epoch 164] train avg loss 0.000766707, dev acc 0.9244, dev avg loss 0.205341, throughput 9.42717K wps
[Epoch 165 Batch 30/162] avg loss 0.000766935, throughput 9.56068K wps
[Epoch 165 Batch 60/162] avg loss 0.000808047, throughput 9.42224K wps
[Epoch 165 Batch 90/162] avg loss 0.000919502, throughput 9.46683K wps
[Epoch 165 Batch 120/162] avg loss 0.000744256, throughput 9.42548K wps
[Epoch 165 Batch 150/162] avg loss 0.000787552, throughput 9.48676K wps
Begin Testing...
[Epoch 165] train avg loss 0.000793315, dev acc 0.9267, dev avg loss 0.206834, throughput 9.45964K wps
[Epoch 166 Batch 30/162] avg loss 0.000733648, throughput 9.66295K wps
[Epoch 166 Batch 60/162] avg loss 0.000742871, throughput 9.48458K wps
[Epoch 166 Batch 90/162] avg loss 0.000780224, throughput 9.62589K wps
[Epoch 166 Batch 120/162] avg loss 0.000757705, throughput 9.4058K wps
[Epoch 166 Batch 150/162] avg loss 0.000784957, throughput 9.47442K wps
Begin Testing...
[Epoch 166] train avg loss 0.00075765, dev acc 0.9289, dev avg loss 0.205511, throughput 9.54243K wps
Observed Improvement.
Begin Testing...
[Epoch 167 Batch 30/162] avg loss 0.000743779, throughput 9.58573K wps
[Epoch 167 Batch 60/162] avg loss 0.00067398, throughput 9.44875K wps
[Epoch 167 Batch 90/162] avg loss 0.000823203, throughput 9.45681K wps
[Epoch 167 Batch 120/162] avg loss 0.000708059, throughput 9.48444K wps
[Epoch 167 Batch 150/162] avg loss 0.000781778, throughput 9.41633K wps
Begin Testing...
[Epoch 167] train avg loss 0.0007515, dev acc 0.9278, dev avg loss 0.205485, throughput 9.4802K wps
[Epoch 168 Batch 30/162] avg loss 0.000814872, throughput 9.55182K wps
[Epoch 168 Batch 60/162] avg loss 0.000685012, throughput 9.46895K wps
[Epoch 168 Batch 90/162] avg loss 0.00073099, throughput 9.34644K wps
[Epoch 168 Batch 120/162] avg loss 0.000729021, throughput 9.42427K wps
[Epoch 168 Batch 150/162] avg loss 0.000834507, throughput 9.50117K wps
Begin Testing...
[Epoch 168] train avg loss 0.000757128, dev acc 0.9289, dev avg loss 0.205992, throughput 9.45682K wps
Observed Improvement.
Begin Testing...
[Epoch 169 Batch 30/162] avg loss 0.0007176, throughput 9.54428K wps
[Epoch 169 Batch 60/162] avg loss 0.000817085, throughput 9.56158K wps
[Epoch 169 Batch 90/162] avg loss 0.000674976, throughput 9.32326K wps
[Epoch 169 Batch 120/162] avg loss 0.000772329, throughput 9.27531K wps
[Epoch 169 Batch 150/162] avg loss 0.000751819, throughput 9.4635K wps
Begin Testing...
[Epoch 169] train avg loss 0.000755091, dev acc 0.9233, dev avg loss 0.205263, throughput 9.42176K wps
[Epoch 170 Batch 30/162] avg loss 0.00073416, throughput 9.50236K wps
[Epoch 170 Batch 60/162] avg loss 0.000707708, throughput 9.3719K wps
[Epoch 170 Batch 90/162] avg loss 0.00068945, throughput 9.56413K wps
[Epoch 170 Batch 120/162] avg loss 0.000671178, throughput 9.43191K wps
[Epoch 170 Batch 150/162] avg loss 0.000807467, throughput 9.41213K wps
Begin Testing...
[Epoch 170] train avg loss 0.000721351, dev acc 0.9289, dev avg loss 0.20604, throughput 9.46098K wps
Observed Improvement.
Begin Testing...
[Epoch 171 Batch 30/162] avg loss 0.000673589, throughput 9.65845K wps
[Epoch 171 Batch 60/162] avg loss 0.000763474, throughput 9.55471K wps
[Epoch 171 Batch 90/162] avg loss 0.000823721, throughput 9.58077K wps
[Epoch 171 Batch 120/162] avg loss 0.000690431, throughput 9.36608K wps
[Epoch 171 Batch 150/162] avg loss 0.000763297, throughput 9.58085K wps
Begin Testing...
[Epoch 171] train avg loss 0.000736637, dev acc 0.9222, dev avg loss 0.205743, throughput 9.53383K wps
[Epoch 172 Batch 30/162] avg loss 0.000654849, throughput 9.5314K wps
[Epoch 172 Batch 60/162] avg loss 0.000748247, throughput 9.23611K wps
[Epoch 172 Batch 90/162] avg loss 0.000702072, throughput 9.42079K wps
[Epoch 172 Batch 120/162] avg loss 0.000697684, throughput 9.46839K wps
[Epoch 172 Batch 150/162] avg loss 0.00069075, throughput 9.37675K wps
Begin Testing...
[Epoch 172] train avg loss 0.000710038, dev acc 0.9267, dev avg loss 0.207979, throughput 9.42767K wps
[Epoch 173 Batch 30/162] avg loss 0.000626582, throughput 9.77174K wps
[Epoch 173 Batch 60/162] avg loss 0.000728995, throughput 9.50931K wps
[Epoch 173 Batch 90/162] avg loss 0.00068716, throughput 9.49452K wps
[Epoch 173 Batch 120/162] avg loss 0.000714657, throughput 9.37675K wps
[Epoch 173 Batch 150/162] avg loss 0.000664465, throughput 9.29997K wps
Begin Testing...
[Epoch 173] train avg loss 0.000682774, dev acc 0.9289, dev avg loss 0.206349, throughput 9.48535K wps
Observed Improvement.
Begin Testing...
[Epoch 174 Batch 30/162] avg loss 0.000740221, throughput 9.56209K wps
[Epoch 174 Batch 60/162] avg loss 0.000690439, throughput 9.31056K wps
[Epoch 174 Batch 90/162] avg loss 0.000722441, throughput 9.49293K wps
[Epoch 174 Batch 120/162] avg loss 0.000620718, throughput 9.47802K wps
[Epoch 174 Batch 150/162] avg loss 0.000742748, throughput 9.46055K wps
Begin Testing...
[Epoch 174] train avg loss 0.000700674, dev acc 0.9267, dev avg loss 0.206004, throughput 9.45661K wps
[Epoch 175 Batch 30/162] avg loss 0.000673981, throughput 9.41989K wps
[Epoch 175 Batch 60/162] avg loss 0.000660247, throughput 9.39125K wps
[Epoch 175 Batch 90/162] avg loss 0.000760396, throughput 9.5171K wps
[Epoch 175 Batch 120/162] avg loss 0.000707504, throughput 9.5156K wps
[Epoch 175 Batch 150/162] avg loss 0.000714897, throughput 9.36114K wps
Begin Testing...
[Epoch 175] train avg loss 0.000712366, dev acc 0.9278, dev avg loss 0.208555, throughput 9.43218K wps
[Epoch 176 Batch 30/162] avg loss 0.000778857, throughput 9.59421K wps
[Epoch 176 Batch 60/162] avg loss 0.000727038, throughput 9.39782K wps
[Epoch 176 Batch 90/162] avg loss 0.000674844, throughput 9.32626K wps
[Epoch 176 Batch 120/162] avg loss 0.000660369, throughput 9.56783K wps
[Epoch 176 Batch 150/162] avg loss 0.000739998, throughput 9.38982K wps
Begin Testing...
[Epoch 176] train avg loss 0.000716046, dev acc 0.9278, dev avg loss 0.20822, throughput 9.44843K wps
[Epoch 177 Batch 30/162] avg loss 0.000678428, throughput 9.58359K wps
[Epoch 177 Batch 60/162] avg loss 0.000698532, throughput 9.42352K wps
[Epoch 177 Batch 90/162] avg loss 0.000690659, throughput 9.49164K wps
[Epoch 177 Batch 120/162] avg loss 0.000659115, throughput 9.39805K wps
[Epoch 177 Batch 150/162] avg loss 0.000693169, throughput 9.40968K wps
Begin Testing...
[Epoch 177] train avg loss 0.00069543, dev acc 0.9278, dev avg loss 0.207497, throughput 9.47459K wps
[Epoch 178 Batch 30/162] avg loss 0.000746484, throughput 9.74935K wps
[Epoch 178 Batch 60/162] avg loss 0.000736255, throughput 9.50901K wps
[Epoch 178 Batch 90/162] avg loss 0.000589108, throughput 9.50845K wps
[Epoch 178 Batch 120/162] avg loss 0.000730224, throughput 9.47371K wps
[Epoch 178 Batch 150/162] avg loss 0.000733984, throughput 9.46451K wps
Begin Testing...
[Epoch 178] train avg loss 0.000696745, dev acc 0.9289, dev avg loss 0.206933, throughput 9.52636K wps
Observed Improvement.
Begin Testing...
[Epoch 179 Batch 30/162] avg loss 0.000718516, throughput 9.58062K wps
[Epoch 179 Batch 60/162] avg loss 0.000675776, throughput 9.37707K wps
[Epoch 179 Batch 90/162] avg loss 0.000673322, throughput 9.46623K wps
[Epoch 179 Batch 120/162] avg loss 0.000738529, throughput 9.4373K wps
[Epoch 179 Batch 150/162] avg loss 0.000734194, throughput 9.54815K wps
Begin Testing...
[Epoch 179] train avg loss 0.000707145, dev acc 0.9278, dev avg loss 0.208353, throughput 9.47702K wps
[Epoch 180 Batch 30/162] avg loss 0.000610996, throughput 9.65409K wps
[Epoch 180 Batch 60/162] avg loss 0.000749, throughput 9.55319K wps
[Epoch 180 Batch 90/162] avg loss 0.000652009, throughput 9.36428K wps
[Epoch 180 Batch 120/162] avg loss 0.000675604, throughput 9.49928K wps
[Epoch 180 Batch 150/162] avg loss 0.000694875, throughput 9.51038K wps
Begin Testing...
[Epoch 180] train avg loss 0.000691013, dev acc 0.9278, dev avg loss 0.207631, throughput 9.51908K wps
[Epoch 181 Batch 30/162] avg loss 0.000588224, throughput 9.59914K wps
[Epoch 181 Batch 60/162] avg loss 0.000635872, throughput 9.46946K wps
[Epoch 181 Batch 90/162] avg loss 0.00069915, throughput 9.41339K wps
[Epoch 181 Batch 120/162] avg loss 0.000753804, throughput 9.37621K wps
[Epoch 181 Batch 150/162] avg loss 0.000662609, throughput 9.41879K wps
Begin Testing...
[Epoch 181] train avg loss 0.000663274, dev acc 0.9278, dev avg loss 0.208206, throughput 9.46404K wps
[Epoch 182 Batch 30/162] avg loss 0.000666476, throughput 9.47303K wps
[Epoch 182 Batch 60/162] avg loss 0.000627907, throughput 9.38286K wps
[Epoch 182 Batch 90/162] avg loss 0.000625086, throughput 9.40081K wps
[Epoch 182 Batch 120/162] avg loss 0.000703524, throughput 9.41173K wps
[Epoch 182 Batch 150/162] avg loss 0.000635931, throughput 9.52645K wps
Begin Testing...
[Epoch 182] train avg loss 0.000645783, dev acc 0.9256, dev avg loss 0.206906, throughput 9.42507K wps
[Epoch 183 Batch 30/162] avg loss 0.000705587, throughput 9.65051K wps
[Epoch 183 Batch 60/162] avg loss 0.000615413, throughput 9.33434K wps
[Epoch 183 Batch 90/162] avg loss 0.000668193, throughput 9.55033K wps
[Epoch 183 Batch 120/162] avg loss 0.000605602, throughput 9.39778K wps
[Epoch 183 Batch 150/162] avg loss 0.000683876, throughput 9.41532K wps
Begin Testing...
[Epoch 183] train avg loss 0.000657171, dev acc 0.9278, dev avg loss 0.20733, throughput 9.4606K wps
[Epoch 184 Batch 30/162] avg loss 0.000613128, throughput 9.86033K wps
[Epoch 184 Batch 60/162] avg loss 0.000661191, throughput 9.5636K wps
[Epoch 184 Batch 90/162] avg loss 0.000670968, throughput 9.58659K wps
[Epoch 184 Batch 120/162] avg loss 0.000616377, throughput 9.37526K wps
[Epoch 184 Batch 150/162] avg loss 0.000678205, throughput 9.52692K wps
Begin Testing...
[Epoch 184] train avg loss 0.00064571, dev acc 0.9278, dev avg loss 0.207498, throughput 9.56145K wps
[Epoch 185 Batch 30/162] avg loss 0.000593121, throughput 9.7672K wps
[Epoch 185 Batch 60/162] avg loss 0.000601184, throughput 9.49687K wps
[Epoch 185 Batch 90/162] avg loss 0.000684217, throughput 9.57266K wps
[Epoch 185 Batch 120/162] avg loss 0.00068065, throughput 9.27643K wps
[Epoch 185 Batch 150/162] avg loss 0.000664552, throughput 9.42686K wps
Begin Testing...
[Epoch 185] train avg loss 0.000644593, dev acc 0.9278, dev avg loss 0.208221, throughput 9.49571K wps
[Epoch 186 Batch 30/162] avg loss 0.000637585, throughput 9.61942K wps
[Epoch 186 Batch 60/162] avg loss 0.000675925, throughput 9.56263K wps
[Epoch 186 Batch 90/162] avg loss 0.000603797, throughput 9.41987K wps
[Epoch 186 Batch 120/162] avg loss 0.00057505, throughput 9.37527K wps
[Epoch 186 Batch 150/162] avg loss 0.000590562, throughput 9.38524K wps
Begin Testing...
[Epoch 186] train avg loss 0.00062278, dev acc 0.9289, dev avg loss 0.208064, throughput 9.47337K wps
Observed Improvement.
Begin Testing...
[Epoch 187 Batch 30/162] avg loss 0.000647179, throughput 9.64779K wps
[Epoch 187 Batch 60/162] avg loss 0.000534438, throughput 9.36101K wps
[Epoch 187 Batch 90/162] avg loss 0.000608947, throughput 9.4966K wps
[Epoch 187 Batch 120/162] avg loss 0.000657451, throughput 9.37471K wps
[Epoch 187 Batch 150/162] avg loss 0.00063951, throughput 9.30218K wps
Begin Testing...
[Epoch 187] train avg loss 0.000626374, dev acc 0.9300, dev avg loss 0.208404, throughput 9.43009K wps
Observed Improvement.
Begin Testing...
[Epoch 188 Batch 30/162] avg loss 0.00063143, throughput 9.5983K wps
[Epoch 188 Batch 60/162] avg loss 0.000670443, throughput 9.46867K wps
[Epoch 188 Batch 90/162] avg loss 0.000635734, throughput 9.53786K wps
[Epoch 188 Batch 120/162] avg loss 0.000631539, throughput 9.50352K wps
[Epoch 188 Batch 150/162] avg loss 0.000787645, throughput 9.50386K wps
Begin Testing...
[Epoch 188] train avg loss 0.000667255, dev acc 0.9289, dev avg loss 0.2075, throughput 9.51423K wps
[Epoch 189 Batch 30/162] avg loss 0.000687224, throughput 9.60754K wps
[Epoch 189 Batch 60/162] avg loss 0.000580855, throughput 9.43297K wps
[Epoch 189 Batch 90/162] avg loss 0.0006352, throughput 9.63995K wps
[Epoch 189 Batch 120/162] avg loss 0.000645874, throughput 9.50713K wps
[Epoch 189 Batch 150/162] avg loss 0.000597618, throughput 9.31089K wps
Begin Testing...
[Epoch 189] train avg loss 0.000638436, dev acc 0.9256, dev avg loss 0.207325, throughput 9.49553K wps
[Epoch 190 Batch 30/162] avg loss 0.000627836, throughput 9.76987K wps
[Epoch 190 Batch 60/162] avg loss 0.000501036, throughput 9.39468K wps
[Epoch 190 Batch 90/162] avg loss 0.00064301, throughput 9.53015K wps
[Epoch 190 Batch 120/162] avg loss 0.000637962, throughput 9.34064K wps
[Epoch 190 Batch 150/162] avg loss 0.000662991, throughput 9.49812K wps
Begin Testing...
[Epoch 190] train avg loss 0.000619237, dev acc 0.9289, dev avg loss 0.208361, throughput 9.49916K wps
[Epoch 191 Batch 30/162] avg loss 0.000548056, throughput 9.67687K wps
[Epoch 191 Batch 60/162] avg loss 0.000643805, throughput 9.4346K wps
[Epoch 191 Batch 90/162] avg loss 0.000575352, throughput 9.46137K wps
[Epoch 191 Batch 120/162] avg loss 0.000668765, throughput 9.582K wps
[Epoch 191 Batch 150/162] avg loss 0.000614356, throughput 9.51936K wps
Begin Testing...
[Epoch 191] train avg loss 0.000607639, dev acc 0.9289, dev avg loss 0.209567, throughput 9.53768K wps
[Epoch 192 Batch 30/162] avg loss 0.000613261, throughput 9.66633K wps
[Epoch 192 Batch 60/162] avg loss 0.000604304, throughput 9.59052K wps
[Epoch 192 Batch 90/162] avg loss 0.000610424, throughput 9.39293K wps
[Epoch 192 Batch 120/162] avg loss 0.000678899, throughput 9.55306K wps
[Epoch 192 Batch 150/162] avg loss 0.000636774, throughput 9.475K wps
Begin Testing...
[Epoch 192] train avg loss 0.000624002, dev acc 0.9289, dev avg loss 0.208557, throughput 9.54161K wps
[Epoch 193 Batch 30/162] avg loss 0.000609266, throughput 9.66367K wps
[Epoch 193 Batch 60/162] avg loss 0.000664505, throughput 9.46784K wps
[Epoch 193 Batch 90/162] avg loss 0.000607476, throughput 9.48806K wps
[Epoch 193 Batch 120/162] avg loss 0.000591541, throughput 9.5508K wps
[Epoch 193 Batch 150/162] avg loss 0.0005407, throughput 9.34428K wps
Begin Testing...
[Epoch 193] train avg loss 0.000612354, dev acc 0.9278, dev avg loss 0.208242, throughput 9.5217K wps
[Epoch 194 Batch 30/162] avg loss 0.000658979, throughput 9.63338K wps
[Epoch 194 Batch 60/162] avg loss 0.000580899, throughput 9.44413K wps
[Epoch 194 Batch 90/162] avg loss 0.00060238, throughput 9.32181K wps
[Epoch 194 Batch 120/162] avg loss 0.000564177, throughput 9.52364K wps
[Epoch 194 Batch 150/162] avg loss 0.000567308, throughput 9.39769K wps
Begin Testing...
[Epoch 194] train avg loss 0.000583511, dev acc 0.9289, dev avg loss 0.209278, throughput 9.45241K wps
[Epoch 195 Batch 30/162] avg loss 0.000659279, throughput 9.60089K wps
[Epoch 195 Batch 60/162] avg loss 0.000579547, throughput 9.42961K wps
[Epoch 195 Batch 90/162] avg loss 0.000579195, throughput 9.60515K wps
[Epoch 195 Batch 120/162] avg loss 0.000496326, throughput 9.53598K wps
[Epoch 195 Batch 150/162] avg loss 0.000553938, throughput 9.31662K wps
Begin Testing...
[Epoch 195] train avg loss 0.000573327, dev acc 0.9278, dev avg loss 0.210618, throughput 9.49732K wps
[Epoch 196 Batch 30/162] avg loss 0.000599829, throughput 9.67967K wps
[Epoch 196 Batch 60/162] avg loss 0.000623847, throughput 9.49694K wps
[Epoch 196 Batch 90/162] avg loss 0.000566855, throughput 9.56408K wps
[Epoch 196 Batch 120/162] avg loss 0.000600928, throughput 9.44838K wps
[Epoch 196 Batch 150/162] avg loss 0.00057985, throughput 9.58742K wps
Begin Testing...
[Epoch 196] train avg loss 0.000592891, dev acc 0.9267, dev avg loss 0.209227, throughput 9.56175K wps
[Epoch 197 Batch 30/162] avg loss 0.000600422, throughput 9.59716K wps
[Epoch 197 Batch 60/162] avg loss 0.000559956, throughput 9.4545K wps
[Epoch 197 Batch 90/162] avg loss 0.0006541, throughput 9.4957K wps
[Epoch 197 Batch 120/162] avg loss 0.000607976, throughput 9.3918K wps
[Epoch 197 Batch 150/162] avg loss 0.000576404, throughput 9.49248K wps
Begin Testing...
[Epoch 197] train avg loss 0.000611443, dev acc 0.9300, dev avg loss 0.209414, throughput 9.49809K wps
Observed Improvement.
Begin Testing...
[Epoch 198 Batch 30/162] avg loss 0.000553406, throughput 9.79464K wps
[Epoch 198 Batch 60/162] avg loss 0.000559714, throughput 9.51104K wps
[Epoch 198 Batch 90/162] avg loss 0.00051738, throughput 9.35305K wps
[Epoch 198 Batch 120/162] avg loss 0.000566188, throughput 9.39311K wps
[Epoch 198 Batch 150/162] avg loss 0.000609753, throughput 9.36603K wps
Begin Testing...
[Epoch 198] train avg loss 0.000557862, dev acc 0.9289, dev avg loss 0.209707, throughput 9.47758K wps
[Epoch 199 Batch 30/162] avg loss 0.000607188, throughput 9.46696K wps
[Epoch 199 Batch 60/162] avg loss 0.000644563, throughput 9.42667K wps
[Epoch 199 Batch 90/162] avg loss 0.000577037, throughput 9.41084K wps
[Epoch 199 Batch 120/162] avg loss 0.000567405, throughput 9.62297K wps
[Epoch 199 Batch 150/162] avg loss 0.000580255, throughput 9.35673K wps
Begin Testing...
[Epoch 199] train avg loss 0.000595575, dev acc 0.9300, dev avg loss 0.209837, throughput 9.46521K wps
Observed Improvement.
Begin Testing...
Test loss 0.209254, test acc 0.9220
Total time cost 444.17s
[Epoch 0 Batch 30/162] avg loss 0.013829, throughput 7.208K wps
[Epoch 0 Batch 60/162] avg loss 0.0136989, throughput 9.29081K wps
[Epoch 0 Batch 90/162] avg loss 0.0137143, throughput 9.43193K wps
[Epoch 0 Batch 120/162] avg loss 0.0136175, throughput 9.18434K wps
[Epoch 0 Batch 150/162] avg loss 0.0132864, throughput 9.53365K wps
Begin Testing...
[Epoch 0] train avg loss 0.0136096, dev acc 0.7678, dev avg loss 0.659978, throughput 8.87515K wps
Observed Improvement.
Begin Testing...
[Epoch 1 Batch 30/162] avg loss 0.0132097, throughput 9.56007K wps
[Epoch 1 Batch 60/162] avg loss 0.0130817, throughput 9.45131K wps
[Epoch 1 Batch 90/162] avg loss 0.0129974, throughput 9.31261K wps
[Epoch 1 Batch 120/162] avg loss 0.0129093, throughput 9.39633K wps
[Epoch 1 Batch 150/162] avg loss 0.0128788, throughput 9.52237K wps
Begin Testing...
[Epoch 1] train avg loss 0.0130047, dev acc 0.8500, dev avg loss 0.628756, throughput 9.44338K wps
Observed Improvement.
Begin Testing...
[Epoch 2 Batch 30/162] avg loss 0.0127451, throughput 9.48562K wps
[Epoch 2 Batch 60/162] avg loss 0.0125132, throughput 9.39321K wps
[Epoch 2 Batch 90/162] avg loss 0.0124621, throughput 9.50934K wps
[Epoch 2 Batch 120/162] avg loss 0.0122962, throughput 9.56354K wps
[Epoch 2 Batch 150/162] avg loss 0.012162, throughput 9.3427K wps
Begin Testing...
[Epoch 2] train avg loss 0.0124054, dev acc 0.8789, dev avg loss 0.59394, throughput 9.43472K wps
Observed Improvement.
Begin Testing...
[Epoch 3 Batch 30/162] avg loss 0.0119624, throughput 9.61283K wps
[Epoch 3 Batch 60/162] avg loss 0.0117301, throughput 9.26657K wps
[Epoch 3 Batch 90/162] avg loss 0.0117134, throughput 9.45701K wps
[Epoch 3 Batch 120/162] avg loss 0.0115655, throughput 9.57762K wps
[Epoch 3 Batch 150/162] avg loss 0.0113824, throughput 9.37345K wps
Begin Testing...
[Epoch 3] train avg loss 0.0116558, dev acc 0.8789, dev avg loss 0.554149, throughput 9.46399K wps
Observed Improvement.
Begin Testing...
[Epoch 4 Batch 30/162] avg loss 0.0113272, throughput 9.59266K wps
[Epoch 4 Batch 60/162] avg loss 0.0112233, throughput 9.40818K wps
[Epoch 4 Batch 90/162] avg loss 0.0108828, throughput 9.3778K wps
[Epoch 4 Batch 120/162] avg loss 0.0106731, throughput 9.25345K wps
[Epoch 4 Batch 150/162] avg loss 0.0106371, throughput 9.32118K wps
Begin Testing...
[Epoch 4] train avg loss 0.0109225, dev acc 0.8844, dev avg loss 0.511193, throughput 9.38419K wps
Observed Improvement.
Begin Testing...
[Epoch 5 Batch 30/162] avg loss 0.0104534, throughput 9.68679K wps
[Epoch 5 Batch 60/162] avg loss 0.010326, throughput 9.46981K wps
[Epoch 5 Batch 90/162] avg loss 0.0100933, throughput 9.44781K wps
[Epoch 5 Batch 120/162] avg loss 0.00977012, throughput 9.3716K wps
[Epoch 5 Batch 150/162] avg loss 0.00964586, throughput 9.52851K wps
Begin Testing...
[Epoch 5] train avg loss 0.0100131, dev acc 0.8867, dev avg loss 0.469111, throughput 9.49147K wps
Observed Improvement.
Begin Testing...
[Epoch 6 Batch 30/162] avg loss 0.00959034, throughput 9.4926K wps
[Epoch 6 Batch 60/162] avg loss 0.00936672, throughput 9.28209K wps
[Epoch 6 Batch 90/162] avg loss 0.009264, throughput 9.36387K wps
[Epoch 6 Batch 120/162] avg loss 0.00933306, throughput 9.46275K wps
[Epoch 6 Batch 150/162] avg loss 0.00912951, throughput 9.58345K wps
Begin Testing...
[Epoch 6] train avg loss 0.00928592, dev acc 0.8844, dev avg loss 0.432725, throughput 9.41603K wps
[Epoch 7 Batch 30/162] avg loss 0.00891287, throughput 9.53072K wps
[Epoch 7 Batch 60/162] avg loss 0.0087248, throughput 9.30298K wps
[Epoch 7 Batch 90/162] avg loss 0.00860508, throughput 9.43795K wps
[Epoch 7 Batch 120/162] avg loss 0.00848804, throughput 9.43848K wps
[Epoch 7 Batch 150/162] avg loss 0.00842134, throughput 9.18999K wps
Begin Testing...
[Epoch 7] train avg loss 0.00860394, dev acc 0.8900, dev avg loss 0.40124, throughput 9.38133K wps
Observed Improvement.
Begin Testing...
[Epoch 8 Batch 30/162] avg loss 0.00823696, throughput 9.57921K wps
[Epoch 8 Batch 60/162] avg loss 0.00785191, throughput 9.60768K wps
[Epoch 8 Batch 90/162] avg loss 0.00812602, throughput 9.57244K wps
[Epoch 8 Batch 120/162] avg loss 0.00798649, throughput 9.41999K wps
[Epoch 8 Batch 150/162] avg loss 0.00801619, throughput 9.52847K wps
Begin Testing...
[Epoch 8] train avg loss 0.00806028, dev acc 0.8922, dev avg loss 0.37552, throughput 9.54923K wps
Observed Improvement.
Begin Testing...
[Epoch 9 Batch 30/162] avg loss 0.00773257, throughput 9.56491K wps
[Epoch 9 Batch 60/162] avg loss 0.00757291, throughput 9.33588K wps
[Epoch 9 Batch 90/162] avg loss 0.0076651, throughput 9.38282K wps
[Epoch 9 Batch 120/162] avg loss 0.00757729, throughput 9.30426K wps
[Epoch 9 Batch 150/162] avg loss 0.00782699, throughput 9.63207K wps
Begin Testing...
[Epoch 9] train avg loss 0.00766465, dev acc 0.8933, dev avg loss 0.356858, throughput 9.44644K wps
Observed Improvement.
Begin Testing...
[Epoch 10 Batch 30/162] avg loss 0.00773072, throughput 9.52281K wps
[Epoch 10 Batch 60/162] avg loss 0.00718771, throughput 9.35585K wps
[Epoch 10 Batch 90/162] avg loss 0.00731817, throughput 9.49756K wps
[Epoch 10 Batch 120/162] avg loss 0.00731091, throughput 9.26938K wps
[Epoch 10 Batch 150/162] avg loss 0.00735855, throughput 9.32419K wps
Begin Testing...
[Epoch 10] train avg loss 0.00734001, dev acc 0.8900, dev avg loss 0.341538, throughput 9.40179K wps
[Epoch 11 Batch 30/162] avg loss 0.00738112, throughput 9.42457K wps
[Epoch 11 Batch 60/162] avg loss 0.00693597, throughput 9.31871K wps
[Epoch 11 Batch 90/162] avg loss 0.00698534, throughput 9.44242K wps
[Epoch 11 Batch 120/162] avg loss 0.00691637, throughput 9.34626K wps
[Epoch 11 Batch 150/162] avg loss 0.0065747, throughput 9.52079K wps
Begin Testing...
[Epoch 11] train avg loss 0.00694758, dev acc 0.8967, dev avg loss 0.328893, throughput 9.41109K wps
Observed Improvement.
Begin Testing...
[Epoch 12 Batch 30/162] avg loss 0.00697718, throughput 9.80512K wps
[Epoch 12 Batch 60/162] avg loss 0.00685821, throughput 9.39265K wps
[Epoch 12 Batch 90/162] avg loss 0.00640822, throughput 9.56119K wps
[Epoch 12 Batch 120/162] avg loss 0.00679509, throughput 9.39306K wps
[Epoch 12 Batch 150/162] avg loss 0.00663711, throughput 9.44449K wps
Begin Testing...
[Epoch 12] train avg loss 0.00673237, dev acc 0.8967, dev avg loss 0.319303, throughput 9.50352K wps
Observed Improvement.
Begin Testing...
[Epoch 13 Batch 30/162] avg loss 0.00631625, throughput 9.59615K wps
[Epoch 13 Batch 60/162] avg loss 0.00673149, throughput 9.45037K wps
[Epoch 13 Batch 90/162] avg loss 0.00643841, throughput 9.24484K wps
[Epoch 13 Batch 120/162] avg loss 0.00662944, throughput 9.50795K wps
[Epoch 13 Batch 150/162] avg loss 0.00629367, throughput 9.2873K wps
Begin Testing...
[Epoch 13] train avg loss 0.00646299, dev acc 0.8933, dev avg loss 0.311517, throughput 9.40485K wps
[Epoch 14 Batch 30/162] avg loss 0.00650618, throughput 9.62758K wps
[Epoch 14 Batch 60/162] avg loss 0.00640435, throughput 9.37971K wps
[Epoch 14 Batch 90/162] avg loss 0.00598794, throughput 9.51621K wps
[Epoch 14 Batch 120/162] avg loss 0.00604007, throughput 9.32783K wps
[Epoch 14 Batch 150/162] avg loss 0.00672447, throughput 9.31069K wps
Begin Testing...
[Epoch 14] train avg loss 0.00631478, dev acc 0.8922, dev avg loss 0.304424, throughput 9.4222K wps
[Epoch 15 Batch 30/162] avg loss 0.00603298, throughput 9.56445K wps
[Epoch 15 Batch 60/162] avg loss 0.00597876, throughput 9.31321K wps
[Epoch 15 Batch 90/162] avg loss 0.00658161, throughput 9.45428K wps
[Epoch 15 Batch 120/162] avg loss 0.00596256, throughput 9.41808K wps
[Epoch 15 Batch 150/162] avg loss 0.00612577, throughput 9.44126K wps
Begin Testing...
[Epoch 15] train avg loss 0.00616233, dev acc 0.8967, dev avg loss 0.298807, throughput 9.4311K wps
Observed Improvement.
Begin Testing...
[Epoch 16 Batch 30/162] avg loss 0.00619658, throughput 9.67663K wps
[Epoch 16 Batch 60/162] avg loss 0.00582975, throughput 9.36452K wps
[Epoch 16 Batch 90/162] avg loss 0.005749, throughput 9.51498K wps
[Epoch 16 Batch 120/162] avg loss 0.00585696, throughput 9.47719K wps
[Epoch 16 Batch 150/162] avg loss 0.00598491, throughput 9.53261K wps
Begin Testing...
[Epoch 16] train avg loss 0.00596373, dev acc 0.8956, dev avg loss 0.293864, throughput 9.4888K wps
[Epoch 17 Batch 30/162] avg loss 0.00596259, throughput 9.7391K wps
[Epoch 17 Batch 60/162] avg loss 0.00603559, throughput 9.24237K wps
[Epoch 17 Batch 90/162] avg loss 0.00572383, throughput 9.56008K wps
[Epoch 17 Batch 120/162] avg loss 0.00607349, throughput 9.31242K wps
[Epoch 17 Batch 150/162] avg loss 0.00543478, throughput 9.37332K wps
Begin Testing...
[Epoch 17] train avg loss 0.00582219, dev acc 0.8967, dev avg loss 0.289581, throughput 9.44802K wps
Observed Improvement.
Begin Testing...
[Epoch 18 Batch 30/162] avg loss 0.00581034, throughput 9.59483K wps
[Epoch 18 Batch 60/162] avg loss 0.00579778, throughput 9.33808K wps
[Epoch 18 Batch 90/162] avg loss 0.00554913, throughput 9.57699K wps
[Epoch 18 Batch 120/162] avg loss 0.00554941, throughput 9.39781K wps
[Epoch 18 Batch 150/162] avg loss 0.00560081, throughput 9.28885K wps
Begin Testing...
[Epoch 18] train avg loss 0.00569511, dev acc 0.8989, dev avg loss 0.285296, throughput 9.44187K wps
Observed Improvement.
Begin Testing...
[Epoch 19 Batch 30/162] avg loss 0.00552723, throughput 9.47309K wps
[Epoch 19 Batch 60/162] avg loss 0.00556946, throughput 9.33096K wps
[Epoch 19 Batch 90/162] avg loss 0.0054623, throughput 9.35771K wps
[Epoch 19 Batch 120/162] avg loss 0.00583238, throughput 9.41862K wps
[Epoch 19 Batch 150/162] avg loss 0.0056404, throughput 9.62151K wps
Begin Testing...
[Epoch 19] train avg loss 0.00556018, dev acc 0.8967, dev avg loss 0.282706, throughput 9.42479K wps
[Epoch 20 Batch 30/162] avg loss 0.00569937, throughput 9.41919K wps
[Epoch 20 Batch 60/162] avg loss 0.00546633, throughput 9.32727K wps
[Epoch 20 Batch 90/162] avg loss 0.00527158, throughput 9.33251K wps
[Epoch 20 Batch 120/162] avg loss 0.00496667, throughput 9.4064K wps
[Epoch 20 Batch 150/162] avg loss 0.00569327, throughput 9.46654K wps
Begin Testing...
[Epoch 20] train avg loss 0.00548253, dev acc 0.8989, dev avg loss 0.278738, throughput 9.38916K wps
Observed Improvement.
Begin Testing...
[Epoch 21 Batch 30/162] avg loss 0.00528185, throughput 9.6302K wps
[Epoch 21 Batch 60/162] avg loss 0.00554339, throughput 9.32466K wps
[Epoch 21 Batch 90/162] avg loss 0.00541017, throughput 9.43865K wps
[Epoch 21 Batch 120/162] avg loss 0.00534882, throughput 9.36237K wps
[Epoch 21 Batch 150/162] avg loss 0.00504109, throughput 9.42391K wps
Begin Testing...
[Epoch 21] train avg loss 0.00532251, dev acc 0.9022, dev avg loss 0.275674, throughput 9.43628K wps
Observed Improvement.
Begin Testing...
[Epoch 22 Batch 30/162] avg loss 0.00526041, throughput 9.44638K wps
[Epoch 22 Batch 60/162] avg loss 0.00516438, throughput 9.37734K wps
[Epoch 22 Batch 90/162] avg loss 0.00533399, throughput 9.42346K wps
[Epoch 22 Batch 120/162] avg loss 0.00531593, throughput 9.62042K wps
[Epoch 22 Batch 150/162] avg loss 0.00524851, throughput 9.48451K wps
Begin Testing...
[Epoch 22] train avg loss 0.00527129, dev acc 0.9011, dev avg loss 0.27327, throughput 9.4712K wps
[Epoch 23 Batch 30/162] avg loss 0.00506747, throughput 9.48955K wps
[Epoch 23 Batch 60/162] avg loss 0.00497361, throughput 9.3259K wps
[Epoch 23 Batch 90/162] avg loss 0.00480812, throughput 9.43743K wps
[Epoch 23 Batch 120/162] avg loss 0.00532305, throughput 9.59634K wps
[Epoch 23 Batch 150/162] avg loss 0.00542523, throughput 9.19439K wps
Begin Testing...
[Epoch 23] train avg loss 0.00512465, dev acc 0.8956, dev avg loss 0.270454, throughput 9.40284K wps
[Epoch 24 Batch 30/162] avg loss 0.00561421, throughput 9.63612K wps
[Epoch 24 Batch 60/162] avg loss 0.00473299, throughput 9.50276K wps
[Epoch 24 Batch 90/162] avg loss 0.00516071, throughput 9.23087K wps
[Epoch 24 Batch 120/162] avg loss 0.0051325, throughput 9.53771K wps
[Epoch 24 Batch 150/162] avg loss 0.00497035, throughput 9.2516K wps
Begin Testing...
[Epoch 24] train avg loss 0.00510758, dev acc 0.9000, dev avg loss 0.268241, throughput 9.41771K wps
[Epoch 25 Batch 30/162] avg loss 0.00525126, throughput 9.56689K wps
[Epoch 25 Batch 60/162] avg loss 0.00534225, throughput 9.51869K wps
[Epoch 25 Batch 90/162] avg loss 0.00472423, throughput 9.62562K wps
[Epoch 25 Batch 120/162] avg loss 0.00454245, throughput 9.50176K wps
[Epoch 25 Batch 150/162] avg loss 0.00500999, throughput 9.40196K wps
Begin Testing...
[Epoch 25] train avg loss 0.00492288, dev acc 0.8922, dev avg loss 0.265998, throughput 9.50497K wps
[Epoch 26 Batch 30/162] avg loss 0.00509621, throughput 9.69421K wps
[Epoch 26 Batch 60/162] avg loss 0.00490369, throughput 9.4407K wps
[Epoch 26 Batch 90/162] avg loss 0.00472783, throughput 9.49493K wps
[Epoch 26 Batch 120/162] avg loss 0.00493606, throughput 9.63868K wps
[Epoch 26 Batch 150/162] avg loss 0.00476271, throughput 9.48098K wps
Begin Testing...
[Epoch 26] train avg loss 0.00485542, dev acc 0.9000, dev avg loss 0.263824, throughput 9.52859K wps
[Epoch 27 Batch 30/162] avg loss 0.00488316, throughput 9.70698K wps
[Epoch 27 Batch 60/162] avg loss 0.00480363, throughput 9.3343K wps
[Epoch 27 Batch 90/162] avg loss 0.0047383, throughput 9.34885K wps
[Epoch 27 Batch 120/162] avg loss 0.00459245, throughput 9.36099K wps
[Epoch 27 Batch 150/162] avg loss 0.00486574, throughput 9.50831K wps
Begin Testing...
[Epoch 27] train avg loss 0.00474982, dev acc 0.9011, dev avg loss 0.262033, throughput 9.45638K wps
[Epoch 28 Batch 30/162] avg loss 0.00475912, throughput 9.6094K wps
[Epoch 28 Batch 60/162] avg loss 0.00464061, throughput 9.58565K wps
[Epoch 28 Batch 90/162] avg loss 0.0046418, throughput 9.37864K wps
[Epoch 28 Batch 120/162] avg loss 0.0047531, throughput 9.31665K wps
[Epoch 28 Batch 150/162] avg loss 0.00459064, throughput 9.51769K wps
Begin Testing...
[Epoch 28] train avg loss 0.00468322, dev acc 0.8967, dev avg loss 0.260006, throughput 9.48099K wps
[Epoch 29 Batch 30/162] avg loss 0.00470761, throughput 9.68041K wps
[Epoch 29 Batch 60/162] avg loss 0.00426105, throughput 9.25475K wps
[Epoch 29 Batch 90/162] avg loss 0.00465676, throughput 9.47957K wps
[Epoch 29 Batch 120/162] avg loss 0.00445583, throughput 9.58242K wps
[Epoch 29 Batch 150/162] avg loss 0.0047887, throughput 9.40571K wps
Begin Testing...
[Epoch 29] train avg loss 0.00459929, dev acc 0.8933, dev avg loss 0.258276, throughput 9.46931K wps
[Epoch 30 Batch 30/162] avg loss 0.00449429, throughput 9.62177K wps
[Epoch 30 Batch 60/162] avg loss 0.00462392, throughput 9.59086K wps
[Epoch 30 Batch 90/162] avg loss 0.00462789, throughput 9.62458K wps
[Epoch 30 Batch 120/162] avg loss 0.00460854, throughput 9.28737K wps
[Epoch 30 Batch 150/162] avg loss 0.0040971, throughput 9.33964K wps
Begin Testing...
[Epoch 30] train avg loss 0.00444876, dev acc 0.8978, dev avg loss 0.256638, throughput 9.4894K wps
[Epoch 31 Batch 30/162] avg loss 0.00422288, throughput 9.65116K wps
[Epoch 31 Batch 60/162] avg loss 0.00437051, throughput 9.3125K wps
[Epoch 31 Batch 90/162] avg loss 0.00449187, throughput 9.41309K wps
[Epoch 31 Batch 120/162] avg loss 0.00488993, throughput 9.22687K wps
[Epoch 31 Batch 150/162] avg loss 0.00421197, throughput 9.34753K wps
Begin Testing...
[Epoch 31] train avg loss 0.00446768, dev acc 0.8933, dev avg loss 0.255506, throughput 9.38925K wps
[Epoch 32 Batch 30/162] avg loss 0.00431977, throughput 9.78082K wps
[Epoch 32 Batch 60/162] avg loss 0.00413386, throughput 9.48972K wps
[Epoch 32 Batch 90/162] avg loss 0.0043643, throughput 9.4618K wps
[Epoch 32 Batch 120/162] avg loss 0.00439172, throughput 9.54022K wps
[Epoch 32 Batch 150/162] avg loss 0.00427458, throughput 9.39454K wps
Begin Testing...
[Epoch 32] train avg loss 0.00432458, dev acc 0.9000, dev avg loss 0.254483, throughput 9.53656K wps
[Epoch 33 Batch 30/162] avg loss 0.00420768, throughput 9.6258K wps
[Epoch 33 Batch 60/162] avg loss 0.0042365, throughput 9.38269K wps
[Epoch 33 Batch 90/162] avg loss 0.00421376, throughput 9.26031K wps
[Epoch 33 Batch 120/162] avg loss 0.00408605, throughput 9.33573K wps
[Epoch 33 Batch 150/162] avg loss 0.004319, throughput 9.51122K wps
Begin Testing...
[Epoch 33] train avg loss 0.00426273, dev acc 0.8933, dev avg loss 0.253094, throughput 9.41099K wps
[Epoch 34 Batch 30/162] avg loss 0.00418267, throughput 9.81242K wps
[Epoch 34 Batch 60/162] avg loss 0.00436134, throughput 9.32125K wps
[Epoch 34 Batch 90/162] avg loss 0.00400369, throughput 9.50111K wps
[Epoch 34 Batch 120/162] avg loss 0.0044739, throughput 9.34642K wps
[Epoch 34 Batch 150/162] avg loss 0.00392229, throughput 9.59992K wps
Begin Testing...
[Epoch 34] train avg loss 0.00420666, dev acc 0.8956, dev avg loss 0.251108, throughput 9.51202K wps
[Epoch 35 Batch 30/162] avg loss 0.00426826, throughput 9.57279K wps
[Epoch 35 Batch 60/162] avg loss 0.00389888, throughput 9.34467K wps
[Epoch 35 Batch 90/162] avg loss 0.00411956, throughput 9.38085K wps
[Epoch 35 Batch 120/162] avg loss 0.0044266, throughput 9.53033K wps
[Epoch 35 Batch 150/162] avg loss 0.00441975, throughput 9.55266K wps
Begin Testing...
[Epoch 35] train avg loss 0.00418664, dev acc 0.8967, dev avg loss 0.25014, throughput 9.44957K wps
[Epoch 36 Batch 30/162] avg loss 0.00400617, throughput 9.6392K wps
[Epoch 36 Batch 60/162] avg loss 0.0043667, throughput 9.27619K wps
[Epoch 36 Batch 90/162] avg loss 0.00426298, throughput 9.37363K wps
[Epoch 36 Batch 120/162] avg loss 0.00398219, throughput 9.37211K wps
[Epoch 36 Batch 150/162] avg loss 0.00413953, throughput 9.49164K wps
Begin Testing...
[Epoch 36] train avg loss 0.00416012, dev acc 0.8978, dev avg loss 0.248519, throughput 9.41985K wps
[Epoch 37 Batch 30/162] avg loss 0.00395778, throughput 9.53247K wps
[Epoch 37 Batch 60/162] avg loss 0.00449907, throughput 9.33843K wps
[Epoch 37 Batch 90/162] avg loss 0.0038147, throughput 9.59091K wps
[Epoch 37 Batch 120/162] avg loss 0.0038871, throughput 9.41651K wps
[Epoch 37 Batch 150/162] avg loss 0.0040264, throughput 9.28328K wps
Begin Testing...
[Epoch 37] train avg loss 0.00403053, dev acc 0.8978, dev avg loss 0.247093, throughput 9.42625K wps
[Epoch 38 Batch 30/162] avg loss 0.00416211, throughput 9.65905K wps
[Epoch 38 Batch 60/162] avg loss 0.00366363, throughput 9.60486K wps
[Epoch 38 Batch 90/162] avg loss 0.00393216, throughput 9.49057K wps
[Epoch 38 Batch 120/162] avg loss 0.00392939, throughput 9.49013K wps
[Epoch 38 Batch 150/162] avg loss 0.00433943, throughput 9.45351K wps
Begin Testing...
[Epoch 38] train avg loss 0.00397535, dev acc 0.8944, dev avg loss 0.24619, throughput 9.54007K wps
[Epoch 39 Batch 30/162] avg loss 0.00396276, throughput 9.49706K wps
[Epoch 39 Batch 60/162] avg loss 0.00376785, throughput 9.48969K wps
[Epoch 39 Batch 90/162] avg loss 0.00369409, throughput 9.52108K wps
[Epoch 39 Batch 120/162] avg loss 0.00424024, throughput 9.52121K wps
[Epoch 39 Batch 150/162] avg loss 0.00385241, throughput 9.39873K wps
Begin Testing...
[Epoch 39] train avg loss 0.00390477, dev acc 0.8978, dev avg loss 0.245064, throughput 9.45901K wps
[Epoch 40 Batch 30/162] avg loss 0.00401372, throughput 9.55509K wps
[Epoch 40 Batch 60/162] avg loss 0.00383773, throughput 9.49152K wps
[Epoch 40 Batch 90/162] avg loss 0.00372368, throughput 9.52907K wps
[Epoch 40 Batch 120/162] avg loss 0.00397325, throughput 9.51281K wps
[Epoch 40 Batch 150/162] avg loss 0.00372412, throughput 9.39829K wps
Begin Testing...
[Epoch 40] train avg loss 0.00385668, dev acc 0.8956, dev avg loss 0.244043, throughput 9.50928K wps
[Epoch 41 Batch 30/162] avg loss 0.00381114, throughput 9.48418K wps
[Epoch 41 Batch 60/162] avg loss 0.00385449, throughput 9.2948K wps
[Epoch 41 Batch 90/162] avg loss 0.00395028, throughput 9.42033K wps
[Epoch 41 Batch 120/162] avg loss 0.00377134, throughput 9.44006K wps
[Epoch 41 Batch 150/162] avg loss 0.00345964, throughput 9.41592K wps
Begin Testing...
[Epoch 41] train avg loss 0.00374949, dev acc 0.9000, dev avg loss 0.242757, throughput 9.40384K wps
[Epoch 42 Batch 30/162] avg loss 0.00340458, throughput 9.80481K wps
[Epoch 42 Batch 60/162] avg loss 0.00381874, throughput 9.5051K wps
[Epoch 42 Batch 90/162] avg loss 0.00415661, throughput 9.51603K wps
[Epoch 42 Batch 120/162] avg loss 0.00360212, throughput 9.34425K wps
[Epoch 42 Batch 150/162] avg loss 0.00351236, throughput 9.48508K wps
Begin Testing...
[Epoch 42] train avg loss 0.00369169, dev acc 0.9022, dev avg loss 0.241956, throughput 9.54115K wps
Observed Improvement.
Begin Testing...
[Epoch 43 Batch 30/162] avg loss 0.00371243, throughput 9.54453K wps
[Epoch 43 Batch 60/162] avg loss 0.00344379, throughput 9.61315K wps
[Epoch 43 Batch 90/162] avg loss 0.00367914, throughput 9.39592K wps
[Epoch 43 Batch 120/162] avg loss 0.0038382, throughput 9.38073K wps
[Epoch 43 Batch 150/162] avg loss 0.00377529, throughput 9.45854K wps
Begin Testing...
[Epoch 43] train avg loss 0.00369587, dev acc 0.8967, dev avg loss 0.241521, throughput 9.48115K wps
[Epoch 44 Batch 30/162] avg loss 0.00341677, throughput 9.63596K wps
[Epoch 44 Batch 60/162] avg loss 0.0037098, throughput 9.2818K wps
[Epoch 44 Batch 90/162] avg loss 0.00383354, throughput 9.50801K wps
[Epoch 44 Batch 120/162] avg loss 0.00351972, throughput 9.50503K wps
[Epoch 44 Batch 150/162] avg loss 0.00347858, throughput 9.40406K wps
Begin Testing...
[Epoch 44] train avg loss 0.00359543, dev acc 0.9022, dev avg loss 0.239972, throughput 9.45154K wps
Observed Improvement.
Begin Testing...
[Epoch 45 Batch 30/162] avg loss 0.00343004, throughput 9.70942K wps
[Epoch 45 Batch 60/162] avg loss 0.00348672, throughput 9.41686K wps
[Epoch 45 Batch 90/162] avg loss 0.00360608, throughput 9.32892K wps
[Epoch 45 Batch 120/162] avg loss 0.00353152, throughput 9.33254K wps
[Epoch 45 Batch 150/162] avg loss 0.00357817, throughput 9.49348K wps
Begin Testing...
[Epoch 45] train avg loss 0.00356873, dev acc 0.9022, dev avg loss 0.238594, throughput 9.45121K wps
Observed Improvement.
Begin Testing...
[Epoch 46 Batch 30/162] avg loss 0.00357299, throughput 9.70748K wps
[Epoch 46 Batch 60/162] avg loss 0.00356739, throughput 9.53977K wps
[Epoch 46 Batch 90/162] avg loss 0.00370183, throughput 9.38615K wps
[Epoch 46 Batch 120/162] avg loss 0.0035029, throughput 9.46609K wps
[Epoch 46 Batch 150/162] avg loss 0.00346978, throughput 9.44931K wps
Begin Testing...
[Epoch 46] train avg loss 0.0035717, dev acc 0.9022, dev avg loss 0.238007, throughput 9.51018K wps
Observed Improvement.
Begin Testing...
[Epoch 47 Batch 30/162] avg loss 0.00336332, throughput 9.62587K wps
[Epoch 47 Batch 60/162] avg loss 0.00335629, throughput 9.52317K wps
[Epoch 47 Batch 90/162] avg loss 0.003664, throughput 9.35081K wps
[Epoch 47 Batch 120/162] avg loss 0.00358625, throughput 9.43895K wps
[Epoch 47 Batch 150/162] avg loss 0.00367986, throughput 9.20644K wps
Begin Testing...
[Epoch 47] train avg loss 0.00352338, dev acc 0.9033, dev avg loss 0.237191, throughput 9.43431K wps
Observed Improvement.
Begin Testing...
[Epoch 48 Batch 30/162] avg loss 0.00338516, throughput 9.58287K wps
[Epoch 48 Batch 60/162] avg loss 0.00329225, throughput 9.45025K wps
[Epoch 48 Batch 90/162] avg loss 0.00346309, throughput 9.43295K wps
[Epoch 48 Batch 120/162] avg loss 0.00329374, throughput 9.54128K wps
[Epoch 48 Batch 150/162] avg loss 0.00339367, throughput 9.48371K wps
Begin Testing...
[Epoch 48] train avg loss 0.0033251, dev acc 0.9033, dev avg loss 0.236375, throughput 9.49072K wps
Observed Improvement.
Begin Testing...
[Epoch 49 Batch 30/162] avg loss 0.00332905, throughput 9.65562K wps
[Epoch 49 Batch 60/162] avg loss 0.00328514, throughput 9.21749K wps
[Epoch 49 Batch 90/162] avg loss 0.00358098, throughput 9.45913K wps
[Epoch 49 Batch 120/162] avg loss 0.00328841, throughput 9.38916K wps
[Epoch 49 Batch 150/162] avg loss 0.00314091, throughput 9.26344K wps
Begin Testing...
[Epoch 49] train avg loss 0.00333849, dev acc 0.9000, dev avg loss 0.236004, throughput 9.41798K wps
[Epoch 50 Batch 30/162] avg loss 0.00324784, throughput 9.53356K wps
[Epoch 50 Batch 60/162] avg loss 0.00338077, throughput 9.37234K wps
[Epoch 50 Batch 90/162] avg loss 0.00332034, throughput 9.55948K wps
[Epoch 50 Batch 120/162] avg loss 0.00320163, throughput 9.65082K wps
[Epoch 50 Batch 150/162] avg loss 0.00367385, throughput 9.48855K wps
Begin Testing...
[Epoch 50] train avg loss 0.00334383, dev acc 0.9033, dev avg loss 0.234703, throughput 9.52477K wps
Observed Improvement.
Begin Testing...
[Epoch 51 Batch 30/162] avg loss 0.00312833, throughput 9.68638K wps
[Epoch 51 Batch 60/162] avg loss 0.00328858, throughput 9.43724K wps
[Epoch 51 Batch 90/162] avg loss 0.00327053, throughput 9.53924K wps
[Epoch 51 Batch 120/162] avg loss 0.00335872, throughput 9.5596K wps
[Epoch 51 Batch 150/162] avg loss 0.00294471, throughput 9.637K wps
Begin Testing...
[Epoch 51] train avg loss 0.00321714, dev acc 0.9044, dev avg loss 0.234071, throughput 9.54202K wps
Observed Improvement.
Begin Testing...
[Epoch 52 Batch 30/162] avg loss 0.00304537, throughput 9.84218K wps
[Epoch 52 Batch 60/162] avg loss 0.00306324, throughput 9.56457K wps
[Epoch 52 Batch 90/162] avg loss 0.00340771, throughput 9.53341K wps
[Epoch 52 Batch 120/162] avg loss 0.00317916, throughput 9.42146K wps
[Epoch 52 Batch 150/162] avg loss 0.00326698, throughput 9.33892K wps
Begin Testing...
[Epoch 52] train avg loss 0.00321891, dev acc 0.9067, dev avg loss 0.233562, throughput 9.54812K wps
Observed Improvement.
Begin Testing...
[Epoch 53 Batch 30/162] avg loss 0.00304132, throughput 9.72912K wps
[Epoch 53 Batch 60/162] avg loss 0.00342059, throughput 9.52993K wps
[Epoch 53 Batch 90/162] avg loss 0.00326972, throughput 9.47436K wps
[Epoch 53 Batch 120/162] avg loss 0.00327138, throughput 9.43446K wps
[Epoch 53 Batch 150/162] avg loss 0.00282514, throughput 9.42306K wps
Begin Testing...
[Epoch 53] train avg loss 0.00317478, dev acc 0.9044, dev avg loss 0.23284, throughput 9.51214K wps
[Epoch 54 Batch 30/162] avg loss 0.00315169, throughput 9.58181K wps
[Epoch 54 Batch 60/162] avg loss 0.00300704, throughput 9.49448K wps
[Epoch 54 Batch 90/162] avg loss 0.00308731, throughput 9.42524K wps
[Epoch 54 Batch 120/162] avg loss 0.00279734, throughput 9.42356K wps
[Epoch 54 Batch 150/162] avg loss 0.00331404, throughput 9.37764K wps
Begin Testing...
[Epoch 54] train avg loss 0.00306876, dev acc 0.9056, dev avg loss 0.231836, throughput 9.44323K wps
[Epoch 55 Batch 30/162] avg loss 0.00281606, throughput 9.68995K wps
[Epoch 55 Batch 60/162] avg loss 0.00324277, throughput 9.27316K wps
[Epoch 55 Batch 90/162] avg loss 0.00333104, throughput 9.30764K wps
[Epoch 55 Batch 120/162] avg loss 0.00307817, throughput 9.54116K wps
[Epoch 55 Batch 150/162] avg loss 0.00302023, throughput 9.36106K wps
Begin Testing...
[Epoch 55] train avg loss 0.00309483, dev acc 0.9044, dev avg loss 0.233717, throughput 9.42232K wps
[Epoch 56 Batch 30/162] avg loss 0.00311346, throughput 9.51313K wps
[Epoch 56 Batch 60/162] avg loss 0.00301776, throughput 9.60004K wps
[Epoch 56 Batch 90/162] avg loss 0.00320084, throughput 9.50125K wps
[Epoch 56 Batch 120/162] avg loss 0.00278142, throughput 9.2911K wps
[Epoch 56 Batch 150/162] avg loss 0.00314581, throughput 9.31839K wps
Begin Testing...
[Epoch 56] train avg loss 0.00306238, dev acc 0.9056, dev avg loss 0.230773, throughput 9.43364K wps
[Epoch 57 Batch 30/162] avg loss 0.00300886, throughput 9.65841K wps
[Epoch 57 Batch 60/162] avg loss 0.00288217, throughput 9.417K wps
[Epoch 57 Batch 90/162] avg loss 0.00308621, throughput 9.37019K wps
[Epoch 57 Batch 120/162] avg loss 0.00316749, throughput 9.26365K wps
[Epoch 57 Batch 150/162] avg loss 0.00281351, throughput 9.36883K wps
Begin Testing...
[Epoch 57] train avg loss 0.00299467, dev acc 0.9067, dev avg loss 0.230207, throughput 9.40837K wps
Observed Improvement.
Begin Testing...
[Epoch 58 Batch 30/162] avg loss 0.00299431, throughput 9.76199K wps
[Epoch 58 Batch 60/162] avg loss 0.00282411, throughput 9.50064K wps
[Epoch 58 Batch 90/162] avg loss 0.00295165, throughput 9.59821K wps
[Epoch 58 Batch 120/162] avg loss 0.00279171, throughput 9.53881K wps
[Epoch 58 Batch 150/162] avg loss 0.00304129, throughput 9.55481K wps
Begin Testing...
[Epoch 58] train avg loss 0.00291889, dev acc 0.9067, dev avg loss 0.229415, throughput 9.58692K wps
Observed Improvement.
Begin Testing...
[Epoch 59 Batch 30/162] avg loss 0.00292177, throughput 9.77472K wps
[Epoch 59 Batch 60/162] avg loss 0.00281067, throughput 9.31075K wps
[Epoch 59 Batch 90/162] avg loss 0.00298035, throughput 9.32372K wps
[Epoch 59 Batch 120/162] avg loss 0.00286139, throughput 9.32079K wps
[Epoch 59 Batch 150/162] avg loss 0.00287484, throughput 9.4594K wps
Begin Testing...
[Epoch 59] train avg loss 0.00291011, dev acc 0.9033, dev avg loss 0.229813, throughput 9.43401K wps
[Epoch 60 Batch 30/162] avg loss 0.00274751, throughput 9.65562K wps
[Epoch 60 Batch 60/162] avg loss 0.00297057, throughput 9.29029K wps
[Epoch 60 Batch 90/162] avg loss 0.00309006, throughput 9.47881K wps
[Epoch 60 Batch 120/162] avg loss 0.00273039, throughput 9.34194K wps
[Epoch 60 Batch 150/162] avg loss 0.00280782, throughput 9.45636K wps
Begin Testing...
[Epoch 60] train avg loss 0.00287456, dev acc 0.9078, dev avg loss 0.228875, throughput 9.42405K wps
Observed Improvement.
Begin Testing...
[Epoch 61 Batch 30/162] avg loss 0.00285105, throughput 9.53147K wps
[Epoch 61 Batch 60/162] avg loss 0.00292641, throughput 9.29196K wps
[Epoch 61 Batch 90/162] avg loss 0.00265302, throughput 9.4929K wps
[Epoch 61 Batch 120/162] avg loss 0.00280257, throughput 9.43319K wps
[Epoch 61 Batch 150/162] avg loss 0.0028003, throughput 9.395K wps
Begin Testing...
[Epoch 61] train avg loss 0.00279155, dev acc 0.9100, dev avg loss 0.227657, throughput 9.43359K wps
Observed Improvement.
Begin Testing...
[Epoch 62 Batch 30/162] avg loss 0.00287637, throughput 9.66193K wps
[Epoch 62 Batch 60/162] avg loss 0.00262742, throughput 9.39944K wps
[Epoch 62 Batch 90/162] avg loss 0.00280082, throughput 9.52576K wps
[Epoch 62 Batch 120/162] avg loss 0.00263081, throughput 9.43111K wps
[Epoch 62 Batch 150/162] avg loss 0.00267903, throughput 9.40929K wps
Begin Testing...
[Epoch 62] train avg loss 0.00275114, dev acc 0.9078, dev avg loss 0.227671, throughput 9.47403K wps
[Epoch 63 Batch 30/162] avg loss 0.00271683, throughput 9.55298K wps
[Epoch 63 Batch 60/162] avg loss 0.00282988, throughput 9.59907K wps
[Epoch 63 Batch 90/162] avg loss 0.00269918, throughput 9.41993K wps
[Epoch 63 Batch 120/162] avg loss 0.00271857, throughput 9.40832K wps
[Epoch 63 Batch 150/162] avg loss 0.00271426, throughput 9.54929K wps
Begin Testing...
[Epoch 63] train avg loss 0.00273382, dev acc 0.9044, dev avg loss 0.227331, throughput 9.51732K wps
[Epoch 64 Batch 30/162] avg loss 0.00260102, throughput 9.58995K wps
[Epoch 64 Batch 60/162] avg loss 0.00257094, throughput 9.5072K wps
[Epoch 64 Batch 90/162] avg loss 0.00288627, throughput 9.44931K wps
[Epoch 64 Batch 120/162] avg loss 0.0028859, throughput 9.63203K wps
[Epoch 64 Batch 150/162] avg loss 0.00263134, throughput 9.59836K wps
Begin Testing...
[Epoch 64] train avg loss 0.0027179, dev acc 0.9089, dev avg loss 0.225894, throughput 9.53597K wps
[Epoch 65 Batch 30/162] avg loss 0.00245004, throughput 9.53539K wps
[Epoch 65 Batch 60/162] avg loss 0.00275061, throughput 9.36059K wps
[Epoch 65 Batch 90/162] avg loss 0.00252406, throughput 9.33963K wps
[Epoch 65 Batch 120/162] avg loss 0.00252005, throughput 9.43262K wps
[Epoch 65 Batch 150/162] avg loss 0.00265348, throughput 9.56578K wps
Begin Testing...
[Epoch 65] train avg loss 0.00263024, dev acc 0.9089, dev avg loss 0.226352, throughput 9.44429K wps
[Epoch 66 Batch 30/162] avg loss 0.00261724, throughput 9.62259K wps
[Epoch 66 Batch 60/162] avg loss 0.00258213, throughput 9.47851K wps
[Epoch 66 Batch 90/162] avg loss 0.00244424, throughput 9.54002K wps
[Epoch 66 Batch 120/162] avg loss 0.00264092, throughput 9.4652K wps
[Epoch 66 Batch 150/162] avg loss 0.00260729, throughput 9.34521K wps
Begin Testing...
[Epoch 66] train avg loss 0.00258222, dev acc 0.9056, dev avg loss 0.226091, throughput 9.49081K wps
[Epoch 67 Batch 30/162] avg loss 0.00253401, throughput 9.70008K wps
[Epoch 67 Batch 60/162] avg loss 0.0025713, throughput 9.4671K wps
[Epoch 67 Batch 90/162] avg loss 0.00263792, throughput 9.58472K wps
[Epoch 67 Batch 120/162] avg loss 0.00259544, throughput 9.35013K wps
[Epoch 67 Batch 150/162] avg loss 0.00253625, throughput 9.56424K wps
Begin Testing...
[Epoch 67] train avg loss 0.00258191, dev acc 0.9056, dev avg loss 0.224853, throughput 9.53606K wps
[Epoch 68 Batch 30/162] avg loss 0.00262541, throughput 9.45497K wps
[Epoch 68 Batch 60/162] avg loss 0.00223576, throughput 9.50722K wps
[Epoch 68 Batch 90/162] avg loss 0.00265467, throughput 9.48961K wps
[Epoch 68 Batch 120/162] avg loss 0.0026805, throughput 9.35227K wps
[Epoch 68 Batch 150/162] avg loss 0.00247413, throughput 9.34828K wps
Begin Testing...
[Epoch 68] train avg loss 0.00250969, dev acc 0.9089, dev avg loss 0.224803, throughput 9.42636K wps
[Epoch 69 Batch 30/162] avg loss 0.00234652, throughput 9.61719K wps
[Epoch 69 Batch 60/162] avg loss 0.00249058, throughput 9.32232K wps
[Epoch 69 Batch 90/162] avg loss 0.00265199, throughput 9.58094K wps
[Epoch 69 Batch 120/162] avg loss 0.00250144, throughput 9.56603K wps
[Epoch 69 Batch 150/162] avg loss 0.00244254, throughput 9.30794K wps
Begin Testing...
[Epoch 69] train avg loss 0.00252046, dev acc 0.9078, dev avg loss 0.225191, throughput 9.48686K wps
[Epoch 70 Batch 30/162] avg loss 0.0027333, throughput 9.76137K wps
[Epoch 70 Batch 60/162] avg loss 0.00234354, throughput 9.39371K wps
[Epoch 70 Batch 90/162] avg loss 0.00244186, throughput 9.41765K wps
[Epoch 70 Batch 120/162] avg loss 0.00249614, throughput 9.45938K wps
[Epoch 70 Batch 150/162] avg loss 0.00242335, throughput 9.40657K wps
Begin Testing...
[Epoch 70] train avg loss 0.00249483, dev acc 0.9067, dev avg loss 0.223972, throughput 9.48917K wps
[Epoch 71 Batch 30/162] avg loss 0.00265116, throughput 9.70562K wps
[Epoch 71 Batch 60/162] avg loss 0.00219598, throughput 9.5973K wps
[Epoch 71 Batch 90/162] avg loss 0.00264071, throughput 9.33711K wps
[Epoch 71 Batch 120/162] avg loss 0.00249406, throughput 9.29141K wps
[Epoch 71 Batch 150/162] avg loss 0.00230302, throughput 9.54167K wps
Begin Testing...
[Epoch 71] train avg loss 0.00246156, dev acc 0.9078, dev avg loss 0.223821, throughput 9.47262K wps
[Epoch 72 Batch 30/162] avg loss 0.00233132, throughput 9.60078K wps
[Epoch 72 Batch 60/162] avg loss 0.00244902, throughput 9.48397K wps
[Epoch 72 Batch 90/162] avg loss 0.00248208, throughput 9.49046K wps
[Epoch 72 Batch 120/162] avg loss 0.00227552, throughput 9.26927K wps
[Epoch 72 Batch 150/162] avg loss 0.00225872, throughput 9.28998K wps
Begin Testing...
[Epoch 72] train avg loss 0.00237168, dev acc 0.9078, dev avg loss 0.223543, throughput 9.4302K wps
[Epoch 73 Batch 30/162] avg loss 0.00263185, throughput 9.81466K wps
[Epoch 73 Batch 60/162] avg loss 0.00225019, throughput 9.23833K wps
[Epoch 73 Batch 90/162] avg loss 0.00229392, throughput 9.47233K wps
[Epoch 73 Batch 120/162] avg loss 0.00245663, throughput 9.32297K wps
[Epoch 73 Batch 150/162] avg loss 0.00234621, throughput 9.24492K wps
Begin Testing...
[Epoch 73] train avg loss 0.0023965, dev acc 0.9078, dev avg loss 0.223035, throughput 9.40381K wps
[Epoch 74 Batch 30/162] avg loss 0.00243304, throughput 9.53845K wps
[Epoch 74 Batch 60/162] avg loss 0.00245765, throughput 9.49401K wps
[Epoch 74 Batch 90/162] avg loss 0.00222335, throughput 9.36038K wps
[Epoch 74 Batch 120/162] avg loss 0.00222151, throughput 9.24534K wps
[Epoch 74 Batch 150/162] avg loss 0.00242142, throughput 9.33622K wps
Begin Testing...
[Epoch 74] train avg loss 0.00238068, dev acc 0.9089, dev avg loss 0.223539, throughput 9.38827K wps
[Epoch 75 Batch 30/162] avg loss 0.00232959, throughput 9.71683K wps
[Epoch 75 Batch 60/162] avg loss 0.00231906, throughput 9.48012K wps
[Epoch 75 Batch 90/162] avg loss 0.00207578, throughput 9.25903K wps
[Epoch 75 Batch 120/162] avg loss 0.00230599, throughput 9.59186K wps
[Epoch 75 Batch 150/162] avg loss 0.00255884, throughput 9.26851K wps
Begin Testing...
[Epoch 75] train avg loss 0.0023027, dev acc 0.9100, dev avg loss 0.222599, throughput 9.45505K wps
Observed Improvement.
Begin Testing...
[Epoch 76 Batch 30/162] avg loss 0.00231372, throughput 9.47676K wps
[Epoch 76 Batch 60/162] avg loss 0.00242213, throughput 9.34452K wps
[Epoch 76 Batch 90/162] avg loss 0.00227477, throughput 9.56446K wps
[Epoch 76 Batch 120/162] avg loss 0.00211569, throughput 9.60293K wps
[Epoch 76 Batch 150/162] avg loss 0.0024557, throughput 9.31244K wps
Begin Testing...
[Epoch 76] train avg loss 0.00231715, dev acc 0.9078, dev avg loss 0.221847, throughput 9.47437K wps
[Epoch 77 Batch 30/162] avg loss 0.00209009, throughput 9.78129K wps
[Epoch 77 Batch 60/162] avg loss 0.00226566, throughput 9.35508K wps
[Epoch 77 Batch 90/162] avg loss 0.00228979, throughput 9.37349K wps
[Epoch 77 Batch 120/162] avg loss 0.00229224, throughput 9.48541K wps
[Epoch 77 Batch 150/162] avg loss 0.00226442, throughput 9.38571K wps
Begin Testing...
[Epoch 77] train avg loss 0.00224203, dev acc 0.9111, dev avg loss 0.222371, throughput 9.45031K wps
Observed Improvement.
Begin Testing...
[Epoch 78 Batch 30/162] avg loss 0.00234424, throughput 9.58064K wps
[Epoch 78 Batch 60/162] avg loss 0.00235085, throughput 9.37106K wps
[Epoch 78 Batch 90/162] avg loss 0.00209641, throughput 9.40242K wps
[Epoch 78 Batch 120/162] avg loss 0.00222298, throughput 9.56422K wps
[Epoch 78 Batch 150/162] avg loss 0.00207928, throughput 9.4193K wps
Begin Testing...
[Epoch 78] train avg loss 0.00221071, dev acc 0.9100, dev avg loss 0.222119, throughput 9.47201K wps
[Epoch 79 Batch 30/162] avg loss 0.00235271, throughput 9.52035K wps
[Epoch 79 Batch 60/162] avg loss 0.00209923, throughput 9.40336K wps
[Epoch 79 Batch 90/162] avg loss 0.0019856, throughput 9.29455K wps
[Epoch 79 Batch 120/162] avg loss 0.00214341, throughput 9.31275K wps
[Epoch 79 Batch 150/162] avg loss 0.00233086, throughput 9.4199K wps
Begin Testing...
[Epoch 79] train avg loss 0.0021831, dev acc 0.9089, dev avg loss 0.221286, throughput 9.37384K wps
[Epoch 80 Batch 30/162] avg loss 0.00212989, throughput 9.66752K wps
[Epoch 80 Batch 60/162] avg loss 0.00233351, throughput 9.42017K wps
[Epoch 80 Batch 90/162] avg loss 0.00204925, throughput 9.53371K wps
[Epoch 80 Batch 120/162] avg loss 0.00208745, throughput 9.50871K wps
[Epoch 80 Batch 150/162] avg loss 0.00204086, throughput 9.30398K wps
Begin Testing...
[Epoch 80] train avg loss 0.00213919, dev acc 0.9078, dev avg loss 0.221146, throughput 9.4751K wps
[Epoch 81 Batch 30/162] avg loss 0.00222365, throughput 9.53221K wps
[Epoch 81 Batch 60/162] avg loss 0.00195803, throughput 9.48797K wps
[Epoch 81 Batch 90/162] avg loss 0.00210109, throughput 9.4905K wps
[Epoch 81 Batch 120/162] avg loss 0.00215135, throughput 9.40405K wps
[Epoch 81 Batch 150/162] avg loss 0.00189482, throughput 9.43953K wps
Begin Testing...
[Epoch 81] train avg loss 0.00208873, dev acc 0.9100, dev avg loss 0.22161, throughput 9.45935K wps
[Epoch 82 Batch 30/162] avg loss 0.00216291, throughput 9.70562K wps
[Epoch 82 Batch 60/162] avg loss 0.00205441, throughput 9.41718K wps
[Epoch 82 Batch 90/162] avg loss 0.00215618, throughput 9.35488K wps
[Epoch 82 Batch 120/162] avg loss 0.00202197, throughput 9.35648K wps
[Epoch 82 Batch 150/162] avg loss 0.0019107, throughput 9.6142K wps
Begin Testing...
[Epoch 82] train avg loss 0.00204379, dev acc 0.9089, dev avg loss 0.22071, throughput 9.48537K wps
[Epoch 83 Batch 30/162] avg loss 0.00222992, throughput 9.45501K wps
[Epoch 83 Batch 60/162] avg loss 0.0018562, throughput 9.42028K wps
[Epoch 83 Batch 90/162] avg loss 0.00232759, throughput 9.22708K wps
[Epoch 83 Batch 120/162] avg loss 0.00210041, throughput 9.30703K wps
[Epoch 83 Batch 150/162] avg loss 0.0020873, throughput 9.36673K wps
Begin Testing...
[Epoch 83] train avg loss 0.00209961, dev acc 0.9089, dev avg loss 0.220238, throughput 9.3445K wps
[Epoch 84 Batch 30/162] avg loss 0.00193394, throughput 9.49083K wps
[Epoch 84 Batch 60/162] avg loss 0.00178724, throughput 9.42035K wps
[Epoch 84 Batch 90/162] avg loss 0.002191, throughput 9.42442K wps
[Epoch 84 Batch 120/162] avg loss 0.00190365, throughput 9.56577K wps
[Epoch 84 Batch 150/162] avg loss 0.00216141, throughput 9.55071K wps
Begin Testing...
[Epoch 84] train avg loss 0.00200584, dev acc 0.9089, dev avg loss 0.221608, throughput 9.50437K wps
[Epoch 85 Batch 30/162] avg loss 0.00200855, throughput 9.66676K wps
[Epoch 85 Batch 60/162] avg loss 0.00228305, throughput 9.41849K wps
[Epoch 85 Batch 90/162] avg loss 0.00196109, throughput 9.47651K wps
[Epoch 85 Batch 120/162] avg loss 0.00201678, throughput 9.46363K wps
[Epoch 85 Batch 150/162] avg loss 0.00191863, throughput 9.54773K wps
Begin Testing...
[Epoch 85] train avg loss 0.00204643, dev acc 0.9100, dev avg loss 0.220292, throughput 9.50108K wps
[Epoch 86 Batch 30/162] avg loss 0.00213809, throughput 9.60812K wps
[Epoch 86 Batch 60/162] avg loss 0.00216389, throughput 9.28713K wps
[Epoch 86 Batch 90/162] avg loss 0.00181319, throughput 9.56062K wps
[Epoch 86 Batch 120/162] avg loss 0.00190878, throughput 9.55331K wps
[Epoch 86 Batch 150/162] avg loss 0.0020424, throughput 9.34361K wps
Begin Testing...
[Epoch 86] train avg loss 0.00201799, dev acc 0.9078, dev avg loss 0.219585, throughput 9.45225K wps
[Epoch 87 Batch 30/162] avg loss 0.0017795, throughput 9.56219K wps
[Epoch 87 Batch 60/162] avg loss 0.00199864, throughput 9.40179K wps
[Epoch 87 Batch 90/162] avg loss 0.00214422, throughput 9.38433K wps
[Epoch 87 Batch 120/162] avg loss 0.00171401, throughput 9.47148K wps
[Epoch 87 Batch 150/162] avg loss 0.00192982, throughput 9.31495K wps
Begin Testing...
[Epoch 87] train avg loss 0.00193888, dev acc 0.9122, dev avg loss 0.2195, throughput 9.42037K wps
Observed Improvement.
Begin Testing...
[Epoch 88 Batch 30/162] avg loss 0.00191884, throughput 9.58688K wps
[Epoch 88 Batch 60/162] avg loss 0.00176907, throughput 9.21836K wps
[Epoch 88 Batch 90/162] avg loss 0.00200286, throughput 9.33015K wps
[Epoch 88 Batch 120/162] avg loss 0.00210091, throughput 9.3336K wps
[Epoch 88 Batch 150/162] avg loss 0.00207507, throughput 9.48237K wps
Begin Testing...
[Epoch 88] train avg loss 0.00198079, dev acc 0.9089, dev avg loss 0.220665, throughput 9.39067K wps
[Epoch 89 Batch 30/162] avg loss 0.00184786, throughput 9.5012K wps
[Epoch 89 Batch 60/162] avg loss 0.00212888, throughput 9.42002K wps
[Epoch 89 Batch 90/162] avg loss 0.00199185, throughput 9.27916K wps
[Epoch 89 Batch 120/162] avg loss 0.00176154, throughput 9.51382K wps
[Epoch 89 Batch 150/162] avg loss 0.00172582, throughput 9.41007K wps
Begin Testing...
[Epoch 89] train avg loss 0.00188618, dev acc 0.9089, dev avg loss 0.219125, throughput 9.41021K wps
[Epoch 90 Batch 30/162] avg loss 0.00182844, throughput 9.78588K wps
[Epoch 90 Batch 60/162] avg loss 0.00191285, throughput 9.37479K wps
[Epoch 90 Batch 90/162] avg loss 0.00192626, throughput 9.52528K wps
[Epoch 90 Batch 120/162] avg loss 0.00185552, throughput 9.56085K wps
[Epoch 90 Batch 150/162] avg loss 0.00185076, throughput 9.56648K wps
Begin Testing...
[Epoch 90] train avg loss 0.00189062, dev acc 0.9089, dev avg loss 0.219254, throughput 9.54624K wps
[Epoch 91 Batch 30/162] avg loss 0.00195068, throughput 9.50265K wps
[Epoch 91 Batch 60/162] avg loss 0.00160475, throughput 9.34737K wps
[Epoch 91 Batch 90/162] avg loss 0.00200673, throughput 9.32374K wps
[Epoch 91 Batch 120/162] avg loss 0.00181465, throughput 9.41979K wps
[Epoch 91 Batch 150/162] avg loss 0.00189604, throughput 9.41508K wps
Begin Testing...
[Epoch 91] train avg loss 0.00186969, dev acc 0.9111, dev avg loss 0.21887, throughput 9.38969K wps
[Epoch 92 Batch 30/162] avg loss 0.00181704, throughput 9.77774K wps
[Epoch 92 Batch 60/162] avg loss 0.00161336, throughput 9.34604K wps
[Epoch 92 Batch 90/162] avg loss 0.00201436, throughput 9.3396K wps
[Epoch 92 Batch 120/162] avg loss 0.00207533, throughput 9.39426K wps
[Epoch 92 Batch 150/162] avg loss 0.00172785, throughput 9.29946K wps
Begin Testing...
[Epoch 92] train avg loss 0.0018499, dev acc 0.9111, dev avg loss 0.219325, throughput 9.4302K wps
[Epoch 93 Batch 30/162] avg loss 0.00182003, throughput 9.67273K wps
[Epoch 93 Batch 60/162] avg loss 0.00173343, throughput 9.35819K wps
[Epoch 93 Batch 90/162] avg loss 0.00174688, throughput 9.31267K wps
[Epoch 93 Batch 120/162] avg loss 0.00175878, throughput 9.35214K wps
[Epoch 93 Batch 150/162] avg loss 0.00187643, throughput 9.30837K wps
Begin Testing...
[Epoch 93] train avg loss 0.00178763, dev acc 0.9111, dev avg loss 0.219115, throughput 9.3952K wps
[Epoch 94 Batch 30/162] avg loss 0.00190417, throughput 9.55282K wps
[Epoch 94 Batch 60/162] avg loss 0.00160619, throughput 9.3391K wps
[Epoch 94 Batch 90/162] avg loss 0.0019931, throughput 9.62708K wps
[Epoch 94 Batch 120/162] avg loss 0.00181494, throughput 9.46653K wps
[Epoch 94 Batch 150/162] avg loss 0.001738, throughput 9.27947K wps
Begin Testing...
[Epoch 94] train avg loss 0.00182563, dev acc 0.9100, dev avg loss 0.218182, throughput 9.45202K wps
[Epoch 95 Batch 30/162] avg loss 0.00182958, throughput 9.45947K wps
[Epoch 95 Batch 60/162] avg loss 0.00184266, throughput 9.3421K wps
[Epoch 95 Batch 90/162] avg loss 0.00182976, throughput 9.25142K wps
[Epoch 95 Batch 120/162] avg loss 0.00162568, throughput 9.41526K wps
[Epoch 95 Batch 150/162] avg loss 0.00173006, throughput 9.52948K wps
Begin Testing...
[Epoch 95] train avg loss 0.00177541, dev acc 0.9089, dev avg loss 0.218616, throughput 9.4042K wps
[Epoch 96 Batch 30/162] avg loss 0.00173373, throughput 9.46913K wps
[Epoch 96 Batch 60/162] avg loss 0.00185007, throughput 9.39829K wps
[Epoch 96 Batch 90/162] avg loss 0.00174506, throughput 9.59549K wps
[Epoch 96 Batch 120/162] avg loss 0.00166967, throughput 9.47551K wps
[Epoch 96 Batch 150/162] avg loss 0.00190528, throughput 9.3913K wps
Begin Testing...
[Epoch 96] train avg loss 0.00178073, dev acc 0.9111, dev avg loss 0.2181, throughput 9.46895K wps
[Epoch 97 Batch 30/162] avg loss 0.00163419, throughput 9.6564K wps
[Epoch 97 Batch 60/162] avg loss 0.00172592, throughput 9.64543K wps
[Epoch 97 Batch 90/162] avg loss 0.00159437, throughput 9.39163K wps
[Epoch 97 Batch 120/162] avg loss 0.00157453, throughput 9.46961K wps
[Epoch 97 Batch 150/162] avg loss 0.00160716, throughput 9.28141K wps
Begin Testing...
[Epoch 97] train avg loss 0.00164425, dev acc 0.9100, dev avg loss 0.218739, throughput 9.47423K wps
[Epoch 98 Batch 30/162] avg loss 0.00154644, throughput 9.67402K wps
[Epoch 98 Batch 60/162] avg loss 0.00161139, throughput 9.55952K wps
[Epoch 98 Batch 90/162] avg loss 0.00163716, throughput 9.29929K wps
[Epoch 98 Batch 120/162] avg loss 0.00199379, throughput 9.45583K wps
[Epoch 98 Batch 150/162] avg loss 0.0015358, throughput 9.58802K wps
Begin Testing...
[Epoch 98] train avg loss 0.00167753, dev acc 0.9089, dev avg loss 0.218332, throughput 9.50315K wps
[Epoch 99 Batch 30/162] avg loss 0.0017181, throughput 9.66166K wps
[Epoch 99 Batch 60/162] avg loss 0.0014925, throughput 9.65079K wps
[Epoch 99 Batch 90/162] avg loss 0.00169776, throughput 9.36586K wps
[Epoch 99 Batch 120/162] avg loss 0.00167437, throughput 9.34196K wps
[Epoch 99 Batch 150/162] avg loss 0.00164303, throughput 9.4401K wps
Begin Testing...
[Epoch 99] train avg loss 0.00163912, dev acc 0.9089, dev avg loss 0.21797, throughput 9.47254K wps
[Epoch 100 Batch 30/162] avg loss 0.00172311, throughput 9.69474K wps
[Epoch 100 Batch 60/162] avg loss 0.0015106, throughput 9.37041K wps
[Epoch 100 Batch 90/162] avg loss 0.00171667, throughput 9.36718K wps
[Epoch 100 Batch 120/162] avg loss 0.00158198, throughput 9.57183K wps
[Epoch 100 Batch 150/162] avg loss 0.00160873, throughput 9.41018K wps
Begin Testing...
[Epoch 100] train avg loss 0.00163292, dev acc 0.9111, dev avg loss 0.218249, throughput 9.469K wps
[Epoch 101 Batch 30/162] avg loss 0.00169912, throughput 9.67916K wps
[Epoch 101 Batch 60/162] avg loss 0.00162226, throughput 9.41336K wps
[Epoch 101 Batch 90/162] avg loss 0.00161771, throughput 9.47554K wps
[Epoch 101 Batch 120/162] avg loss 0.00156766, throughput 9.45444K wps
[Epoch 101 Batch 150/162] avg loss 0.00160106, throughput 9.42783K wps
Begin Testing...
[Epoch 101] train avg loss 0.00161875, dev acc 0.9122, dev avg loss 0.21838, throughput 9.48707K wps
Observed Improvement.
Begin Testing...
[Epoch 102 Batch 30/162] avg loss 0.00164463, throughput 9.71282K wps
[Epoch 102 Batch 60/162] avg loss 0.00163987, throughput 9.4949K wps
[Epoch 102 Batch 90/162] avg loss 0.00187043, throughput 9.49972K wps
[Epoch 102 Batch 120/162] avg loss 0.00173755, throughput 9.55358K wps
[Epoch 102 Batch 150/162] avg loss 0.00167217, throughput 9.47727K wps
Begin Testing...
[Epoch 102] train avg loss 0.0016931, dev acc 0.9089, dev avg loss 0.217314, throughput 9.54381K wps
[Epoch 103 Batch 30/162] avg loss 0.00162459, throughput 9.8075K wps
[Epoch 103 Batch 60/162] avg loss 0.0014097, throughput 9.6176K wps
[Epoch 103 Batch 90/162] avg loss 0.00184504, throughput 9.6187K wps
[Epoch 103 Batch 120/162] avg loss 0.00163595, throughput 9.39052K wps
[Epoch 103 Batch 150/162] avg loss 0.0013957, throughput 9.4538K wps
Begin Testing...
[Epoch 103] train avg loss 0.00158749, dev acc 0.9089, dev avg loss 0.217954, throughput 9.56766K wps
[Epoch 104 Batch 30/162] avg loss 0.00145649, throughput 9.62778K wps
[Epoch 104 Batch 60/162] avg loss 0.00160015, throughput 9.24468K wps
[Epoch 104 Batch 90/162] avg loss 0.00156663, throughput 9.4886K wps
[Epoch 104 Batch 120/162] avg loss 0.00159552, throughput 9.35882K wps
[Epoch 104 Batch 150/162] avg loss 0.00161709, throughput 9.39604K wps
Begin Testing...
[Epoch 104] train avg loss 0.00157035, dev acc 0.9100, dev avg loss 0.217412, throughput 9.42192K wps
[Epoch 105 Batch 30/162] avg loss 0.00139521, throughput 9.60093K wps
[Epoch 105 Batch 60/162] avg loss 0.00161099, throughput 9.50275K wps
[Epoch 105 Batch 90/162] avg loss 0.00133284, throughput 9.52292K wps
[Epoch 105 Batch 120/162] avg loss 0.00166887, throughput 9.37K wps
[Epoch 105 Batch 150/162] avg loss 0.00165047, throughput 9.45376K wps
Begin Testing...
[Epoch 105] train avg loss 0.00153814, dev acc 0.9100, dev avg loss 0.218134, throughput 9.49395K wps
[Epoch 106 Batch 30/162] avg loss 0.00155629, throughput 9.6182K wps
[Epoch 106 Batch 60/162] avg loss 0.00153939, throughput 9.29398K wps
[Epoch 106 Batch 90/162] avg loss 0.00165987, throughput 9.41273K wps
[Epoch 106 Batch 120/162] avg loss 0.00137204, throughput 9.43521K wps
[Epoch 106 Batch 150/162] avg loss 0.00163072, throughput 9.33766K wps
Begin Testing...
[Epoch 106] train avg loss 0.00155549, dev acc 0.9100, dev avg loss 0.218352, throughput 9.42299K wps
[Epoch 107 Batch 30/162] avg loss 0.00146346, throughput 9.55194K wps
[Epoch 107 Batch 60/162] avg loss 0.00158311, throughput 9.4073K wps
[Epoch 107 Batch 90/162] avg loss 0.00161533, throughput 9.39201K wps
[Epoch 107 Batch 120/162] avg loss 0.0015741, throughput 9.58495K wps
[Epoch 107 Batch 150/162] avg loss 0.00146978, throughput 9.56825K wps
Begin Testing...
[Epoch 107] train avg loss 0.00154305, dev acc 0.9089, dev avg loss 0.217034, throughput 9.48997K wps
[Epoch 108 Batch 30/162] avg loss 0.00150228, throughput 9.43922K wps
[Epoch 108 Batch 60/162] avg loss 0.00154515, throughput 9.42667K wps
[Epoch 108 Batch 90/162] avg loss 0.00152148, throughput 9.68687K wps
[Epoch 108 Batch 120/162] avg loss 0.00140496, throughput 9.26175K wps
[Epoch 108 Batch 150/162] avg loss 0.0016133, throughput 9.51938K wps
Begin Testing...
[Epoch 108] train avg loss 0.00150288, dev acc 0.9089, dev avg loss 0.217453, throughput 9.45065K wps
[Epoch 109 Batch 30/162] avg loss 0.00165414, throughput 9.59127K wps
[Epoch 109 Batch 60/162] avg loss 0.00159648, throughput 9.57575K wps
[Epoch 109 Batch 90/162] avg loss 0.00152362, throughput 9.63366K wps
[Epoch 109 Batch 120/162] avg loss 0.00152375, throughput 9.39203K wps
[Epoch 109 Batch 150/162] avg loss 0.00144737, throughput 9.46188K wps
Begin Testing...
[Epoch 109] train avg loss 0.00153026, dev acc 0.9100, dev avg loss 0.217437, throughput 9.51891K wps
[Epoch 110 Batch 30/162] avg loss 0.0015049, throughput 9.72175K wps
[Epoch 110 Batch 60/162] avg loss 0.00136787, throughput 9.48326K wps
[Epoch 110 Batch 90/162] avg loss 0.00142602, throughput 9.43564K wps
[Epoch 110 Batch 120/162] avg loss 0.00152032, throughput 9.40788K wps
[Epoch 110 Batch 150/162] avg loss 0.00144865, throughput 9.39506K wps
Begin Testing...
[Epoch 110] train avg loss 0.00145118, dev acc 0.9100, dev avg loss 0.21718, throughput 9.46977K wps
[Epoch 111 Batch 30/162] avg loss 0.00156247, throughput 9.75965K wps
[Epoch 111 Batch 60/162] avg loss 0.00135436, throughput 9.3549K wps
[Epoch 111 Batch 90/162] avg loss 0.00139131, throughput 9.40562K wps
[Epoch 111 Batch 120/162] avg loss 0.0012776, throughput 9.42576K wps
[Epoch 111 Batch 150/162] avg loss 0.00149204, throughput 9.3029K wps
Begin Testing...
[Epoch 111] train avg loss 0.00142168, dev acc 0.9089, dev avg loss 0.217852, throughput 9.43446K wps
[Epoch 112 Batch 30/162] avg loss 0.00130094, throughput 9.66906K wps
[Epoch 112 Batch 60/162] avg loss 0.00142185, throughput 9.35126K wps
[Epoch 112 Batch 90/162] avg loss 0.00145422, throughput 9.31391K wps
[Epoch 112 Batch 120/162] avg loss 0.00159254, throughput 9.50384K wps
[Epoch 112 Batch 150/162] avg loss 0.0014078, throughput 9.2938K wps
Begin Testing...
[Epoch 112] train avg loss 0.00145227, dev acc 0.9100, dev avg loss 0.217085, throughput 9.43721K wps
[Epoch 113 Batch 30/162] avg loss 0.00133559, throughput 9.677K wps
[Epoch 113 Batch 60/162] avg loss 0.00131785, throughput 9.25178K wps
[Epoch 113 Batch 90/162] avg loss 0.00166246, throughput 9.46375K wps
[Epoch 113 Batch 120/162] avg loss 0.00129416, throughput 9.30584K wps
[Epoch 113 Batch 150/162] avg loss 0.00136864, throughput 9.46217K wps
Begin Testing...
[Epoch 113] train avg loss 0.00138733, dev acc 0.9111, dev avg loss 0.217175, throughput 9.42529K wps
[Epoch 114 Batch 30/162] avg loss 0.00140878, throughput 9.72216K wps
[Epoch 114 Batch 60/162] avg loss 0.00143006, throughput 9.35059K wps
[Epoch 114 Batch 90/162] avg loss 0.00127332, throughput 9.32625K wps
[Epoch 114 Batch 120/162] avg loss 0.00145155, throughput 9.49572K wps
[Epoch 114 Batch 150/162] avg loss 0.00131222, throughput 9.42428K wps
Begin Testing...
[Epoch 114] train avg loss 0.0013734, dev acc 0.9100, dev avg loss 0.21746, throughput 9.47131K wps
[Epoch 115 Batch 30/162] avg loss 0.00142746, throughput 9.69614K wps
[Epoch 115 Batch 60/162] avg loss 0.00141877, throughput 9.27915K wps
[Epoch 115 Batch 90/162] avg loss 0.00136466, throughput 9.48261K wps
[Epoch 115 Batch 120/162] avg loss 0.00133827, throughput 9.41853K wps
[Epoch 115 Batch 150/162] avg loss 0.00128208, throughput 9.29366K wps
Begin Testing...
[Epoch 115] train avg loss 0.00136657, dev acc 0.9089, dev avg loss 0.218063, throughput 9.44898K wps
[Epoch 116 Batch 30/162] avg loss 0.00112552, throughput 9.58448K wps
[Epoch 116 Batch 60/162] avg loss 0.00147145, throughput 9.30466K wps
[Epoch 116 Batch 90/162] avg loss 0.00138053, throughput 9.30639K wps
[Epoch 116 Batch 120/162] avg loss 0.0014145, throughput 9.44499K wps
[Epoch 116 Batch 150/162] avg loss 0.00130903, throughput 9.39252K wps
Begin Testing...
[Epoch 116] train avg loss 0.00133986, dev acc 0.9089, dev avg loss 0.217129, throughput 9.421K wps
[Epoch 117 Batch 30/162] avg loss 0.00129296, throughput 9.71931K wps
[Epoch 117 Batch 60/162] avg loss 0.0013259, throughput 9.60314K wps
[Epoch 117 Batch 90/162] avg loss 0.00145878, throughput 9.58249K wps
[Epoch 117 Batch 120/162] avg loss 0.0011844, throughput 9.29866K wps
[Epoch 117 Batch 150/162] avg loss 0.00137257, throughput 9.42946K wps
Begin Testing...
[Epoch 117] train avg loss 0.00130454, dev acc 0.9089, dev avg loss 0.217913, throughput 9.5217K wps
[Epoch 118 Batch 30/162] avg loss 0.00123469, throughput 9.70501K wps
[Epoch 118 Batch 60/162] avg loss 0.00134196, throughput 9.42962K wps
[Epoch 118 Batch 90/162] avg loss 0.00129808, throughput 9.37701K wps
[Epoch 118 Batch 120/162] avg loss 0.00142418, throughput 9.44727K wps
[Epoch 118 Batch 150/162] avg loss 0.00132794, throughput 9.41216K wps
Begin Testing...
[Epoch 118] train avg loss 0.00132358, dev acc 0.9100, dev avg loss 0.216944, throughput 9.4682K wps
[Epoch 119 Batch 30/162] avg loss 0.00139347, throughput 9.75187K wps
[Epoch 119 Batch 60/162] avg loss 0.00131741, throughput 9.50871K wps
[Epoch 119 Batch 90/162] avg loss 0.00129081, throughput 9.27596K wps
[Epoch 119 Batch 120/162] avg loss 0.00123133, throughput 9.52415K wps
[Epoch 119 Batch 150/162] avg loss 0.00123535, throughput 9.49749K wps
Begin Testing...
[Epoch 119] train avg loss 0.00128746, dev acc 0.9089, dev avg loss 0.21805, throughput 9.49739K wps
[Epoch 120 Batch 30/162] avg loss 0.00127597, throughput 9.72208K wps
[Epoch 120 Batch 60/162] avg loss 0.0013207, throughput 9.27743K wps
[Epoch 120 Batch 90/162] avg loss 0.00114378, throughput 9.42126K wps
[Epoch 120 Batch 120/162] avg loss 0.0011446, throughput 9.51508K wps
[Epoch 120 Batch 150/162] avg loss 0.00132742, throughput 9.48949K wps
Begin Testing...
[Epoch 120] train avg loss 0.00124889, dev acc 0.9100, dev avg loss 0.217462, throughput 9.46879K wps
[Epoch 121 Batch 30/162] avg loss 0.00124552, throughput 9.47329K wps
[Epoch 121 Batch 60/162] avg loss 0.00127297, throughput 9.49334K wps
[Epoch 121 Batch 90/162] avg loss 0.00126577, throughput 9.51955K wps
[Epoch 121 Batch 120/162] avg loss 0.00123644, throughput 9.34649K wps
[Epoch 121 Batch 150/162] avg loss 0.00136504, throughput 9.35592K wps
Begin Testing...
[Epoch 121] train avg loss 0.00128798, dev acc 0.9089, dev avg loss 0.217517, throughput 9.42624K wps
[Epoch 122 Batch 30/162] avg loss 0.00114948, throughput 9.51817K wps
[Epoch 122 Batch 60/162] avg loss 0.00134988, throughput 9.3742K wps
[Epoch 122 Batch 90/162] avg loss 0.00120601, throughput 9.36498K wps
[Epoch 122 Batch 120/162] avg loss 0.00113344, throughput 9.61576K wps
[Epoch 122 Batch 150/162] avg loss 0.00127018, throughput 9.29356K wps
Begin Testing...
[Epoch 122] train avg loss 0.00123499, dev acc 0.9100, dev avg loss 0.217309, throughput 9.42922K wps
[Epoch 123 Batch 30/162] avg loss 0.00146148, throughput 9.52159K wps
[Epoch 123 Batch 60/162] avg loss 0.00115455, throughput 9.32833K wps
[Epoch 123 Batch 90/162] avg loss 0.00131893, throughput 9.42686K wps
[Epoch 123 Batch 120/162] avg loss 0.00112549, throughput 9.24923K wps
[Epoch 123 Batch 150/162] avg loss 0.00121221, throughput 9.26204K wps
Begin Testing...
[Epoch 123] train avg loss 0.00125224, dev acc 0.9100, dev avg loss 0.218865, throughput 9.34269K wps
[Epoch 124 Batch 30/162] avg loss 0.00129036, throughput 9.58569K wps
[Epoch 124 Batch 60/162] avg loss 0.0013083, throughput 9.39481K wps
[Epoch 124 Batch 90/162] avg loss 0.00122164, throughput 9.42329K wps
[Epoch 124 Batch 120/162] avg loss 0.0011718, throughput 9.45915K wps
[Epoch 124 Batch 150/162] avg loss 0.00127394, throughput 9.48249K wps
Begin Testing...
[Epoch 124] train avg loss 0.00125948, dev acc 0.9089, dev avg loss 0.217134, throughput 9.45046K wps
[Epoch 125 Batch 30/162] avg loss 0.0012259, throughput 9.60296K wps
[Epoch 125 Batch 60/162] avg loss 0.00128314, throughput 9.40371K wps
[Epoch 125 Batch 90/162] avg loss 0.00116839, throughput 9.35168K wps
[Epoch 125 Batch 120/162] avg loss 0.00122757, throughput 9.47643K wps
[Epoch 125 Batch 150/162] avg loss 0.00128965, throughput 9.55298K wps
Begin Testing...
[Epoch 125] train avg loss 0.00123511, dev acc 0.9111, dev avg loss 0.217287, throughput 9.45913K wps
[Epoch 126 Batch 30/162] avg loss 0.00130119, throughput 9.72892K wps
[Epoch 126 Batch 60/162] avg loss 0.00116235, throughput 9.32456K wps
[Epoch 126 Batch 90/162] avg loss 0.00123073, throughput 9.61512K wps
[Epoch 126 Batch 120/162] avg loss 0.00115433, throughput 9.43827K wps
[Epoch 126 Batch 150/162] avg loss 0.00112324, throughput 9.53365K wps
Begin Testing...
[Epoch 126] train avg loss 0.00119951, dev acc 0.9089, dev avg loss 0.21799, throughput 9.52235K wps
[Epoch 127 Batch 30/162] avg loss 0.00120257, throughput 9.49586K wps
[Epoch 127 Batch 60/162] avg loss 0.00113058, throughput 9.40719K wps
[Epoch 127 Batch 90/162] avg loss 0.00109187, throughput 9.3013K wps
[Epoch 127 Batch 120/162] avg loss 0.00104284, throughput 9.32138K wps
[Epoch 127 Batch 150/162] avg loss 0.00119972, throughput 9.37736K wps
Begin Testing...
[Epoch 127] train avg loss 0.00113237, dev acc 0.9089, dev avg loss 0.217814, throughput 9.38909K wps
[Epoch 128 Batch 30/162] avg loss 0.00111884, throughput 9.49609K wps
[Epoch 128 Batch 60/162] avg loss 0.00124483, throughput 9.39658K wps
[Epoch 128 Batch 90/162] avg loss 0.00121978, throughput 9.41221K wps
[Epoch 128 Batch 120/162] avg loss 0.00120005, throughput 9.36342K wps
[Epoch 128 Batch 150/162] avg loss 0.00115952, throughput 9.57743K wps
Begin Testing...
[Epoch 128] train avg loss 0.00119598, dev acc 0.9100, dev avg loss 0.217309, throughput 9.43128K wps
[Epoch 129 Batch 30/162] avg loss 0.00119354, throughput 9.6803K wps
[Epoch 129 Batch 60/162] avg loss 0.00116596, throughput 9.57089K wps
[Epoch 129 Batch 90/162] avg loss 0.00117384, throughput 9.52933K wps
[Epoch 129 Batch 120/162] avg loss 0.0012215, throughput 9.54391K wps
[Epoch 129 Batch 150/162] avg loss 0.00115888, throughput 9.39225K wps
Begin Testing...
[Epoch 129] train avg loss 0.00117451, dev acc 0.9111, dev avg loss 0.217039, throughput 9.53453K wps
[Epoch 130 Batch 30/162] avg loss 0.00113525, throughput 9.53826K wps
[Epoch 130 Batch 60/162] avg loss 0.00109202, throughput 9.34225K wps
[Epoch 130 Batch 90/162] avg loss 0.00122072, throughput 9.50685K wps
[Epoch 130 Batch 120/162] avg loss 0.00130126, throughput 9.24645K wps
[Epoch 130 Batch 150/162] avg loss 0.00113453, throughput 9.3894K wps
Begin Testing...
[Epoch 130] train avg loss 0.00116383, dev acc 0.9089, dev avg loss 0.218549, throughput 9.42608K wps
[Epoch 131 Batch 30/162] avg loss 0.0010676, throughput 9.6262K wps
[Epoch 131 Batch 60/162] avg loss 0.00106278, throughput 9.30048K wps
[Epoch 131 Batch 90/162] avg loss 0.00115892, throughput 9.42219K wps
[Epoch 131 Batch 120/162] avg loss 0.00116847, throughput 9.29837K wps
[Epoch 131 Batch 150/162] avg loss 0.0012069, throughput 9.27848K wps
Begin Testing...
[Epoch 131] train avg loss 0.0011205, dev acc 0.9089, dev avg loss 0.216988, throughput 9.37876K wps
[Epoch 132 Batch 30/162] avg loss 0.0012381, throughput 9.75097K wps
[Epoch 132 Batch 60/162] avg loss 0.00122365, throughput 9.44826K wps
[Epoch 132 Batch 90/162] avg loss 0.00110947, throughput 9.4605K wps
[Epoch 132 Batch 120/162] avg loss 0.000998513, throughput 9.43535K wps
[Epoch 132 Batch 150/162] avg loss 0.0011189, throughput 9.5952K wps
Begin Testing...
[Epoch 132] train avg loss 0.00114755, dev acc 0.9089, dev avg loss 0.216688, throughput 9.53571K wps
[Epoch 133 Batch 30/162] avg loss 0.00117865, throughput 9.66553K wps
[Epoch 133 Batch 60/162] avg loss 0.00106288, throughput 9.42586K wps
[Epoch 133 Batch 90/162] avg loss 0.000953371, throughput 9.35705K wps
[Epoch 133 Batch 120/162] avg loss 0.00111177, throughput 9.48834K wps
[Epoch 133 Batch 150/162] avg loss 0.00114493, throughput 9.38552K wps
Begin Testing...
[Epoch 133] train avg loss 0.00108688, dev acc 0.9111, dev avg loss 0.217284, throughput 9.46816K wps
[Epoch 134 Batch 30/162] avg loss 0.000904694, throughput 9.43243K wps
[Epoch 134 Batch 60/162] avg loss 0.00119572, throughput 9.4211K wps
[Epoch 134 Batch 90/162] avg loss 0.00113766, throughput 9.5462K wps
[Epoch 134 Batch 120/162] avg loss 0.00106033, throughput 9.4957K wps
[Epoch 134 Batch 150/162] avg loss 0.0010659, throughput 9.43953K wps
Begin Testing...
[Epoch 134] train avg loss 0.0010698, dev acc 0.9089, dev avg loss 0.217711, throughput 9.47922K wps
[Epoch 135 Batch 30/162] avg loss 0.00111522, throughput 9.70661K wps
[Epoch 135 Batch 60/162] avg loss 0.0011044, throughput 9.42163K wps
[Epoch 135 Batch 90/162] avg loss 0.00110222, throughput 9.42674K wps
[Epoch 135 Batch 120/162] avg loss 0.0010946, throughput 9.45893K wps
[Epoch 135 Batch 150/162] avg loss 0.00110953, throughput 9.57379K wps
Begin Testing...
[Epoch 135] train avg loss 0.00109447, dev acc 0.9111, dev avg loss 0.217595, throughput 9.50785K wps
[Epoch 136 Batch 30/162] avg loss 0.00107158, throughput 9.48522K wps
[Epoch 136 Batch 60/162] avg loss 0.0011078, throughput 9.33245K wps
[Epoch 136 Batch 90/162] avg loss 0.00110636, throughput 9.54667K wps
[Epoch 136 Batch 120/162] avg loss 0.00110274, throughput 9.44976K wps
[Epoch 136 Batch 150/162] avg loss 0.00104357, throughput 9.512K wps
Begin Testing...
[Epoch 136] train avg loss 0.00108329, dev acc 0.9122, dev avg loss 0.216948, throughput 9.44993K wps
Observed Improvement.
Begin Testing...
[Epoch 137 Batch 30/162] avg loss 0.000940322, throughput 9.51545K wps
[Epoch 137 Batch 60/162] avg loss 0.00106296, throughput 9.49015K wps
[Epoch 137 Batch 90/162] avg loss 0.00108467, throughput 9.2239K wps
[Epoch 137 Batch 120/162] avg loss 0.00094846, throughput 9.50138K wps
[Epoch 137 Batch 150/162] avg loss 0.00120922, throughput 9.53847K wps
Begin Testing...
[Epoch 137] train avg loss 0.0010412, dev acc 0.9122, dev avg loss 0.21761, throughput 9.45085K wps
Observed Improvement.
Begin Testing...
[Epoch 138 Batch 30/162] avg loss 0.00118121, throughput 9.36168K wps
[Epoch 138 Batch 60/162] avg loss 0.00106648, throughput 9.38741K wps
[Epoch 138 Batch 90/162] avg loss 0.000955106, throughput 9.48402K wps
[Epoch 138 Batch 120/162] avg loss 0.000998304, throughput 9.2639K wps
[Epoch 138 Batch 150/162] avg loss 0.0010982, throughput 9.29988K wps
Begin Testing...
[Epoch 138] train avg loss 0.00106742, dev acc 0.9089, dev avg loss 0.21723, throughput 9.3808K wps
[Epoch 139 Batch 30/162] avg loss 0.000961995, throughput 9.49769K wps
[Epoch 139 Batch 60/162] avg loss 0.0010174, throughput 9.33512K wps
[Epoch 139 Batch 90/162] avg loss 0.00103828, throughput 9.32203K wps
[Epoch 139 Batch 120/162] avg loss 0.00106634, throughput 9.43057K wps
[Epoch 139 Batch 150/162] avg loss 0.00101449, throughput 9.38289K wps
Begin Testing...
[Epoch 139] train avg loss 0.00101744, dev acc 0.9089, dev avg loss 0.217293, throughput 9.38719K wps
[Epoch 140 Batch 30/162] avg loss 0.00114081, throughput 9.86684K wps
[Epoch 140 Batch 60/162] avg loss 0.00102723, throughput 9.31189K wps
[Epoch 140 Batch 90/162] avg loss 0.00103639, throughput 9.41228K wps
[Epoch 140 Batch 120/162] avg loss 0.000972539, throughput 9.30971K wps
[Epoch 140 Batch 150/162] avg loss 0.000912421, throughput 9.32654K wps
Begin Testing...
[Epoch 140] train avg loss 0.0010202, dev acc 0.9089, dev avg loss 0.217688, throughput 9.45764K wps
[Epoch 141 Batch 30/162] avg loss 0.00103549, throughput 9.73595K wps
[Epoch 141 Batch 60/162] avg loss 0.0010371, throughput 9.39658K wps
[Epoch 141 Batch 90/162] avg loss 0.000950104, throughput 9.6164K wps
[Epoch 141 Batch 120/162] avg loss 0.000889611, throughput 9.39145K wps
[Epoch 141 Batch 150/162] avg loss 0.0010607, throughput 9.36084K wps
Begin Testing...
[Epoch 141] train avg loss 0.000984185, dev acc 0.9078, dev avg loss 0.217527, throughput 9.48863K wps
[Epoch 142 Batch 30/162] avg loss 0.000941143, throughput 9.55422K wps
[Epoch 142 Batch 60/162] avg loss 0.00097777, throughput 9.5314K wps
[Epoch 142 Batch 90/162] avg loss 0.00102914, throughput 9.29749K wps
[Epoch 142 Batch 120/162] avg loss 0.00109697, throughput 9.39802K wps
[Epoch 142 Batch 150/162] avg loss 0.00109091, throughput 9.45697K wps
Begin Testing...
[Epoch 142] train avg loss 0.00102798, dev acc 0.9100, dev avg loss 0.217129, throughput 9.45113K wps
[Epoch 143 Batch 30/162] avg loss 0.000966385, throughput 9.56571K wps
[Epoch 143 Batch 60/162] avg loss 0.00101092, throughput 9.44759K wps
[Epoch 143 Batch 90/162] avg loss 0.00098039, throughput 9.32376K wps
[Epoch 143 Batch 120/162] avg loss 0.00105542, throughput 9.57046K wps
[Epoch 143 Batch 150/162] avg loss 0.000971644, throughput 9.51946K wps
Begin Testing...
[Epoch 143] train avg loss 0.00100034, dev acc 0.9100, dev avg loss 0.217436, throughput 9.45421K wps
[Epoch 144 Batch 30/162] avg loss 0.000906368, throughput 9.51821K wps
[Epoch 144 Batch 60/162] avg loss 0.00100944, throughput 9.40628K wps
[Epoch 144 Batch 90/162] avg loss 0.000974349, throughput 9.55796K wps
[Epoch 144 Batch 120/162] avg loss 0.00115251, throughput 9.32236K wps
[Epoch 144 Batch 150/162] avg loss 0.000899508, throughput 9.23209K wps
Begin Testing...
[Epoch 144] train avg loss 0.000995989, dev acc 0.9089, dev avg loss 0.21771, throughput 9.39699K wps
[Epoch 145 Batch 30/162] avg loss 0.000943166, throughput 9.52754K wps
[Epoch 145 Batch 60/162] avg loss 0.000986776, throughput 9.56874K wps
[Epoch 145 Batch 90/162] avg loss 0.000947049, throughput 9.27751K wps
[Epoch 145 Batch 120/162] avg loss 0.000978188, throughput 9.53353K wps
[Epoch 145 Batch 150/162] avg loss 0.0010268, throughput 9.33957K wps
Begin Testing...
[Epoch 145] train avg loss 0.000968056, dev acc 0.9089, dev avg loss 0.217637, throughput 9.43889K wps
[Epoch 146 Batch 30/162] avg loss 0.000933939, throughput 9.5603K wps
[Epoch 146 Batch 60/162] avg loss 0.00093129, throughput 9.47407K wps
[Epoch 146 Batch 90/162] avg loss 0.0010716, throughput 9.35156K wps
[Epoch 146 Batch 120/162] avg loss 0.00100839, throughput 9.42933K wps
[Epoch 146 Batch 150/162] avg loss 0.000961902, throughput 9.47344K wps
Begin Testing...
[Epoch 146] train avg loss 0.000974634, dev acc 0.9100, dev avg loss 0.219246, throughput 9.45393K wps
[Epoch 147 Batch 30/162] avg loss 0.000950461, throughput 9.67325K wps
[Epoch 147 Batch 60/162] avg loss 0.00088621, throughput 9.50358K wps
[Epoch 147 Batch 90/162] avg loss 0.000957361, throughput 9.45755K wps
[Epoch 147 Batch 120/162] avg loss 0.00100865, throughput 9.25315K wps
[Epoch 147 Batch 150/162] avg loss 0.000987927, throughput 9.31008K wps
Begin Testing...
[Epoch 147] train avg loss 0.000950391, dev acc 0.9100, dev avg loss 0.218379, throughput 9.42298K wps
[Epoch 148 Batch 30/162] avg loss 0.000908402, throughput 9.53067K wps
[Epoch 148 Batch 60/162] avg loss 0.0010014, throughput 9.35246K wps
[Epoch 148 Batch 90/162] avg loss 0.000947491, throughput 9.60037K wps
[Epoch 148 Batch 120/162] avg loss 0.000870628, throughput 9.2917K wps
[Epoch 148 Batch 150/162] avg loss 0.00100155, throughput 9.37988K wps
Begin Testing...
[Epoch 148] train avg loss 0.000949928, dev acc 0.9089, dev avg loss 0.218625, throughput 9.43507K wps
[Epoch 149 Batch 30/162] avg loss 0.00105856, throughput 9.63001K wps
[Epoch 149 Batch 60/162] avg loss 0.00085787, throughput 9.30589K wps
[Epoch 149 Batch 90/162] avg loss 0.00093853, throughput 9.46787K wps
[Epoch 149 Batch 120/162] avg loss 0.000855349, throughput 9.53084K wps
[Epoch 149 Batch 150/162] avg loss 0.000920344, throughput 9.23146K wps
Begin Testing...
[Epoch 149] train avg loss 0.000931005, dev acc 0.9111, dev avg loss 0.218596, throughput 9.42009K wps
[Epoch 150 Batch 30/162] avg loss 0.000891096, throughput 9.53878K wps
[Epoch 150 Batch 60/162] avg loss 0.000887642, throughput 9.3763K wps
[Epoch 150 Batch 90/162] avg loss 0.000944568, throughput 9.45483K wps
[Epoch 150 Batch 120/162] avg loss 0.000974617, throughput 9.4183K wps
[Epoch 150 Batch 150/162] avg loss 0.000922618, throughput 9.38648K wps
Begin Testing...
[Epoch 150] train avg loss 0.000937083, dev acc 0.9067, dev avg loss 0.218028, throughput 9.42299K wps
[Epoch 151 Batch 30/162] avg loss 0.000851337, throughput 9.67938K wps
[Epoch 151 Batch 60/162] avg loss 0.000914555, throughput 9.35129K wps
[Epoch 151 Batch 90/162] avg loss 0.000945184, throughput 9.50554K wps
[Epoch 151 Batch 120/162] avg loss 0.000979686, throughput 9.59568K wps
[Epoch 151 Batch 150/162] avg loss 0.000900089, throughput 9.2392K wps
Begin Testing...
[Epoch 151] train avg loss 0.000922251, dev acc 0.9111, dev avg loss 0.218506, throughput 9.45755K wps
[Epoch 152 Batch 30/162] avg loss 0.000950433, throughput 9.74775K wps
[Epoch 152 Batch 60/162] avg loss 0.000835245, throughput 9.21597K wps
[Epoch 152 Batch 90/162] avg loss 0.000947667, throughput 9.53754K wps
[Epoch 152 Batch 120/162] avg loss 0.000934707, throughput 9.42621K wps
[Epoch 152 Batch 150/162] avg loss 0.000993374, throughput 9.48486K wps
Begin Testing...
[Epoch 152] train avg loss 0.000933004, dev acc 0.9078, dev avg loss 0.218225, throughput 9.4579K wps
[Epoch 153 Batch 30/162] avg loss 0.000828409, throughput 9.53519K wps
[Epoch 153 Batch 60/162] avg loss 0.000891951, throughput 9.48393K wps
[Epoch 153 Batch 90/162] avg loss 0.000893427, throughput 9.14582K wps
[Epoch 153 Batch 120/162] avg loss 0.000809085, throughput 9.75108K wps
[Epoch 153 Batch 150/162] avg loss 0.000999089, throughput 9.47275K wps
Begin Testing...
[Epoch 153] train avg loss 0.000883462, dev acc 0.9089, dev avg loss 0.218527, throughput 9.47224K wps
[Epoch 154 Batch 30/162] avg loss 0.000845741, throughput 9.43714K wps
[Epoch 154 Batch 60/162] avg loss 0.000831011, throughput 9.50977K wps
[Epoch 154 Batch 90/162] avg loss 0.0008771, throughput 9.59822K wps
[Epoch 154 Batch 120/162] avg loss 0.000885701, throughput 9.46457K wps
[Epoch 154 Batch 150/162] avg loss 0.000892621, throughput 9.3738K wps
Begin Testing...
[Epoch 154] train avg loss 0.00087296, dev acc 0.9133, dev avg loss 0.218458, throughput 9.46598K wps
Observed Improvement.
Begin Testing...
[Epoch 155 Batch 30/162] avg loss 0.000847974, throughput 9.5499K wps
[Epoch 155 Batch 60/162] avg loss 0.00079981, throughput 9.48361K wps
[Epoch 155 Batch 90/162] avg loss 0.000778227, throughput 9.58621K wps
[Epoch 155 Batch 120/162] avg loss 0.000905361, throughput 9.40582K wps
[Epoch 155 Batch 150/162] avg loss 0.000890989, throughput 9.34891K wps
Begin Testing...
[Epoch 155] train avg loss 0.000839313, dev acc 0.9078, dev avg loss 0.218741, throughput 9.48278K wps
[Epoch 156 Batch 30/162] avg loss 0.000806234, throughput 9.67188K wps
[Epoch 156 Batch 60/162] avg loss 0.000751866, throughput 9.32927K wps
[Epoch 156 Batch 90/162] avg loss 0.000888378, throughput 9.47809K wps
[Epoch 156 Batch 120/162] avg loss 0.000921799, throughput 9.31785K wps
[Epoch 156 Batch 150/162] avg loss 0.000805108, throughput 9.37184K wps
Begin Testing...
[Epoch 156] train avg loss 0.000838097, dev acc 0.9089, dev avg loss 0.218654, throughput 9.42775K wps
[Epoch 157 Batch 30/162] avg loss 0.000885883, throughput 9.55811K wps
[Epoch 157 Batch 60/162] avg loss 0.000858526, throughput 9.46356K wps
[Epoch 157 Batch 90/162] avg loss 0.000837578, throughput 9.38363K wps
[Epoch 157 Batch 120/162] avg loss 0.000820341, throughput 9.67253K wps
[Epoch 157 Batch 150/162] avg loss 0.000864976, throughput 9.29977K wps
Begin Testing...
[Epoch 157] train avg loss 0.000853967, dev acc 0.9111, dev avg loss 0.219512, throughput 9.47096K wps
[Epoch 158 Batch 30/162] avg loss 0.000791664, throughput 9.67544K wps
[Epoch 158 Batch 60/162] avg loss 0.000753529, throughput 9.5553K wps
[Epoch 158 Batch 90/162] avg loss 0.000947337, throughput 9.5057K wps
[Epoch 158 Batch 120/162] avg loss 0.000838943, throughput 9.33654K wps
[Epoch 158 Batch 150/162] avg loss 0.00083272, throughput 9.55026K wps
Begin Testing...
[Epoch 158] train avg loss 0.000843226, dev acc 0.9111, dev avg loss 0.219408, throughput 9.51548K wps
[Epoch 159 Batch 30/162] avg loss 0.000829187, throughput 9.59708K wps
[Epoch 159 Batch 60/162] avg loss 0.000829272, throughput 9.48448K wps
[Epoch 159 Batch 90/162] avg loss 0.000735563, throughput 9.53659K wps
[Epoch 159 Batch 120/162] avg loss 0.000784052, throughput 9.55814K wps
[Epoch 159 Batch 150/162] avg loss 0.000908451, throughput 9.41571K wps
Begin Testing...
[Epoch 159] train avg loss 0.000823674, dev acc 0.9100, dev avg loss 0.218803, throughput 9.50258K wps
[Epoch 160 Batch 30/162] avg loss 0.000787179, throughput 9.42681K wps
[Epoch 160 Batch 60/162] avg loss 0.000834185, throughput 9.32166K wps
[Epoch 160 Batch 90/162] avg loss 0.000898912, throughput 9.31049K wps
[Epoch 160 Batch 120/162] avg loss 0.000814403, throughput 9.42526K wps
[Epoch 160 Batch 150/162] avg loss 0.000838858, throughput 9.5899K wps
Begin Testing...
[Epoch 160] train avg loss 0.000828333, dev acc 0.9111, dev avg loss 0.219191, throughput 9.43231K wps
[Epoch 161 Batch 30/162] avg loss 0.000851416, throughput 9.43269K wps
[Epoch 161 Batch 60/162] avg loss 0.000920611, throughput 9.27625K wps
[Epoch 161 Batch 90/162] avg loss 0.000794992, throughput 9.38488K wps
[Epoch 161 Batch 120/162] avg loss 0.000774601, throughput 9.43766K wps
[Epoch 161 Batch 150/162] avg loss 0.000804285, throughput 9.32578K wps
Begin Testing...
[Epoch 161] train avg loss 0.000835438, dev acc 0.9111, dev avg loss 0.219739, throughput 9.36997K wps
[Epoch 162 Batch 30/162] avg loss 0.000862112, throughput 9.60328K wps
[Epoch 162 Batch 60/162] avg loss 0.000833104, throughput 9.27332K wps
[Epoch 162 Batch 90/162] avg loss 0.00081285, throughput 9.35418K wps
[Epoch 162 Batch 120/162] avg loss 0.000778443, throughput 9.58729K wps
[Epoch 162 Batch 150/162] avg loss 0.000814885, throughput 9.3892K wps
Begin Testing...
[Epoch 162] train avg loss 0.000820512, dev acc 0.9100, dev avg loss 0.218348, throughput 9.44421K wps
[Epoch 163 Batch 30/162] avg loss 0.000791981, throughput 9.5537K wps
[Epoch 163 Batch 60/162] avg loss 0.000898185, throughput 9.54914K wps
[Epoch 163 Batch 90/162] avg loss 0.000780961, throughput 9.57133K wps
[Epoch 163 Batch 120/162] avg loss 0.000841725, throughput 9.39339K wps
[Epoch 163 Batch 150/162] avg loss 0.000760108, throughput 9.30701K wps
Begin Testing...
[Epoch 163] train avg loss 0.000813174, dev acc 0.9111, dev avg loss 0.219537, throughput 9.48837K wps
[Epoch 164 Batch 30/162] avg loss 0.000847384, throughput 9.83443K wps
[Epoch 164 Batch 60/162] avg loss 0.00080522, throughput 9.61268K wps
[Epoch 164 Batch 90/162] avg loss 0.000812692, throughput 9.33775K wps
[Epoch 164 Batch 120/162] avg loss 0.000813333, throughput 9.42308K wps
[Epoch 164 Batch 150/162] avg loss 0.000731988, throughput 9.31522K wps
Begin Testing...
[Epoch 164] train avg loss 0.000802866, dev acc 0.9111, dev avg loss 0.219343, throughput 9.47823K wps
[Epoch 165 Batch 30/162] avg loss 0.000857584, throughput 9.51634K wps
[Epoch 165 Batch 60/162] avg loss 0.000776183, throughput 9.39221K wps
[Epoch 165 Batch 90/162] avg loss 0.000788945, throughput 9.54414K wps
[Epoch 165 Batch 120/162] avg loss 0.000816463, throughput 9.34575K wps
[Epoch 165 Batch 150/162] avg loss 0.00073906, throughput 9.31274K wps
Begin Testing...
[Epoch 165] train avg loss 0.000805754, dev acc 0.9111, dev avg loss 0.219014, throughput 9.42161K wps
[Epoch 166 Batch 30/162] avg loss 0.000714372, throughput 9.64993K wps
[Epoch 166 Batch 60/162] avg loss 0.000799931, throughput 9.53911K wps
[Epoch 166 Batch 90/162] avg loss 0.000805026, throughput 9.41001K wps
[Epoch 166 Batch 120/162] avg loss 0.000820947, throughput 9.44277K wps
[Epoch 166 Batch 150/162] avg loss 0.000772819, throughput 9.5152K wps
Begin Testing...
[Epoch 166] train avg loss 0.00077849, dev acc 0.9122, dev avg loss 0.218807, throughput 9.48666K wps
[Epoch 167 Batch 30/162] avg loss 0.000833017, throughput 9.51937K wps
[Epoch 167 Batch 60/162] avg loss 0.000722526, throughput 9.48404K wps
[Epoch 167 Batch 90/162] avg loss 0.000870129, throughput 9.26493K wps
[Epoch 167 Batch 120/162] avg loss 0.000687699, throughput 9.62627K wps
[Epoch 167 Batch 150/162] avg loss 0.000729848, throughput 9.56914K wps
Begin Testing...
[Epoch 167] train avg loss 0.00077113, dev acc 0.9111, dev avg loss 0.219221, throughput 9.47374K wps
[Epoch 168 Batch 30/162] avg loss 0.000832638, throughput 9.46693K wps
[Epoch 168 Batch 60/162] avg loss 0.000869838, throughput 9.41975K wps
[Epoch 168 Batch 90/162] avg loss 0.000763797, throughput 9.31674K wps
[Epoch 168 Batch 120/162] avg loss 0.00077279, throughput 9.41454K wps
[Epoch 168 Batch 150/162] avg loss 0.000916034, throughput 9.54536K wps
Begin Testing...
[Epoch 168] train avg loss 0.000812443, dev acc 0.9122, dev avg loss 0.218801, throughput 9.44904K wps
[Epoch 169 Batch 30/162] avg loss 0.00078582, throughput 9.68088K wps
[Epoch 169 Batch 60/162] avg loss 0.00086975, throughput 9.46113K wps
[Epoch 169 Batch 90/162] avg loss 0.000692748, throughput 9.54727K wps
[Epoch 169 Batch 120/162] avg loss 0.00070873, throughput 9.52783K wps
[Epoch 169 Batch 150/162] avg loss 0.000867694, throughput 9.56831K wps
Begin Testing...
[Epoch 169] train avg loss 0.000786686, dev acc 0.9111, dev avg loss 0.219378, throughput 9.55645K wps
[Epoch 170 Batch 30/162] avg loss 0.000700153, throughput 9.49153K wps
[Epoch 170 Batch 60/162] avg loss 0.000660965, throughput 9.53755K wps
[Epoch 170 Batch 90/162] avg loss 0.000817943, throughput 9.24648K wps
[Epoch 170 Batch 120/162] avg loss 0.000876069, throughput 9.59876K wps
[Epoch 170 Batch 150/162] avg loss 0.000761341, throughput 9.35327K wps
Begin Testing...
[Epoch 170] train avg loss 0.000763576, dev acc 0.9111, dev avg loss 0.219284, throughput 9.46053K wps
[Epoch 171 Batch 30/162] avg loss 0.00071208, throughput 9.50346K wps
[Epoch 171 Batch 60/162] avg loss 0.000735963, throughput 9.38448K wps
[Epoch 171 Batch 90/162] avg loss 0.000858362, throughput 9.35254K wps
[Epoch 171 Batch 120/162] avg loss 0.000817886, throughput 9.42085K wps
[Epoch 171 Batch 150/162] avg loss 0.000727119, throughput 9.49865K wps
Begin Testing...
[Epoch 171] train avg loss 0.000759, dev acc 0.9133, dev avg loss 0.219348, throughput 9.43984K wps
Observed Improvement.
Begin Testing...
[Epoch 172 Batch 30/162] avg loss 0.000668514, throughput 9.78753K wps
[Epoch 172 Batch 60/162] avg loss 0.000855718, throughput 9.60511K wps
[Epoch 172 Batch 90/162] avg loss 0.000720326, throughput 9.43928K wps
[Epoch 172 Batch 120/162] avg loss 0.000823596, throughput 9.52169K wps
[Epoch 172 Batch 150/162] avg loss 0.000740463, throughput 9.34638K wps
Begin Testing...
[Epoch 172] train avg loss 0.000764416, dev acc 0.9100, dev avg loss 0.219732, throughput 9.51908K wps
[Epoch 173 Batch 30/162] avg loss 0.000748294, throughput 9.82582K wps
[Epoch 173 Batch 60/162] avg loss 0.000646413, throughput 9.49K wps
[Epoch 173 Batch 90/162] avg loss 0.000722156, throughput 9.47378K wps
[Epoch 173 Batch 120/162] avg loss 0.000707626, throughput 9.2572K wps
[Epoch 173 Batch 150/162] avg loss 0.000665613, throughput 9.35911K wps
Begin Testing...
[Epoch 173] train avg loss 0.000702009, dev acc 0.9111, dev avg loss 0.220906, throughput 9.48383K wps
[Epoch 174 Batch 30/162] avg loss 0.00065913, throughput 9.56889K wps
[Epoch 174 Batch 60/162] avg loss 0.00079676, throughput 9.42198K wps
[Epoch 174 Batch 90/162] avg loss 0.000734391, throughput 9.50835K wps
[Epoch 174 Batch 120/162] avg loss 0.000747235, throughput 9.54263K wps
[Epoch 174 Batch 150/162] avg loss 0.000734081, throughput 9.63617K wps
Begin Testing...
[Epoch 174] train avg loss 0.000731815, dev acc 0.9100, dev avg loss 0.220846, throughput 9.5316K wps
[Epoch 175 Batch 30/162] avg loss 0.000723426, throughput 9.61397K wps
[Epoch 175 Batch 60/162] avg loss 0.000643744, throughput 9.35773K wps
[Epoch 175 Batch 90/162] avg loss 0.000810967, throughput 9.52385K wps
[Epoch 175 Batch 120/162] avg loss 0.000692436, throughput 9.43366K wps
[Epoch 175 Batch 150/162] avg loss 0.000675295, throughput 9.26942K wps
Begin Testing...
[Epoch 175] train avg loss 0.000701854, dev acc 0.9111, dev avg loss 0.220468, throughput 9.43276K wps
[Epoch 176 Batch 30/162] avg loss 0.000706459, throughput 9.76601K wps
[Epoch 176 Batch 60/162] avg loss 0.00081076, throughput 9.2771K wps
[Epoch 176 Batch 90/162] avg loss 0.000668239, throughput 9.57896K wps
[Epoch 176 Batch 120/162] avg loss 0.000790366, throughput 9.54263K wps
[Epoch 176 Batch 150/162] avg loss 0.000678171, throughput 9.49162K wps
Begin Testing...
[Epoch 176] train avg loss 0.000725452, dev acc 0.9100, dev avg loss 0.220934, throughput 9.52951K wps
[Epoch 177 Batch 30/162] avg loss 0.000656914, throughput 9.48847K wps
[Epoch 177 Batch 60/162] avg loss 0.000675079, throughput 9.46237K wps
[Epoch 177 Batch 90/162] avg loss 0.000707773, throughput 9.41983K wps
[Epoch 177 Batch 120/162] avg loss 0.00073217, throughput 9.53725K wps
[Epoch 177 Batch 150/162] avg loss 0.000644849, throughput 9.61461K wps
Begin Testing...
[Epoch 177] train avg loss 0.000690147, dev acc 0.9122, dev avg loss 0.220507, throughput 9.51285K wps
[Epoch 178 Batch 30/162] avg loss 0.000734028, throughput 9.43348K wps
[Epoch 178 Batch 60/162] avg loss 0.000790013, throughput 9.26171K wps
[Epoch 178 Batch 90/162] avg loss 0.000688852, throughput 9.45157K wps
[Epoch 178 Batch 120/162] avg loss 0.000729231, throughput 9.56385K wps
[Epoch 178 Batch 150/162] avg loss 0.000683801, throughput 9.62795K wps
Begin Testing...
[Epoch 178] train avg loss 0.000723559, dev acc 0.9089, dev avg loss 0.220602, throughput 9.46845K wps
[Epoch 179 Batch 30/162] avg loss 0.000783897, throughput 9.61771K wps
[Epoch 179 Batch 60/162] avg loss 0.000733344, throughput 9.35488K wps
[Epoch 179 Batch 90/162] avg loss 0.000649741, throughput 9.50528K wps
[Epoch 179 Batch 120/162] avg loss 0.000690017, throughput 9.49378K wps
[Epoch 179 Batch 150/162] avg loss 0.000801969, throughput 9.4481K wps
Begin Testing...
[Epoch 179] train avg loss 0.000726005, dev acc 0.9100, dev avg loss 0.221108, throughput 9.47856K wps
[Epoch 180 Batch 30/162] avg loss 0.000793934, throughput 9.63853K wps
[Epoch 180 Batch 60/162] avg loss 0.000678616, throughput 9.39852K wps
[Epoch 180 Batch 90/162] avg loss 0.000633588, throughput 9.32578K wps
[Epoch 180 Batch 120/162] avg loss 0.000698826, throughput 9.34528K wps
[Epoch 180 Batch 150/162] avg loss 0.000651103, throughput 9.51358K wps
Begin Testing...
[Epoch 180] train avg loss 0.000693971, dev acc 0.9122, dev avg loss 0.220552, throughput 9.45198K wps
[Epoch 181 Batch 30/162] avg loss 0.000655821, throughput 9.51276K wps
[Epoch 181 Batch 60/162] avg loss 0.000700056, throughput 9.49053K wps
[Epoch 181 Batch 90/162] avg loss 0.000682758, throughput 9.45994K wps
[Epoch 181 Batch 120/162] avg loss 0.000625761, throughput 9.41032K wps
[Epoch 181 Batch 150/162] avg loss 0.000709177, throughput 9.39926K wps
Begin Testing...
[Epoch 181] train avg loss 0.00067662, dev acc 0.9111, dev avg loss 0.221071, throughput 9.46415K wps
[Epoch 182 Batch 30/162] avg loss 0.000645041, throughput 9.57483K wps
[Epoch 182 Batch 60/162] avg loss 0.00059718, throughput 9.59006K wps
[Epoch 182 Batch 90/162] avg loss 0.000656296, throughput 9.32107K wps
[Epoch 182 Batch 120/162] avg loss 0.000735413, throughput 9.31927K wps
[Epoch 182 Batch 150/162] avg loss 0.000681141, throughput 9.38273K wps
Begin Testing...
[Epoch 182] train avg loss 0.000654909, dev acc 0.9100, dev avg loss 0.224301, throughput 9.43853K wps
[Epoch 183 Batch 30/162] avg loss 0.000741379, throughput 9.63121K wps
[Epoch 183 Batch 60/162] avg loss 0.000681086, throughput 9.45125K wps
[Epoch 183 Batch 90/162] avg loss 0.000707458, throughput 9.43953K wps
[Epoch 183 Batch 120/162] avg loss 0.000622278, throughput 9.58747K wps
[Epoch 183 Batch 150/162] avg loss 0.00065018, throughput 9.5187K wps
Begin Testing...
[Epoch 183] train avg loss 0.000683459, dev acc 0.9122, dev avg loss 0.220982, throughput 9.51112K wps
[Epoch 184 Batch 30/162] avg loss 0.000675088, throughput 9.59213K wps
[Epoch 184 Batch 60/162] avg loss 0.000601738, throughput 9.53689K wps
[Epoch 184 Batch 90/162] avg loss 0.000634056, throughput 9.61285K wps
[Epoch 184 Batch 120/162] avg loss 0.000649196, throughput 9.39535K wps
[Epoch 184 Batch 150/162] avg loss 0.000703888, throughput 9.38726K wps
Begin Testing...
[Epoch 184] train avg loss 0.000661683, dev acc 0.9111, dev avg loss 0.221111, throughput 9.50496K wps
[Epoch 185 Batch 30/162] avg loss 0.000618517, throughput 9.65382K wps
[Epoch 185 Batch 60/162] avg loss 0.000625492, throughput 9.41099K wps
[Epoch 185 Batch 90/162] avg loss 0.000764772, throughput 9.32203K wps
[Epoch 185 Batch 120/162] avg loss 0.000652417, throughput 9.44632K wps
[Epoch 185 Batch 150/162] avg loss 0.000707376, throughput 9.50616K wps
Begin Testing...
[Epoch 185] train avg loss 0.000668261, dev acc 0.9111, dev avg loss 0.221245, throughput 9.47147K wps
[Epoch 186 Batch 30/162] avg loss 0.000547317, throughput 9.48102K wps
[Epoch 186 Batch 60/162] avg loss 0.000744615, throughput 9.48275K wps
[Epoch 186 Batch 90/162] avg loss 0.000679294, throughput 9.46677K wps
[Epoch 186 Batch 120/162] avg loss 0.000662117, throughput 9.32649K wps
[Epoch 186 Batch 150/162] avg loss 0.000625509, throughput 9.41767K wps
Begin Testing...
[Epoch 186] train avg loss 0.000649875, dev acc 0.9111, dev avg loss 0.221843, throughput 9.42952K wps
[Epoch 187 Batch 30/162] avg loss 0.000623454, throughput 9.46475K wps
[Epoch 187 Batch 60/162] avg loss 0.000711432, throughput 9.55529K wps
[Epoch 187 Batch 90/162] avg loss 0.000576701, throughput 9.50823K wps
[Epoch 187 Batch 120/162] avg loss 0.000793025, throughput 9.45893K wps
[Epoch 187 Batch 150/162] avg loss 0.000557966, throughput 9.37504K wps
Begin Testing...
[Epoch 187] train avg loss 0.00065417, dev acc 0.9100, dev avg loss 0.222138, throughput 9.46014K wps
[Epoch 188 Batch 30/162] avg loss 0.000670797, throughput 9.63472K wps
[Epoch 188 Batch 60/162] avg loss 0.000610627, throughput 9.47177K wps
[Epoch 188 Batch 90/162] avg loss 0.000626132, throughput 9.3651K wps
[Epoch 188 Batch 120/162] avg loss 0.000687585, throughput 9.42812K wps
[Epoch 188 Batch 150/162] avg loss 0.000674436, throughput 9.50561K wps
Begin Testing...
[Epoch 188] train avg loss 0.000659198, dev acc 0.9111, dev avg loss 0.221231, throughput 9.49565K wps
[Epoch 189 Batch 30/162] avg loss 0.000626368, throughput 9.53522K wps
[Epoch 189 Batch 60/162] avg loss 0.000639245, throughput 9.57837K wps
[Epoch 189 Batch 90/162] avg loss 0.000647012, throughput 9.50315K wps
[Epoch 189 Batch 120/162] avg loss 0.00063006, throughput 9.48926K wps
[Epoch 189 Batch 150/162] avg loss 0.00063728, throughput 9.30732K wps
Begin Testing...
[Epoch 189] train avg loss 0.000635933, dev acc 0.9111, dev avg loss 0.221186, throughput 9.48995K wps
[Epoch 190 Batch 30/162] avg loss 0.000605764, throughput 9.74305K wps
[Epoch 190 Batch 60/162] avg loss 0.00066341, throughput 9.46228K wps
[Epoch 190 Batch 90/162] avg loss 0.000664092, throughput 9.52519K wps
[Epoch 190 Batch 120/162] avg loss 0.000595044, throughput 9.45409K wps
[Epoch 190 Batch 150/162] avg loss 0.000611746, throughput 9.59399K wps
Begin Testing...
[Epoch 190] train avg loss 0.000635765, dev acc 0.9111, dev avg loss 0.22141, throughput 9.537K wps
[Epoch 191 Batch 30/162] avg loss 0.000604996, throughput 9.55262K wps
[Epoch 191 Batch 60/162] avg loss 0.000557252, throughput 9.47108K wps
[Epoch 191 Batch 90/162] avg loss 0.000659998, throughput 9.35844K wps
[Epoch 191 Batch 120/162] avg loss 0.000655206, throughput 9.50682K wps
[Epoch 191 Batch 150/162] avg loss 0.000597222, throughput 9.24463K wps
Begin Testing...
[Epoch 191] train avg loss 0.000624791, dev acc 0.9111, dev avg loss 0.221818, throughput 9.43396K wps
[Epoch 192 Batch 30/162] avg loss 0.000514695, throughput 9.5875K wps
[Epoch 192 Batch 60/162] avg loss 0.000630381, throughput 9.55645K wps
[Epoch 192 Batch 90/162] avg loss 0.000700102, throughput 9.27477K wps
[Epoch 192 Batch 120/162] avg loss 0.000646597, throughput 9.28425K wps
[Epoch 192 Batch 150/162] avg loss 0.000573294, throughput 9.33858K wps
Begin Testing...
[Epoch 192] train avg loss 0.000611497, dev acc 0.9144, dev avg loss 0.221586, throughput 9.40068K wps
Observed Improvement.
Begin Testing...
[Epoch 193 Batch 30/162] avg loss 0.000560191, throughput 9.59658K wps
[Epoch 193 Batch 60/162] avg loss 0.000740667, throughput 9.24633K wps
[Epoch 193 Batch 90/162] avg loss 0.000560823, throughput 9.39799K wps
[Epoch 193 Batch 120/162] avg loss 0.000583794, throughput 9.44473K wps
[Epoch 193 Batch 150/162] avg loss 0.000597312, throughput 9.30449K wps
Begin Testing...
[Epoch 193] train avg loss 0.0006033, dev acc 0.9111, dev avg loss 0.222608, throughput 9.38907K wps
[Epoch 194 Batch 30/162] avg loss 0.000574885, throughput 9.74389K wps
[Epoch 194 Batch 60/162] avg loss 0.000563379, throughput 9.61543K wps
[Epoch 194 Batch 90/162] avg loss 0.000617787, throughput 9.26465K wps
[Epoch 194 Batch 120/162] avg loss 0.000593837, throughput 9.32469K wps
[Epoch 194 Batch 150/162] avg loss 0.000664615, throughput 9.32426K wps
Begin Testing...
[Epoch 194] train avg loss 0.00059914, dev acc 0.9133, dev avg loss 0.222047, throughput 9.44237K wps
[Epoch 195 Batch 30/162] avg loss 0.000616478, throughput 9.57857K wps
[Epoch 195 Batch 60/162] avg loss 0.000602188, throughput 9.36059K wps
[Epoch 195 Batch 90/162] avg loss 0.000577499, throughput 9.44338K wps
[Epoch 195 Batch 120/162] avg loss 0.000594717, throughput 9.38974K wps
[Epoch 195 Batch 150/162] avg loss 0.000598117, throughput 9.31996K wps
Begin Testing...
[Epoch 195] train avg loss 0.000590448, dev acc 0.9144, dev avg loss 0.222472, throughput 9.41188K wps
Observed Improvement.
Begin Testing...
[Epoch 196 Batch 30/162] avg loss 0.000588741, throughput 9.60324K wps
[Epoch 196 Batch 60/162] avg loss 0.000615889, throughput 9.31029K wps
[Epoch 196 Batch 90/162] avg loss 0.000646017, throughput 9.38772K wps
[Epoch 196 Batch 120/162] avg loss 0.000617099, throughput 9.3483K wps
[Epoch 196 Batch 150/162] avg loss 0.000583379, throughput 9.45076K wps
Begin Testing...
[Epoch 196] train avg loss 0.000611141, dev acc 0.9156, dev avg loss 0.222345, throughput 9.41044K wps
Observed Improvement.
Begin Testing...
[Epoch 197 Batch 30/162] avg loss 0.000613499, throughput 9.70863K wps
[Epoch 197 Batch 60/162] avg loss 0.000608815, throughput 9.37012K wps
[Epoch 197 Batch 90/162] avg loss 0.000594961, throughput 9.43718K wps
[Epoch 197 Batch 120/162] avg loss 0.000607892, throughput 9.46398K wps
[Epoch 197 Batch 150/162] avg loss 0.000572639, throughput 9.33094K wps
Begin Testing...
[Epoch 197] train avg loss 0.000604753, dev acc 0.9111, dev avg loss 0.222523, throughput 9.47392K wps
[Epoch 198 Batch 30/162] avg loss 0.000431143, throughput 9.59409K wps
[Epoch 198 Batch 60/162] avg loss 0.000634574, throughput 9.58523K wps
[Epoch 198 Batch 90/162] avg loss 0.000547388, throughput 9.42557K wps
[Epoch 198 Batch 120/162] avg loss 0.000604377, throughput 9.51715K wps
[Epoch 198 Batch 150/162] avg loss 0.000609351, throughput 9.45686K wps
Begin Testing...
[Epoch 198] train avg loss 0.000567412, dev acc 0.9089, dev avg loss 0.224307, throughput 9.49171K wps
[Epoch 199 Batch 30/162] avg loss 0.00055981, throughput 9.54162K wps
[Epoch 199 Batch 60/162] avg loss 0.000588042, throughput 9.35626K wps
[Epoch 199 Batch 90/162] avg loss 0.000593721, throughput 9.43048K wps
[Epoch 199 Batch 120/162] avg loss 0.000594662, throughput 9.40473K wps
[Epoch 199 Batch 150/162] avg loss 0.000569263, throughput 9.42501K wps
Begin Testing...
[Epoch 199] train avg loss 0.000586908, dev acc 0.9111, dev avg loss 0.223292, throughput 9.45381K wps
Test loss 0.204369, test acc 0.9240
Total time cost 440.50s
[Epoch 0 Batch 30/162] avg loss 0.0140069, throughput 7.36917K wps
[Epoch 0 Batch 60/162] avg loss 0.0138984, throughput 9.54449K wps
[Epoch 0 Batch 90/162] avg loss 0.0136342, throughput 9.5696K wps
[Epoch 0 Batch 120/162] avg loss 0.0135677, throughput 9.61726K wps
[Epoch 0 Batch 150/162] avg loss 0.0135645, throughput 9.42766K wps
Begin Testing...
[Epoch 0] train avg loss 0.0137182, dev acc 0.7678, dev avg loss 0.662107, throughput 9.05011K wps
Observed Improvement.
Begin Testing...
[Epoch 1 Batch 30/162] avg loss 0.0133285, throughput 9.6298K wps
[Epoch 1 Batch 60/162] avg loss 0.01315, throughput 9.49642K wps
[Epoch 1 Batch 90/162] avg loss 0.0129559, throughput 9.55913K wps
[Epoch 1 Batch 120/162] avg loss 0.0130114, throughput 9.53114K wps
[Epoch 1 Batch 150/162] avg loss 0.0127979, throughput 9.46374K wps
Begin Testing...
[Epoch 1] train avg loss 0.0130333, dev acc 0.7844, dev avg loss 0.632208, throughput 9.53668K wps
Observed Improvement.
Begin Testing...
[Epoch 2 Batch 30/162] avg loss 0.0126836, throughput 9.48504K wps
[Epoch 2 Batch 60/162] avg loss 0.0126327, throughput 9.3342K wps
[Epoch 2 Batch 90/162] avg loss 0.0124378, throughput 9.56779K wps
[Epoch 2 Batch 120/162] avg loss 0.0121399, throughput 9.58261K wps
[Epoch 2 Batch 150/162] avg loss 0.0120962, throughput 9.43972K wps
Begin Testing...
[Epoch 2] train avg loss 0.0123654, dev acc 0.8533, dev avg loss 0.595958, throughput 9.4803K wps
Observed Improvement.
Begin Testing...
[Epoch 3 Batch 30/162] avg loss 0.0120276, throughput 9.70355K wps
[Epoch 3 Batch 60/162] avg loss 0.0118279, throughput 9.64122K wps
[Epoch 3 Batch 90/162] avg loss 0.011575, throughput 9.39499K wps
[Epoch 3 Batch 120/162] avg loss 0.0115579, throughput 9.4456K wps
[Epoch 3 Batch 150/162] avg loss 0.0114923, throughput 9.40671K wps
Begin Testing...
[Epoch 3] train avg loss 0.0116797, dev acc 0.8556, dev avg loss 0.556526, throughput 9.5267K wps
Observed Improvement.
Begin Testing...
[Epoch 4 Batch 30/162] avg loss 0.0110548, throughput 9.59586K wps
[Epoch 4 Batch 60/162] avg loss 0.01102, throughput 9.48823K wps
[Epoch 4 Batch 90/162] avg loss 0.0108956, throughput 9.38734K wps
[Epoch 4 Batch 120/162] avg loss 0.0108845, throughput 9.27821K wps
[Epoch 4 Batch 150/162] avg loss 0.0104998, throughput 9.53568K wps
Begin Testing...
[Epoch 4] train avg loss 0.0108465, dev acc 0.8622, dev avg loss 0.513422, throughput 9.4484K wps
Observed Improvement.
Begin Testing...
[Epoch 5 Batch 30/162] avg loss 0.0103152, throughput 9.61143K wps
[Epoch 5 Batch 60/162] avg loss 0.0102293, throughput 9.39581K wps
[Epoch 5 Batch 90/162] avg loss 0.0100901, throughput 9.24489K wps
[Epoch 5 Batch 120/162] avg loss 0.00993125, throughput 9.43267K wps
[Epoch 5 Batch 150/162] avg loss 0.00961469, throughput 9.25512K wps
Begin Testing...
[Epoch 5] train avg loss 0.0100175, dev acc 0.8711, dev avg loss 0.470679, throughput 9.39317K wps
Observed Improvement.
Begin Testing...
[Epoch 6 Batch 30/162] avg loss 0.0096784, throughput 9.61991K wps
[Epoch 6 Batch 60/162] avg loss 0.00960195, throughput 9.27989K wps
[Epoch 6 Batch 90/162] avg loss 0.00931351, throughput 9.37002K wps
[Epoch 6 Batch 120/162] avg loss 0.0089509, throughput 9.31943K wps
[Epoch 6 Batch 150/162] avg loss 0.00909913, throughput 9.40027K wps
Begin Testing...
[Epoch 6] train avg loss 0.00930281, dev acc 0.8700, dev avg loss 0.433049, throughput 9.39946K wps
[Epoch 7 Batch 30/162] avg loss 0.00890252, throughput 9.70778K wps
[Epoch 7 Batch 60/162] avg loss 0.00883236, throughput 9.39398K wps
[Epoch 7 Batch 90/162] avg loss 0.00835881, throughput 9.42363K wps
[Epoch 7 Batch 120/162] avg loss 0.00834112, throughput 9.34651K wps
[Epoch 7 Batch 150/162] avg loss 0.00868605, throughput 9.26211K wps
Begin Testing...
[Epoch 7] train avg loss 0.00860678, dev acc 0.8722, dev avg loss 0.401618, throughput 9.41468K wps
Observed Improvement.
Begin Testing...
[Epoch 8 Batch 30/162] avg loss 0.00808057, throughput 9.71866K wps
[Epoch 8 Batch 60/162] avg loss 0.0079983, throughput 9.49226K wps
[Epoch 8 Batch 90/162] avg loss 0.00799038, throughput 9.43594K wps
[Epoch 8 Batch 120/162] avg loss 0.00808982, throughput 9.66972K wps
[Epoch 8 Batch 150/162] avg loss 0.00804835, throughput 9.31564K wps
Begin Testing...
[Epoch 8] train avg loss 0.00802447, dev acc 0.8789, dev avg loss 0.37685, throughput 9.52069K wps
Observed Improvement.
Begin Testing...
[Epoch 9 Batch 30/162] avg loss 0.0077189, throughput 9.68405K wps
[Epoch 9 Batch 60/162] avg loss 0.00759048, throughput 9.42103K wps
[Epoch 9 Batch 90/162] avg loss 0.00756769, throughput 9.51485K wps
[Epoch 9 Batch 120/162] avg loss 0.00741981, throughput 9.55572K wps
[Epoch 9 Batch 150/162] avg loss 0.00765425, throughput 9.35345K wps
Begin Testing...
[Epoch 9] train avg loss 0.0075527, dev acc 0.8800, dev avg loss 0.3568, throughput 9.50005K wps
Observed Improvement.
Begin Testing...
[Epoch 10 Batch 30/162] avg loss 0.00751715, throughput 9.5448K wps
[Epoch 10 Batch 60/162] avg loss 0.00704189, throughput 9.32163K wps
[Epoch 10 Batch 90/162] avg loss 0.00725858, throughput 9.38729K wps
[Epoch 10 Batch 120/162] avg loss 0.00705927, throughput 9.37318K wps
[Epoch 10 Batch 150/162] avg loss 0.0071874, throughput 9.46609K wps
Begin Testing...
[Epoch 10] train avg loss 0.0071888, dev acc 0.8778, dev avg loss 0.342902, throughput 9.44125K wps
[Epoch 11 Batch 30/162] avg loss 0.00697469, throughput 9.58783K wps
[Epoch 11 Batch 60/162] avg loss 0.00708898, throughput 9.41035K wps
[Epoch 11 Batch 90/162] avg loss 0.00683055, throughput 9.62109K wps
[Epoch 11 Batch 120/162] avg loss 0.00692575, throughput 9.35428K wps
[Epoch 11 Batch 150/162] avg loss 0.00662515, throughput 9.39412K wps
Begin Testing...
[Epoch 11] train avg loss 0.00688096, dev acc 0.8822, dev avg loss 0.328868, throughput 9.45901K wps
Observed Improvement.
Begin Testing...
[Epoch 12 Batch 30/162] avg loss 0.00679743, throughput 9.50357K wps
[Epoch 12 Batch 60/162] avg loss 0.00674689, throughput 9.37416K wps
[Epoch 12 Batch 90/162] avg loss 0.00683203, throughput 9.34835K wps
[Epoch 12 Batch 120/162] avg loss 0.00663874, throughput 9.3563K wps
[Epoch 12 Batch 150/162] avg loss 0.00673331, throughput 9.32875K wps
Begin Testing...
[Epoch 12] train avg loss 0.00671956, dev acc 0.8822, dev avg loss 0.318571, throughput 9.38593K wps
Observed Improvement.
Begin Testing...
[Epoch 13 Batch 30/162] avg loss 0.00625132, throughput 9.72615K wps
[Epoch 13 Batch 60/162] avg loss 0.00663428, throughput 9.49473K wps
[Epoch 13 Batch 90/162] avg loss 0.00654897, throughput 9.33174K wps
[Epoch 13 Batch 120/162] avg loss 0.00644648, throughput 9.37604K wps
[Epoch 13 Batch 150/162] avg loss 0.00650013, throughput 9.47905K wps
Begin Testing...
[Epoch 13] train avg loss 0.00645161, dev acc 0.8822, dev avg loss 0.310064, throughput 9.47149K wps
Observed Improvement.
Begin Testing...
[Epoch 14 Batch 30/162] avg loss 0.00629835, throughput 9.77453K wps
[Epoch 14 Batch 60/162] avg loss 0.00644353, throughput 9.24121K wps
[Epoch 14 Batch 90/162] avg loss 0.00630807, throughput 9.37395K wps
[Epoch 14 Batch 120/162] avg loss 0.00608686, throughput 9.4893K wps
[Epoch 14 Batch 150/162] avg loss 0.00627866, throughput 9.32942K wps
Begin Testing...
[Epoch 14] train avg loss 0.00624077, dev acc 0.8856, dev avg loss 0.302529, throughput 9.44467K wps
Observed Improvement.
Begin Testing...
[Epoch 15 Batch 30/162] avg loss 0.00609976, throughput 9.633K wps
[Epoch 15 Batch 60/162] avg loss 0.00594028, throughput 9.38258K wps
[Epoch 15 Batch 90/162] avg loss 0.00593469, throughput 9.30059K wps
[Epoch 15 Batch 120/162] avg loss 0.00611713, throughput 9.39439K wps
[Epoch 15 Batch 150/162] avg loss 0.00626575, throughput 9.48052K wps
Begin Testing...
[Epoch 15] train avg loss 0.00609344, dev acc 0.8900, dev avg loss 0.296641, throughput 9.4513K wps
Observed Improvement.
Begin Testing...
[Epoch 16 Batch 30/162] avg loss 0.00585255, throughput 9.69081K wps
[Epoch 16 Batch 60/162] avg loss 0.00587755, throughput 9.32503K wps
[Epoch 16 Batch 90/162] avg loss 0.00595783, throughput 9.40119K wps
[Epoch 16 Batch 120/162] avg loss 0.00599017, throughput 9.41394K wps
[Epoch 16 Batch 150/162] avg loss 0.00571896, throughput 9.334K wps
Begin Testing...
[Epoch 16] train avg loss 0.00586921, dev acc 0.8900, dev avg loss 0.290993, throughput 9.42237K wps
Observed Improvement.
Begin Testing...
[Epoch 17 Batch 30/162] avg loss 0.00610333, throughput 9.59805K wps
[Epoch 17 Batch 60/162] avg loss 0.00597621, throughput 9.43325K wps
[Epoch 17 Batch 90/162] avg loss 0.00564769, throughput 9.48681K wps
[Epoch 17 Batch 120/162] avg loss 0.0058535, throughput 9.38143K wps
[Epoch 17 Batch 150/162] avg loss 0.00536508, throughput 9.45121K wps
Begin Testing...
[Epoch 17] train avg loss 0.00578944, dev acc 0.8922, dev avg loss 0.286017, throughput 9.47136K wps
Observed Improvement.
Begin Testing...
[Epoch 18 Batch 30/162] avg loss 0.00538034, throughput 9.7146K wps
[Epoch 18 Batch 60/162] avg loss 0.005153, throughput 9.38022K wps
[Epoch 18 Batch 90/162] avg loss 0.00599097, throughput 9.38825K wps
[Epoch 18 Batch 120/162] avg loss 0.00571298, throughput 9.54474K wps
[Epoch 18 Batch 150/162] avg loss 0.00563774, throughput 9.31262K wps
Begin Testing...
[Epoch 18] train avg loss 0.00560174, dev acc 0.8889, dev avg loss 0.281648, throughput 9.44555K wps
[Epoch 19 Batch 30/162] avg loss 0.00561749, throughput 9.62841K wps
[Epoch 19 Batch 60/162] avg loss 0.00573868, throughput 9.29941K wps
[Epoch 19 Batch 90/162] avg loss 0.00559285, throughput 9.37773K wps
[Epoch 19 Batch 120/162] avg loss 0.0055108, throughput 9.29517K wps
[Epoch 19 Batch 150/162] avg loss 0.005272, throughput 9.22967K wps
Begin Testing...
[Epoch 19] train avg loss 0.00552121, dev acc 0.8922, dev avg loss 0.278471, throughput 9.35095K wps
Observed Improvement.
Begin Testing...
[Epoch 20 Batch 30/162] avg loss 0.00551983, throughput 9.53724K wps
[Epoch 20 Batch 60/162] avg loss 0.00541484, throughput 9.40931K wps
[Epoch 20 Batch 90/162] avg loss 0.00549836, throughput 8.45811K wps
[Epoch 20 Batch 120/162] avg loss 0.00550838, throughput 9.27476K wps
[Epoch 20 Batch 150/162] avg loss 0.00557812, throughput 9.35609K wps
Begin Testing...
[Epoch 20] train avg loss 0.00546809, dev acc 0.8900, dev avg loss 0.274994, throughput 9.21718K wps
[Epoch 21 Batch 30/162] avg loss 0.00533628, throughput 9.55324K wps
[Epoch 21 Batch 60/162] avg loss 0.00536702, throughput 9.29962K wps
[Epoch 21 Batch 90/162] avg loss 0.00495739, throughput 9.34445K wps
[Epoch 21 Batch 120/162] avg loss 0.00571434, throughput 9.43448K wps
[Epoch 21 Batch 150/162] avg loss 0.00537062, throughput 9.37577K wps
Begin Testing...
[Epoch 21] train avg loss 0.00533482, dev acc 0.8911, dev avg loss 0.271823, throughput 9.39894K wps
[Epoch 22 Batch 30/162] avg loss 0.0050295, throughput 9.59312K wps
[Epoch 22 Batch 60/162] avg loss 0.00553782, throughput 9.54525K wps
[Epoch 22 Batch 90/162] avg loss 0.00522381, throughput 9.31626K wps
[Epoch 22 Batch 120/162] avg loss 0.0049755, throughput 9.34733K wps
[Epoch 22 Batch 150/162] avg loss 0.00536603, throughput 9.43557K wps
Begin Testing...
[Epoch 22] train avg loss 0.00520627, dev acc 0.8967, dev avg loss 0.269632, throughput 9.46019K wps
Observed Improvement.
Begin Testing...
[Epoch 23 Batch 30/162] avg loss 0.00492685, throughput 9.43279K wps
[Epoch 23 Batch 60/162] avg loss 0.00487685, throughput 9.31221K wps
[Epoch 23 Batch 90/162] avg loss 0.00516623, throughput 9.28358K wps
[Epoch 23 Batch 120/162] avg loss 0.00489556, throughput 9.60979K wps
[Epoch 23 Batch 150/162] avg loss 0.00507732, throughput 9.31583K wps
Begin Testing...
[Epoch 23] train avg loss 0.00498728, dev acc 0.8944, dev avg loss 0.266731, throughput 9.39981K wps
[Epoch 24 Batch 30/162] avg loss 0.00501372, throughput 9.61769K wps
[Epoch 24 Batch 60/162] avg loss 0.00498816, throughput 9.49967K wps
[Epoch 24 Batch 90/162] avg loss 0.00493116, throughput 9.38085K wps
[Epoch 24 Batch 120/162] avg loss 0.00487672, throughput 9.32345K wps
[Epoch 24 Batch 150/162] avg loss 0.00492319, throughput 9.31101K wps
Begin Testing...
[Epoch 24] train avg loss 0.00497444, dev acc 0.8944, dev avg loss 0.264164, throughput 9.42631K wps
[Epoch 25 Batch 30/162] avg loss 0.00486726, throughput 9.85552K wps
[Epoch 25 Batch 60/162] avg loss 0.00516029, throughput 9.56626K wps
[Epoch 25 Batch 90/162] avg loss 0.00450931, throughput 9.33615K wps
[Epoch 25 Batch 120/162] avg loss 0.00473615, throughput 9.3814K wps
[Epoch 25 Batch 150/162] avg loss 0.00511951, throughput 9.32826K wps
Begin Testing...
[Epoch 25] train avg loss 0.00490196, dev acc 0.8933, dev avg loss 0.261544, throughput 9.48674K wps
[Epoch 26 Batch 30/162] avg loss 0.00446047, throughput 9.84988K wps
[Epoch 26 Batch 60/162] avg loss 0.00487787, throughput 9.34218K wps
[Epoch 26 Batch 90/162] avg loss 0.00461985, throughput 9.30203K wps
[Epoch 26 Batch 120/162] avg loss 0.00469245, throughput 9.32401K wps
[Epoch 26 Batch 150/162] avg loss 0.00490017, throughput 9.47403K wps
Begin Testing...
[Epoch 26] train avg loss 0.00473548, dev acc 0.8967, dev avg loss 0.259237, throughput 9.45444K wps
Observed Improvement.
Begin Testing...
[Epoch 27 Batch 30/162] avg loss 0.00474389, throughput 9.65827K wps
[Epoch 27 Batch 60/162] avg loss 0.00467055, throughput 9.41117K wps
[Epoch 27 Batch 90/162] avg loss 0.0048706, throughput 9.38336K wps
[Epoch 27 Batch 120/162] avg loss 0.00466006, throughput 9.52748K wps
[Epoch 27 Batch 150/162] avg loss 0.00474872, throughput 9.48063K wps
Begin Testing...
[Epoch 27] train avg loss 0.00472714, dev acc 0.8956, dev avg loss 0.258566, throughput 9.49305K wps
[Epoch 28 Batch 30/162] avg loss 0.00462495, throughput 9.57619K wps
[Epoch 28 Batch 60/162] avg loss 0.0046153, throughput 9.38147K wps
[Epoch 28 Batch 90/162] avg loss 0.00470163, throughput 9.47896K wps
[Epoch 28 Batch 120/162] avg loss 0.00473551, throughput 9.41783K wps
[Epoch 28 Batch 150/162] avg loss 0.00455591, throughput 9.50105K wps
Begin Testing...
[Epoch 28] train avg loss 0.00467211, dev acc 0.8978, dev avg loss 0.256627, throughput 9.45986K wps
Observed Improvement.
Begin Testing...
[Epoch 29 Batch 30/162] avg loss 0.00472766, throughput 9.64378K wps
[Epoch 29 Batch 60/162] avg loss 0.0042063, throughput 9.61137K wps
[Epoch 29 Batch 90/162] avg loss 0.00442995, throughput 9.30553K wps
[Epoch 29 Batch 120/162] avg loss 0.00449417, throughput 9.47539K wps
[Epoch 29 Batch 150/162] avg loss 0.00457305, throughput 9.55199K wps
Begin Testing...
[Epoch 29] train avg loss 0.00453158, dev acc 0.9000, dev avg loss 0.253277, throughput 9.52575K wps
Observed Improvement.
Begin Testing...
[Epoch 30 Batch 30/162] avg loss 0.00462067, throughput 9.73271K wps
[Epoch 30 Batch 60/162] avg loss 0.00464152, throughput 9.26032K wps
[Epoch 30 Batch 90/162] avg loss 0.00419606, throughput 9.43567K wps
[Epoch 30 Batch 120/162] avg loss 0.00469395, throughput 9.53936K wps
[Epoch 30 Batch 150/162] avg loss 0.0042287, throughput 9.30076K wps
Begin Testing...
[Epoch 30] train avg loss 0.00444922, dev acc 0.9022, dev avg loss 0.251687, throughput 9.43252K wps
Observed Improvement.
Begin Testing...
[Epoch 31 Batch 30/162] avg loss 0.00434919, throughput 9.63879K wps
[Epoch 31 Batch 60/162] avg loss 0.00459829, throughput 9.35843K wps
[Epoch 31 Batch 90/162] avg loss 0.00438623, throughput 9.55179K wps
[Epoch 31 Batch 120/162] avg loss 0.00415008, throughput 9.42293K wps
[Epoch 31 Batch 150/162] avg loss 0.00430403, throughput 9.38469K wps
Begin Testing...
[Epoch 31] train avg loss 0.00438566, dev acc 0.9000, dev avg loss 0.250389, throughput 9.45939K wps
[Epoch 32 Batch 30/162] avg loss 0.00436757, throughput 9.51719K wps
[Epoch 32 Batch 60/162] avg loss 0.00437084, throughput 9.51616K wps
[Epoch 32 Batch 90/162] avg loss 0.00443524, throughput 9.27481K wps
[Epoch 32 Batch 120/162] avg loss 0.00432398, throughput 9.37208K wps
[Epoch 32 Batch 150/162] avg loss 0.00406764, throughput 9.45409K wps
Begin Testing...
[Epoch 32] train avg loss 0.00433712, dev acc 0.9000, dev avg loss 0.249125, throughput 9.43672K wps
[Epoch 33 Batch 30/162] avg loss 0.00410179, throughput 9.50759K wps
[Epoch 33 Batch 60/162] avg loss 0.00424083, throughput 9.51266K wps
[Epoch 33 Batch 90/162] avg loss 0.0044444, throughput 9.269K wps
[Epoch 33 Batch 120/162] avg loss 0.00392629, throughput 9.33449K wps
[Epoch 33 Batch 150/162] avg loss 0.00445924, throughput 9.39245K wps
Begin Testing...
[Epoch 33] train avg loss 0.00420483, dev acc 0.8978, dev avg loss 0.248724, throughput 9.38793K wps
[Epoch 34 Batch 30/162] avg loss 0.00399885, throughput 9.63814K wps
[Epoch 34 Batch 60/162] avg loss 0.00396069, throughput 9.42657K wps
[Epoch 34 Batch 90/162] avg loss 0.00434809, throughput 9.42066K wps
[Epoch 34 Batch 120/162] avg loss 0.0043975, throughput 9.30514K wps
[Epoch 34 Batch 150/162] avg loss 0.00395348, throughput 9.56087K wps
Begin Testing...
[Epoch 34] train avg loss 0.00412942, dev acc 0.9033, dev avg loss 0.245955, throughput 9.46847K wps
Observed Improvement.
Begin Testing...
[Epoch 35 Batch 30/162] avg loss 0.00417799, throughput 9.50394K wps
[Epoch 35 Batch 60/162] avg loss 0.00408284, throughput 9.51067K wps
[Epoch 35 Batch 90/162] avg loss 0.00444418, throughput 9.38307K wps
[Epoch 35 Batch 120/162] avg loss 0.00407536, throughput 9.33818K wps
[Epoch 35 Batch 150/162] avg loss 0.00391685, throughput 9.34936K wps
Begin Testing...
[Epoch 35] train avg loss 0.00414112, dev acc 0.9011, dev avg loss 0.245013, throughput 9.40848K wps
[Epoch 36 Batch 30/162] avg loss 0.00424228, throughput 9.59567K wps
[Epoch 36 Batch 60/162] avg loss 0.00398136, throughput 9.31027K wps
[Epoch 36 Batch 90/162] avg loss 0.00397867, throughput 9.25303K wps
[Epoch 36 Batch 120/162] avg loss 0.00374787, throughput 9.56874K wps
[Epoch 36 Batch 150/162] avg loss 0.00403141, throughput 9.32688K wps
Begin Testing...
[Epoch 36] train avg loss 0.00406164, dev acc 0.9033, dev avg loss 0.243502, throughput 9.39929K wps
Observed Improvement.
Begin Testing...
[Epoch 37 Batch 30/162] avg loss 0.00406017, throughput 9.4679K wps
[Epoch 37 Batch 60/162] avg loss 0.00408843, throughput 9.44796K wps
[Epoch 37 Batch 90/162] avg loss 0.0040087, throughput 9.39715K wps
[Epoch 37 Batch 120/162] avg loss 0.00397594, throughput 9.48287K wps
[Epoch 37 Batch 150/162] avg loss 0.00385451, throughput 9.2435K wps
Begin Testing...
[Epoch 37] train avg loss 0.00400456, dev acc 0.9044, dev avg loss 0.242278, throughput 9.40917K wps
Observed Improvement.
Begin Testing...
[Epoch 38 Batch 30/162] avg loss 0.00383101, throughput 9.77135K wps
[Epoch 38 Batch 60/162] avg loss 0.00409139, throughput 9.40162K wps
[Epoch 38 Batch 90/162] avg loss 0.00386445, throughput 9.38177K wps
[Epoch 38 Batch 120/162] avg loss 0.00379016, throughput 9.39544K wps
[Epoch 38 Batch 150/162] avg loss 0.00368953, throughput 9.53598K wps
Begin Testing...
[Epoch 38] train avg loss 0.00388143, dev acc 0.8944, dev avg loss 0.244049, throughput 9.48127K wps
[Epoch 39 Batch 30/162] avg loss 0.0037532, throughput 9.51779K wps
[Epoch 39 Batch 60/162] avg loss 0.00379816, throughput 9.46835K wps
[Epoch 39 Batch 90/162] avg loss 0.00388524, throughput 9.42865K wps
[Epoch 39 Batch 120/162] avg loss 0.00371377, throughput 9.41585K wps
[Epoch 39 Batch 150/162] avg loss 0.00442531, throughput 9.45539K wps
Begin Testing...
[Epoch 39] train avg loss 0.00390188, dev acc 0.9033, dev avg loss 0.240494, throughput 9.46116K wps
[Epoch 40 Batch 30/162] avg loss 0.0038015, throughput 9.58085K wps
[Epoch 40 Batch 60/162] avg loss 0.00378343, throughput 9.44258K wps
[Epoch 40 Batch 90/162] avg loss 0.00384548, throughput 9.52299K wps
[Epoch 40 Batch 120/162] avg loss 0.0038396, throughput 9.49287K wps
[Epoch 40 Batch 150/162] avg loss 0.00376777, throughput 9.25596K wps
Begin Testing...
[Epoch 40] train avg loss 0.00379062, dev acc 0.9056, dev avg loss 0.239217, throughput 9.44287K wps
Observed Improvement.
Begin Testing...
[Epoch 41 Batch 30/162] avg loss 0.00398003, throughput 9.62944K wps
[Epoch 41 Batch 60/162] avg loss 0.00407411, throughput 9.52752K wps
[Epoch 41 Batch 90/162] avg loss 0.00378415, throughput 9.31445K wps
[Epoch 41 Batch 120/162] avg loss 0.00389275, throughput 9.35069K wps
[Epoch 41 Batch 150/162] avg loss 0.00326517, throughput 9.42276K wps
Begin Testing...
[Epoch 41] train avg loss 0.00377874, dev acc 0.9000, dev avg loss 0.239018, throughput 9.44284K wps
[Epoch 42 Batch 30/162] avg loss 0.00361527, throughput 9.75476K wps
[Epoch 42 Batch 60/162] avg loss 0.00354684, throughput 9.49498K wps
[Epoch 42 Batch 90/162] avg loss 0.00349292, throughput 9.43857K wps
[Epoch 42 Batch 120/162] avg loss 0.00345702, throughput 9.37229K wps
[Epoch 42 Batch 150/162] avg loss 0.00389636, throughput 9.49854K wps
Begin Testing...
[Epoch 42] train avg loss 0.00364849, dev acc 0.8978, dev avg loss 0.238587, throughput 9.4972K wps
[Epoch 43 Batch 30/162] avg loss 0.00379318, throughput 9.46017K wps
[Epoch 43 Batch 60/162] avg loss 0.00362616, throughput 9.31213K wps
[Epoch 43 Batch 90/162] avg loss 0.00387291, throughput 9.40535K wps
[Epoch 43 Batch 120/162] avg loss 0.00340377, throughput 9.51385K wps
[Epoch 43 Batch 150/162] avg loss 0.00354426, throughput 9.63022K wps
Begin Testing...
[Epoch 43] train avg loss 0.00362999, dev acc 0.8989, dev avg loss 0.236725, throughput 9.46503K wps
[Epoch 44 Batch 30/162] avg loss 0.00396009, throughput 9.45772K wps
[Epoch 44 Batch 60/162] avg loss 0.0034284, throughput 9.49513K wps
[Epoch 44 Batch 90/162] avg loss 0.0031975, throughput 9.4879K wps
[Epoch 44 Batch 120/162] avg loss 0.00366218, throughput 9.40196K wps
[Epoch 44 Batch 150/162] avg loss 0.00370804, throughput 9.36773K wps
Begin Testing...
[Epoch 44] train avg loss 0.00359447, dev acc 0.9022, dev avg loss 0.235644, throughput 9.45266K wps
[Epoch 45 Batch 30/162] avg loss 0.00357474, throughput 9.78093K wps
[Epoch 45 Batch 60/162] avg loss 0.00364399, throughput 9.49177K wps
[Epoch 45 Batch 90/162] avg loss 0.00334855, throughput 9.27617K wps
[Epoch 45 Batch 120/162] avg loss 0.00325552, throughput 9.44985K wps
[Epoch 45 Batch 150/162] avg loss 0.00358472, throughput 9.34806K wps
Begin Testing...
[Epoch 45] train avg loss 0.0035184, dev acc 0.9000, dev avg loss 0.235355, throughput 9.47759K wps
[Epoch 46 Batch 30/162] avg loss 0.00347074, throughput 9.68854K wps
[Epoch 46 Batch 60/162] avg loss 0.00338562, throughput 9.31762K wps
[Epoch 46 Batch 90/162] avg loss 0.00369921, throughput 9.41702K wps
[Epoch 46 Batch 120/162] avg loss 0.00359095, throughput 9.56828K wps
[Epoch 46 Batch 150/162] avg loss 0.00354815, throughput 9.22743K wps
Begin Testing...
[Epoch 46] train avg loss 0.00349879, dev acc 0.8989, dev avg loss 0.236272, throughput 9.45936K wps
[Epoch 47 Batch 30/162] avg loss 0.00325151, throughput 9.59809K wps
[Epoch 47 Batch 60/162] avg loss 0.00327018, throughput 9.21793K wps
[Epoch 47 Batch 90/162] avg loss 0.00346037, throughput 9.66793K wps
[Epoch 47 Batch 120/162] avg loss 0.00353189, throughput 9.40153K wps
[Epoch 47 Batch 150/162] avg loss 0.00370354, throughput 9.59388K wps
Begin Testing...
[Epoch 47] train avg loss 0.00344402, dev acc 0.9000, dev avg loss 0.234614, throughput 9.49239K wps
[Epoch 48 Batch 30/162] avg loss 0.0036105, throughput 9.73533K wps
[Epoch 48 Batch 60/162] avg loss 0.00352036, throughput 9.50696K wps
[Epoch 48 Batch 90/162] avg loss 0.00315668, throughput 9.60011K wps
[Epoch 48 Batch 120/162] avg loss 0.00329366, throughput 9.53689K wps
[Epoch 48 Batch 150/162] avg loss 0.00327449, throughput 9.4122K wps
Begin Testing...
[Epoch 48] train avg loss 0.00335908, dev acc 0.8989, dev avg loss 0.233745, throughput 9.55769K wps
[Epoch 49 Batch 30/162] avg loss 0.00343152, throughput 9.54808K wps
[Epoch 49 Batch 60/162] avg loss 0.00348162, throughput 9.30959K wps
[Epoch 49 Batch 90/162] avg loss 0.00330839, throughput 9.54341K wps
[Epoch 49 Batch 120/162] avg loss 0.00358948, throughput 9.49447K wps
[Epoch 49 Batch 150/162] avg loss 0.00334725, throughput 9.37539K wps
Begin Testing...
[Epoch 49] train avg loss 0.0033875, dev acc 0.8978, dev avg loss 0.233107, throughput 9.45418K wps
[Epoch 50 Batch 30/162] avg loss 0.00318679, throughput 9.72929K wps
[Epoch 50 Batch 60/162] avg loss 0.00312677, throughput 9.26746K wps
[Epoch 50 Batch 90/162] avg loss 0.00354316, throughput 9.41891K wps
[Epoch 50 Batch 120/162] avg loss 0.00328846, throughput 9.46169K wps
[Epoch 50 Batch 150/162] avg loss 0.00304651, throughput 9.46377K wps
Begin Testing...
[Epoch 50] train avg loss 0.00322728, dev acc 0.9000, dev avg loss 0.232109, throughput 9.4782K wps
[Epoch 51 Batch 30/162] avg loss 0.003414, throughput 9.77515K wps
[Epoch 51 Batch 60/162] avg loss 0.00298916, throughput 9.55101K wps
[Epoch 51 Batch 90/162] avg loss 0.00327826, throughput 9.48002K wps
[Epoch 51 Batch 120/162] avg loss 0.0031996, throughput 9.21466K wps
[Epoch 51 Batch 150/162] avg loss 0.00322601, throughput 9.68212K wps
Begin Testing...
[Epoch 51] train avg loss 0.00324641, dev acc 0.9067, dev avg loss 0.231436, throughput 9.52582K wps
Observed Improvement.
Begin Testing...
[Epoch 52 Batch 30/162] avg loss 0.00305027, throughput 9.65718K wps
[Epoch 52 Batch 60/162] avg loss 0.00293469, throughput 9.49973K wps
[Epoch 52 Batch 90/162] avg loss 0.00345277, throughput 9.5669K wps
[Epoch 52 Batch 120/162] avg loss 0.00282883, throughput 9.23366K wps
[Epoch 52 Batch 150/162] avg loss 0.00334408, throughput 9.40331K wps
Begin Testing...
[Epoch 52] train avg loss 0.00312488, dev acc 0.9000, dev avg loss 0.231759, throughput 9.46329K wps
[Epoch 53 Batch 30/162] avg loss 0.00307341, throughput 9.7413K wps
[Epoch 53 Batch 60/162] avg loss 0.00286474, throughput 9.60575K wps
[Epoch 53 Batch 90/162] avg loss 0.00335966, throughput 9.46491K wps
[Epoch 53 Batch 120/162] avg loss 0.00336166, throughput 9.56682K wps
[Epoch 53 Batch 150/162] avg loss 0.00305757, throughput 9.31127K wps
Begin Testing...
[Epoch 53] train avg loss 0.0031412, dev acc 0.9033, dev avg loss 0.230269, throughput 9.53474K wps
[Epoch 54 Batch 30/162] avg loss 0.00311497, throughput 9.64457K wps
[Epoch 54 Batch 60/162] avg loss 0.00310818, throughput 9.46324K wps
[Epoch 54 Batch 90/162] avg loss 0.00297833, throughput 9.33916K wps
[Epoch 54 Batch 120/162] avg loss 0.00331121, throughput 9.46711K wps
[Epoch 54 Batch 150/162] avg loss 0.00310923, throughput 9.61887K wps
Begin Testing...
[Epoch 54] train avg loss 0.00310267, dev acc 0.9089, dev avg loss 0.229693, throughput 9.50547K wps
Observed Improvement.
Begin Testing...
[Epoch 55 Batch 30/162] avg loss 0.00286058, throughput 9.70563K wps
[Epoch 55 Batch 60/162] avg loss 0.00332626, throughput 9.47438K wps
[Epoch 55 Batch 90/162] avg loss 0.00275379, throughput 9.4881K wps
[Epoch 55 Batch 120/162] avg loss 0.00282436, throughput 9.49382K wps
[Epoch 55 Batch 150/162] avg loss 0.00318199, throughput 9.42593K wps
Begin Testing...
[Epoch 55] train avg loss 0.0030103, dev acc 0.9000, dev avg loss 0.230146, throughput 9.49507K wps
[Epoch 56 Batch 30/162] avg loss 0.00301154, throughput 9.68543K wps
[Epoch 56 Batch 60/162] avg loss 0.00283591, throughput 9.5704K wps
[Epoch 56 Batch 90/162] avg loss 0.00319223, throughput 9.33979K wps
[Epoch 56 Batch 120/162] avg loss 0.00281098, throughput 9.42443K wps
[Epoch 56 Batch 150/162] avg loss 0.00316716, throughput 9.46798K wps
Begin Testing...
[Epoch 56] train avg loss 0.00300351, dev acc 0.9056, dev avg loss 0.229138, throughput 9.48071K wps
[Epoch 57 Batch 30/162] avg loss 0.00287731, throughput 9.49244K wps
[Epoch 57 Batch 60/162] avg loss 0.00317882, throughput 9.43373K wps
[Epoch 57 Batch 90/162] avg loss 0.00294749, throughput 9.31209K wps
[Epoch 57 Batch 120/162] avg loss 0.00263622, throughput 9.48465K wps
[Epoch 57 Batch 150/162] avg loss 0.0030973, throughput 9.42371K wps
Begin Testing...
[Epoch 57] train avg loss 0.00295966, dev acc 0.9111, dev avg loss 0.228489, throughput 9.44395K wps
Observed Improvement.
Begin Testing...
[Epoch 58 Batch 30/162] avg loss 0.0029442, throughput 9.78403K wps
[Epoch 58 Batch 60/162] avg loss 0.0027786, throughput 9.47745K wps
[Epoch 58 Batch 90/162] avg loss 0.00281949, throughput 9.44369K wps
[Epoch 58 Batch 120/162] avg loss 0.00283737, throughput 9.46841K wps
[Epoch 58 Batch 150/162] avg loss 0.00300873, throughput 9.35371K wps
Begin Testing...
[Epoch 58] train avg loss 0.00288692, dev acc 0.9078, dev avg loss 0.22807, throughput 9.50297K wps
[Epoch 59 Batch 30/162] avg loss 0.00305366, throughput 9.68844K wps
[Epoch 59 Batch 60/162] avg loss 0.00281157, throughput 9.30583K wps
[Epoch 59 Batch 90/162] avg loss 0.00280737, throughput 9.32667K wps
[Epoch 59 Batch 120/162] avg loss 0.00278681, throughput 9.30934K wps
[Epoch 59 Batch 150/162] avg loss 0.00298906, throughput 9.36472K wps
Begin Testing...
[Epoch 59] train avg loss 0.00288401, dev acc 0.9078, dev avg loss 0.2279, throughput 9.4044K wps
[Epoch 60 Batch 30/162] avg loss 0.0026235, throughput 9.52984K wps
[Epoch 60 Batch 60/162] avg loss 0.00286482, throughput 9.53731K wps
[Epoch 60 Batch 90/162] avg loss 0.00281531, throughput 9.41811K wps
[Epoch 60 Batch 120/162] avg loss 0.00277027, throughput 9.37532K wps
[Epoch 60 Batch 150/162] avg loss 0.00283581, throughput 9.36994K wps
Begin Testing...
[Epoch 60] train avg loss 0.00279524, dev acc 0.9100, dev avg loss 0.227302, throughput 9.44034K wps
[Epoch 61 Batch 30/162] avg loss 0.00270716, throughput 9.58616K wps
[Epoch 61 Batch 60/162] avg loss 0.00279457, throughput 9.35632K wps
[Epoch 61 Batch 90/162] avg loss 0.00276284, throughput 9.5807K wps
[Epoch 61 Batch 120/162] avg loss 0.00286912, throughput 9.30973K wps
[Epoch 61 Batch 150/162] avg loss 0.00256892, throughput 9.36881K wps
Begin Testing...
[Epoch 61] train avg loss 0.00275048, dev acc 0.9078, dev avg loss 0.227422, throughput 9.43173K wps
[Epoch 62 Batch 30/162] avg loss 0.00248744, throughput 9.71855K wps
[Epoch 62 Batch 60/162] avg loss 0.00272149, throughput 9.35906K wps
[Epoch 62 Batch 90/162] avg loss 0.00294276, throughput 9.50321K wps
[Epoch 62 Batch 120/162] avg loss 0.00283565, throughput 9.25574K wps
[Epoch 62 Batch 150/162] avg loss 0.00259692, throughput 9.48659K wps
Begin Testing...
[Epoch 62] train avg loss 0.00273754, dev acc 0.9100, dev avg loss 0.226532, throughput 9.46136K wps
[Epoch 63 Batch 30/162] avg loss 0.00236638, throughput 9.45141K wps
[Epoch 63 Batch 60/162] avg loss 0.00294246, throughput 9.50591K wps
[Epoch 63 Batch 90/162] avg loss 0.00267498, throughput 9.60065K wps
[Epoch 63 Batch 120/162] avg loss 0.00293376, throughput 9.34343K wps
[Epoch 63 Batch 150/162] avg loss 0.00284816, throughput 9.28835K wps
Begin Testing...
[Epoch 63] train avg loss 0.00276561, dev acc 0.9122, dev avg loss 0.226463, throughput 9.43645K wps
Observed Improvement.
Begin Testing...
[Epoch 64 Batch 30/162] avg loss 0.00293538, throughput 9.55796K wps
[Epoch 64 Batch 60/162] avg loss 0.00261848, throughput 9.56516K wps
[Epoch 64 Batch 90/162] avg loss 0.00259867, throughput 9.31053K wps
[Epoch 64 Batch 120/162] avg loss 0.00262477, throughput 9.42515K wps
[Epoch 64 Batch 150/162] avg loss 0.00269427, throughput 9.56819K wps
Begin Testing...
[Epoch 64] train avg loss 0.00268228, dev acc 0.9089, dev avg loss 0.226177, throughput 9.49017K wps
[Epoch 65 Batch 30/162] avg loss 0.00249016, throughput 9.60541K wps
[Epoch 65 Batch 60/162] avg loss 0.00302549, throughput 9.59303K wps
[Epoch 65 Batch 90/162] avg loss 0.00261301, throughput 9.31489K wps
[Epoch 65 Batch 120/162] avg loss 0.00253634, throughput 9.50613K wps
[Epoch 65 Batch 150/162] avg loss 0.00281672, throughput 9.35304K wps
Begin Testing...
[Epoch 65] train avg loss 0.00266594, dev acc 0.9111, dev avg loss 0.225831, throughput 9.46192K wps
[Epoch 66 Batch 30/162] avg loss 0.00239761, throughput 9.76571K wps
[Epoch 66 Batch 60/162] avg loss 0.00271206, throughput 9.41482K wps
[Epoch 66 Batch 90/162] avg loss 0.00289738, throughput 9.31728K wps
[Epoch 66 Batch 120/162] avg loss 0.00253317, throughput 9.39058K wps
[Epoch 66 Batch 150/162] avg loss 0.00248738, throughput 9.42371K wps
Begin Testing...
[Epoch 66] train avg loss 0.00259447, dev acc 0.9033, dev avg loss 0.230045, throughput 9.46626K wps
[Epoch 67 Batch 30/162] avg loss 0.00254377, throughput 9.86589K wps
[Epoch 67 Batch 60/162] avg loss 0.00253961, throughput 9.28508K wps
[Epoch 67 Batch 90/162] avg loss 0.00256376, throughput 9.50805K wps
[Epoch 67 Batch 120/162] avg loss 0.00261483, throughput 9.37429K wps
[Epoch 67 Batch 150/162] avg loss 0.00276034, throughput 9.39358K wps
Begin Testing...
[Epoch 67] train avg loss 0.00261711, dev acc 0.9111, dev avg loss 0.225243, throughput 9.49742K wps
[Epoch 68 Batch 30/162] avg loss 0.00228829, throughput 9.70855K wps
[Epoch 68 Batch 60/162] avg loss 0.00262099, throughput 9.54333K wps
[Epoch 68 Batch 90/162] avg loss 0.00266993, throughput 9.48442K wps
[Epoch 68 Batch 120/162] avg loss 0.00248143, throughput 9.32306K wps
[Epoch 68 Batch 150/162] avg loss 0.00252137, throughput 9.40946K wps
Begin Testing...
[Epoch 68] train avg loss 0.00251491, dev acc 0.9100, dev avg loss 0.225425, throughput 9.4809K wps
[Epoch 69 Batch 30/162] avg loss 0.00234518, throughput 9.67962K wps
[Epoch 69 Batch 60/162] avg loss 0.00243195, throughput 9.22304K wps
[Epoch 69 Batch 90/162] avg loss 0.00248518, throughput 9.37461K wps
[Epoch 69 Batch 120/162] avg loss 0.00267432, throughput 9.41972K wps
[Epoch 69 Batch 150/162] avg loss 0.00277203, throughput 9.54107K wps
Begin Testing...
[Epoch 69] train avg loss 0.00251558, dev acc 0.9111, dev avg loss 0.22453, throughput 9.43924K wps
[Epoch 70 Batch 30/162] avg loss 0.00247567, throughput 9.52382K wps
[Epoch 70 Batch 60/162] avg loss 0.00230529, throughput 9.36437K wps
[Epoch 70 Batch 90/162] avg loss 0.00244642, throughput 9.39522K wps
[Epoch 70 Batch 120/162] avg loss 0.00222048, throughput 9.51204K wps
[Epoch 70 Batch 150/162] avg loss 0.00258221, throughput 9.33785K wps
Begin Testing...
[Epoch 70] train avg loss 0.00240433, dev acc 0.9111, dev avg loss 0.224213, throughput 9.41698K wps
[Epoch 71 Batch 30/162] avg loss 0.00242557, throughput 9.80517K wps
[Epoch 71 Batch 60/162] avg loss 0.00232819, throughput 9.23498K wps
[Epoch 71 Batch 90/162] avg loss 0.00263671, throughput 9.48118K wps
[Epoch 71 Batch 120/162] avg loss 0.00240075, throughput 9.39537K wps
[Epoch 71 Batch 150/162] avg loss 0.00254743, throughput 9.41828K wps
Begin Testing...
[Epoch 71] train avg loss 0.00245461, dev acc 0.9100, dev avg loss 0.224458, throughput 9.4696K wps
[Epoch 72 Batch 30/162] avg loss 0.00228725, throughput 9.51625K wps
[Epoch 72 Batch 60/162] avg loss 0.00252462, throughput 9.48741K wps
[Epoch 72 Batch 90/162] avg loss 0.00261999, throughput 9.39599K wps
[Epoch 72 Batch 120/162] avg loss 0.00232309, throughput 9.41302K wps
[Epoch 72 Batch 150/162] avg loss 0.00230151, throughput 9.32338K wps
Begin Testing...
[Epoch 72] train avg loss 0.00242127, dev acc 0.9100, dev avg loss 0.224883, throughput 9.41607K wps
[Epoch 73 Batch 30/162] avg loss 0.00224907, throughput 9.46832K wps
[Epoch 73 Batch 60/162] avg loss 0.00240005, throughput 9.39325K wps
[Epoch 73 Batch 90/162] avg loss 0.00229851, throughput 9.48258K wps
[Epoch 73 Batch 120/162] avg loss 0.00236778, throughput 9.44574K wps
[Epoch 73 Batch 150/162] avg loss 0.00222843, throughput 9.37206K wps
Begin Testing...
[Epoch 73] train avg loss 0.00233474, dev acc 0.9089, dev avg loss 0.224555, throughput 9.41664K wps
[Epoch 74 Batch 30/162] avg loss 0.00232833, throughput 9.61462K wps
[Epoch 74 Batch 60/162] avg loss 0.00231794, throughput 9.51361K wps
[Epoch 74 Batch 90/162] avg loss 0.00218228, throughput 9.5558K wps
[Epoch 74 Batch 120/162] avg loss 0.00231879, throughput 9.28221K wps
[Epoch 74 Batch 150/162] avg loss 0.00228815, throughput 9.37871K wps
Begin Testing...
[Epoch 74] train avg loss 0.0022842, dev acc 0.9100, dev avg loss 0.223896, throughput 9.45028K wps
[Epoch 75 Batch 30/162] avg loss 0.0024462, throughput 9.68592K wps
[Epoch 75 Batch 60/162] avg loss 0.00217882, throughput 9.50356K wps
[Epoch 75 Batch 90/162] avg loss 0.00212187, throughput 9.44907K wps
[Epoch 75 Batch 120/162] avg loss 0.00222226, throughput 9.47761K wps
[Epoch 75 Batch 150/162] avg loss 0.00242205, throughput 9.58809K wps
Begin Testing...
[Epoch 75] train avg loss 0.00228155, dev acc 0.9100, dev avg loss 0.224699, throughput 9.52677K wps
[Epoch 76 Batch 30/162] avg loss 0.00249663, throughput 9.63667K wps
[Epoch 76 Batch 60/162] avg loss 0.00216676, throughput 9.44122K wps
[Epoch 76 Batch 90/162] avg loss 0.00195749, throughput 9.622K wps
[Epoch 76 Batch 120/162] avg loss 0.00215661, throughput 9.33616K wps
[Epoch 76 Batch 150/162] avg loss 0.00224911, throughput 9.35188K wps
Begin Testing...
[Epoch 76] train avg loss 0.00221135, dev acc 0.9100, dev avg loss 0.223008, throughput 9.46154K wps
[Epoch 77 Batch 30/162] avg loss 0.0020198, throughput 9.54873K wps
[Epoch 77 Batch 60/162] avg loss 0.00233853, throughput 9.50284K wps
[Epoch 77 Batch 90/162] avg loss 0.00218323, throughput 9.27507K wps
[Epoch 77 Batch 120/162] avg loss 0.00235394, throughput 9.44151K wps
[Epoch 77 Batch 150/162] avg loss 0.00230878, throughput 9.31058K wps
Begin Testing...
[Epoch 77] train avg loss 0.00223099, dev acc 0.9111, dev avg loss 0.224137, throughput 9.43833K wps
[Epoch 78 Batch 30/162] avg loss 0.00211856, throughput 9.77398K wps
[Epoch 78 Batch 60/162] avg loss 0.00229744, throughput 9.33926K wps
[Epoch 78 Batch 90/162] avg loss 0.00196463, throughput 9.31801K wps
[Epoch 78 Batch 120/162] avg loss 0.00218372, throughput 9.43957K wps
[Epoch 78 Batch 150/162] avg loss 0.00207106, throughput 9.36198K wps
Begin Testing...
[Epoch 78] train avg loss 0.00215366, dev acc 0.9111, dev avg loss 0.223709, throughput 9.43705K wps
[Epoch 79 Batch 30/162] avg loss 0.00205206, throughput 9.63673K wps
[Epoch 79 Batch 60/162] avg loss 0.00214006, throughput 9.25525K wps
[Epoch 79 Batch 90/162] avg loss 0.00222182, throughput 9.32275K wps
[Epoch 79 Batch 120/162] avg loss 0.00210921, throughput 9.35928K wps
[Epoch 79 Batch 150/162] avg loss 0.00229708, throughput 9.42654K wps
Begin Testing...
[Epoch 79] train avg loss 0.00216304, dev acc 0.9100, dev avg loss 0.22304, throughput 9.38989K wps
[Epoch 80 Batch 30/162] avg loss 0.00202311, throughput 9.50889K wps
[Epoch 80 Batch 60/162] avg loss 0.00225087, throughput 9.44872K wps
[Epoch 80 Batch 90/162] avg loss 0.00239755, throughput 9.2521K wps
[Epoch 80 Batch 120/162] avg loss 0.00199517, throughput 9.40409K wps
[Epoch 80 Batch 150/162] avg loss 0.0020565, throughput 9.43119K wps
Begin Testing...
[Epoch 80] train avg loss 0.00212634, dev acc 0.9111, dev avg loss 0.223501, throughput 9.39297K wps
[Epoch 81 Batch 30/162] avg loss 0.0020893, throughput 9.58662K wps
[Epoch 81 Batch 60/162] avg loss 0.00225277, throughput 9.51561K wps
[Epoch 81 Batch 90/162] avg loss 0.00241267, throughput 9.45238K wps
[Epoch 81 Batch 120/162] avg loss 0.00201843, throughput 9.5266K wps
[Epoch 81 Batch 150/162] avg loss 0.0020573, throughput 9.45278K wps
Begin Testing...
[Epoch 81] train avg loss 0.00215459, dev acc 0.9111, dev avg loss 0.223144, throughput 9.52044K wps
[Epoch 82 Batch 30/162] avg loss 0.00201094, throughput 9.85452K wps
[Epoch 82 Batch 60/162] avg loss 0.00202922, throughput 9.25692K wps
[Epoch 82 Batch 90/162] avg loss 0.00204882, throughput 9.45211K wps
[Epoch 82 Batch 120/162] avg loss 0.0022583, throughput 9.23976K wps
[Epoch 82 Batch 150/162] avg loss 0.00202517, throughput 9.60191K wps
Begin Testing...
[Epoch 82] train avg loss 0.00209553, dev acc 0.9078, dev avg loss 0.227246, throughput 9.45543K wps
[Epoch 83 Batch 30/162] avg loss 0.00215706, throughput 9.44428K wps
[Epoch 83 Batch 60/162] avg loss 0.00201779, throughput 9.41742K wps
[Epoch 83 Batch 90/162] avg loss 0.00205798, throughput 9.44165K wps
[Epoch 83 Batch 120/162] avg loss 0.00217492, throughput 9.34198K wps
[Epoch 83 Batch 150/162] avg loss 0.00187376, throughput 9.32838K wps
Begin Testing...
[Epoch 83] train avg loss 0.00204197, dev acc 0.9089, dev avg loss 0.222101, throughput 9.38972K wps
[Epoch 84 Batch 30/162] avg loss 0.00204796, throughput 9.67757K wps
[Epoch 84 Batch 60/162] avg loss 0.00207665, throughput 9.35004K wps
[Epoch 84 Batch 90/162] avg loss 0.00208488, throughput 9.45194K wps
[Epoch 84 Batch 120/162] avg loss 0.00182257, throughput 9.26619K wps
[Epoch 84 Batch 150/162] avg loss 0.00195526, throughput 9.38067K wps
Begin Testing...
[Epoch 84] train avg loss 0.00202065, dev acc 0.9100, dev avg loss 0.222275, throughput 9.44203K wps
[Epoch 85 Batch 30/162] avg loss 0.00180629, throughput 9.67023K wps
[Epoch 85 Batch 60/162] avg loss 0.00194523, throughput 9.57729K wps
[Epoch 85 Batch 90/162] avg loss 0.0021532, throughput 9.51072K wps
[Epoch 85 Batch 120/162] avg loss 0.00195081, throughput 9.33751K wps
[Epoch 85 Batch 150/162] avg loss 0.00192279, throughput 9.24812K wps
Begin Testing...
[Epoch 85] train avg loss 0.00196714, dev acc 0.9078, dev avg loss 0.221798, throughput 9.45448K wps
[Epoch 86 Batch 30/162] avg loss 0.00193722, throughput 9.62516K wps
[Epoch 86 Batch 60/162] avg loss 0.00216401, throughput 9.39089K wps
[Epoch 86 Batch 90/162] avg loss 0.00192283, throughput 9.39995K wps
[Epoch 86 Batch 120/162] avg loss 0.0018353, throughput 9.33341K wps
[Epoch 86 Batch 150/162] avg loss 0.0018735, throughput 9.40926K wps
Begin Testing...
[Epoch 86] train avg loss 0.00194238, dev acc 0.9089, dev avg loss 0.221897, throughput 9.42075K wps
[Epoch 87 Batch 30/162] avg loss 0.00179515, throughput 9.53764K wps
[Epoch 87 Batch 60/162] avg loss 0.00179959, throughput 9.44473K wps
[Epoch 87 Batch 90/162] avg loss 0.00179315, throughput 9.22261K wps
[Epoch 87 Batch 120/162] avg loss 0.00208908, throughput 9.51902K wps
[Epoch 87 Batch 150/162] avg loss 0.00177814, throughput 9.5499K wps
Begin Testing...
[Epoch 87] train avg loss 0.00186595, dev acc 0.9089, dev avg loss 0.221726, throughput 9.44214K wps
[Epoch 88 Batch 30/162] avg loss 0.00206384, throughput 9.46183K wps
[Epoch 88 Batch 60/162] avg loss 0.00193644, throughput 9.46066K wps
[Epoch 88 Batch 90/162] avg loss 0.00178946, throughput 9.38475K wps
[Epoch 88 Batch 120/162] avg loss 0.00173444, throughput 9.49063K wps
[Epoch 88 Batch 150/162] avg loss 0.00198767, throughput 9.42043K wps
Begin Testing...
[Epoch 88] train avg loss 0.00190996, dev acc 0.9100, dev avg loss 0.221572, throughput 9.44708K wps
[Epoch 89 Batch 30/162] avg loss 0.00180881, throughput 9.51313K wps
[Epoch 89 Batch 60/162] avg loss 0.00200014, throughput 9.54673K wps
[Epoch 89 Batch 90/162] avg loss 0.00176425, throughput 9.24385K wps
[Epoch 89 Batch 120/162] avg loss 0.00195862, throughput 9.41936K wps
[Epoch 89 Batch 150/162] avg loss 0.00171618, throughput 9.46988K wps
Begin Testing...
[Epoch 89] train avg loss 0.00184356, dev acc 0.9089, dev avg loss 0.222228, throughput 9.43837K wps
[Epoch 90 Batch 30/162] avg loss 0.00218814, throughput 9.56338K wps
[Epoch 90 Batch 60/162] avg loss 0.00196333, throughput 9.21385K wps
[Epoch 90 Batch 90/162] avg loss 0.00169889, throughput 9.3096K wps
[Epoch 90 Batch 120/162] avg loss 0.0018617, throughput 9.38755K wps
[Epoch 90 Batch 150/162] avg loss 0.0017196, throughput 9.27814K wps
Begin Testing...
[Epoch 90] train avg loss 0.00189339, dev acc 0.9133, dev avg loss 0.224403, throughput 9.37591K wps
Observed Improvement.
Begin Testing...
[Epoch 91 Batch 30/162] avg loss 0.00178204, throughput 9.43565K wps
[Epoch 91 Batch 60/162] avg loss 0.00180133, throughput 9.20636K wps
[Epoch 91 Batch 90/162] avg loss 0.00202408, throughput 9.3333K wps
[Epoch 91 Batch 120/162] avg loss 0.00183943, throughput 9.3836K wps
[Epoch 91 Batch 150/162] avg loss 0.00180091, throughput 9.6583K wps
Begin Testing...
[Epoch 91] train avg loss 0.00184832, dev acc 0.9122, dev avg loss 0.222735, throughput 9.40282K wps
[Epoch 92 Batch 30/162] avg loss 0.00177177, throughput 9.53119K wps
[Epoch 92 Batch 60/162] avg loss 0.00186256, throughput 9.57715K wps
[Epoch 92 Batch 90/162] avg loss 0.00175131, throughput 9.34072K wps
[Epoch 92 Batch 120/162] avg loss 0.00176492, throughput 9.45141K wps
[Epoch 92 Batch 150/162] avg loss 0.00189462, throughput 9.29823K wps
Begin Testing...
[Epoch 92] train avg loss 0.00180633, dev acc 0.9100, dev avg loss 0.221965, throughput 9.42535K wps
[Epoch 93 Batch 30/162] avg loss 0.00181206, throughput 9.55871K wps
[Epoch 93 Batch 60/162] avg loss 0.00173094, throughput 9.45121K wps
[Epoch 93 Batch 90/162] avg loss 0.00175647, throughput 9.39455K wps
[Epoch 93 Batch 120/162] avg loss 0.00184363, throughput 9.31601K wps
[Epoch 93 Batch 150/162] avg loss 0.00189581, throughput 9.36191K wps
Begin Testing...
[Epoch 93] train avg loss 0.00181814, dev acc 0.9100, dev avg loss 0.222432, throughput 9.40368K wps
[Epoch 94 Batch 30/162] avg loss 0.00167533, throughput 9.48769K wps
[Epoch 94 Batch 60/162] avg loss 0.00183529, throughput 9.3326K wps
[Epoch 94 Batch 90/162] avg loss 0.00185893, throughput 9.39838K wps
[Epoch 94 Batch 120/162] avg loss 0.00170035, throughput 9.45219K wps
[Epoch 94 Batch 150/162] avg loss 0.00170485, throughput 9.32434K wps
Begin Testing...
[Epoch 94] train avg loss 0.00175112, dev acc 0.9100, dev avg loss 0.222383, throughput 9.40985K wps
[Epoch 95 Batch 30/162] avg loss 0.00183116, throughput 9.6168K wps
[Epoch 95 Batch 60/162] avg loss 0.00176734, throughput 9.38818K wps
[Epoch 95 Batch 90/162] avg loss 0.0016383, throughput 9.26943K wps
[Epoch 95 Batch 120/162] avg loss 0.00170681, throughput 9.42025K wps
[Epoch 95 Batch 150/162] avg loss 0.00182716, throughput 9.27622K wps
Begin Testing...
[Epoch 95] train avg loss 0.00176316, dev acc 0.9111, dev avg loss 0.22182, throughput 9.39623K wps
[Epoch 96 Batch 30/162] avg loss 0.00178923, throughput 9.59054K wps
[Epoch 96 Batch 60/162] avg loss 0.00165117, throughput 9.32538K wps
[Epoch 96 Batch 90/162] avg loss 0.00178413, throughput 9.32843K wps
[Epoch 96 Batch 120/162] avg loss 0.00158356, throughput 9.32044K wps
[Epoch 96 Batch 150/162] avg loss 0.00175532, throughput 9.49033K wps
Begin Testing...
[Epoch 96] train avg loss 0.00170288, dev acc 0.9089, dev avg loss 0.221998, throughput 9.40377K wps
[Epoch 97 Batch 30/162] avg loss 0.00178649, throughput 9.53962K wps
[Epoch 97 Batch 60/162] avg loss 0.00160091, throughput 9.5448K wps
[Epoch 97 Batch 90/162] avg loss 0.00175841, throughput 9.45058K wps
[Epoch 97 Batch 120/162] avg loss 0.00183563, throughput 9.37682K wps
[Epoch 97 Batch 150/162] avg loss 0.00168899, throughput 9.49673K wps
Begin Testing...
[Epoch 97] train avg loss 0.0017309, dev acc 0.9100, dev avg loss 0.221913, throughput 9.48338K wps
[Epoch 98 Batch 30/162] avg loss 0.00166143, throughput 9.65535K wps
[Epoch 98 Batch 60/162] avg loss 0.00174394, throughput 9.4072K wps
[Epoch 98 Batch 90/162] avg loss 0.00148918, throughput 9.50969K wps
[Epoch 98 Batch 120/162] avg loss 0.00164739, throughput 9.54504K wps
[Epoch 98 Batch 150/162] avg loss 0.00166885, throughput 9.29841K wps
Begin Testing...
[Epoch 98] train avg loss 0.00167471, dev acc 0.9078, dev avg loss 0.221765, throughput 9.47868K wps
[Epoch 99 Batch 30/162] avg loss 0.00179785, throughput 9.68265K wps
[Epoch 99 Batch 60/162] avg loss 0.0016508, throughput 9.44073K wps
[Epoch 99 Batch 90/162] avg loss 0.00163731, throughput 9.46019K wps
[Epoch 99 Batch 120/162] avg loss 0.00154871, throughput 9.35539K wps
[Epoch 99 Batch 150/162] avg loss 0.00160724, throughput 9.34212K wps
Begin Testing...
[Epoch 99] train avg loss 0.00166285, dev acc 0.9133, dev avg loss 0.224643, throughput 9.46026K wps
Observed Improvement.
Begin Testing...
[Epoch 100 Batch 30/162] avg loss 0.0016867, throughput 9.38992K wps
[Epoch 100 Batch 60/162] avg loss 0.00144494, throughput 9.18237K wps
[Epoch 100 Batch 90/162] avg loss 0.0015796, throughput 9.33024K wps
[Epoch 100 Batch 120/162] avg loss 0.00175362, throughput 9.20946K wps
[Epoch 100 Batch 150/162] avg loss 0.00163614, throughput 9.31649K wps
Begin Testing...
[Epoch 100] train avg loss 0.00163861, dev acc 0.9111, dev avg loss 0.222973, throughput 9.28446K wps
[Epoch 101 Batch 30/162] avg loss 0.00155592, throughput 9.5694K wps
[Epoch 101 Batch 60/162] avg loss 0.00154838, throughput 9.52574K wps
[Epoch 101 Batch 90/162] avg loss 0.00163983, throughput 9.27901K wps
[Epoch 101 Batch 120/162] avg loss 0.00149855, throughput 9.58096K wps
[Epoch 101 Batch 150/162] avg loss 0.00161557, throughput 9.49127K wps
Begin Testing...
[Epoch 101] train avg loss 0.00159457, dev acc 0.9111, dev avg loss 0.222961, throughput 9.48217K wps
[Epoch 102 Batch 30/162] avg loss 0.00166855, throughput 9.48211K wps
[Epoch 102 Batch 60/162] avg loss 0.00158453, throughput 9.35853K wps
[Epoch 102 Batch 90/162] avg loss 0.00154465, throughput 9.26111K wps
[Epoch 102 Batch 120/162] avg loss 0.00164105, throughput 9.32157K wps
[Epoch 102 Batch 150/162] avg loss 0.00153149, throughput 9.39485K wps
Begin Testing...
[Epoch 102] train avg loss 0.00159511, dev acc 0.9100, dev avg loss 0.222255, throughput 9.3473K wps
[Epoch 103 Batch 30/162] avg loss 0.00160685, throughput 9.55635K wps
[Epoch 103 Batch 60/162] avg loss 0.00144046, throughput 9.32677K wps
[Epoch 103 Batch 90/162] avg loss 0.00141298, throughput 9.16574K wps
[Epoch 103 Batch 120/162] avg loss 0.00148582, throughput 9.59466K wps
[Epoch 103 Batch 150/162] avg loss 0.00160862, throughput 9.29459K wps
Begin Testing...
[Epoch 103] train avg loss 0.001506, dev acc 0.9111, dev avg loss 0.227087, throughput 9.40768K wps
[Epoch 104 Batch 30/162] avg loss 0.00157667, throughput 9.67024K wps
[Epoch 104 Batch 60/162] avg loss 0.00155269, throughput 9.25444K wps
[Epoch 104 Batch 90/162] avg loss 0.00144932, throughput 9.32201K wps
[Epoch 104 Batch 120/162] avg loss 0.00155954, throughput 9.5227K wps
[Epoch 104 Batch 150/162] avg loss 0.00164006, throughput 9.41261K wps
Begin Testing...
[Epoch 104] train avg loss 0.00157882, dev acc 0.9089, dev avg loss 0.222553, throughput 9.43147K wps
[Epoch 105 Batch 30/162] avg loss 0.00133473, throughput 9.59767K wps
[Epoch 105 Batch 60/162] avg loss 0.00147115, throughput 9.51409K wps
[Epoch 105 Batch 90/162] avg loss 0.00153091, throughput 9.61359K wps
[Epoch 105 Batch 120/162] avg loss 0.00162501, throughput 9.40269K wps
[Epoch 105 Batch 150/162] avg loss 0.00159604, throughput 9.42578K wps
Begin Testing...
[Epoch 105] train avg loss 0.00151743, dev acc 0.9089, dev avg loss 0.221775, throughput 9.49083K wps
[Epoch 106 Batch 30/162] avg loss 0.00155984, throughput 9.77625K wps
[Epoch 106 Batch 60/162] avg loss 0.00155868, throughput 9.27349K wps
[Epoch 106 Batch 90/162] avg loss 0.00155244, throughput 9.34157K wps
[Epoch 106 Batch 120/162] avg loss 0.00155748, throughput 9.3374K wps
[Epoch 106 Batch 150/162] avg loss 0.00149035, throughput 9.41382K wps
Begin Testing...
[Epoch 106] train avg loss 0.00153803, dev acc 0.9089, dev avg loss 0.221765, throughput 9.42445K wps
[Epoch 107 Batch 30/162] avg loss 0.00151909, throughput 9.55543K wps
[Epoch 107 Batch 60/162] avg loss 0.00140774, throughput 9.40742K wps
[Epoch 107 Batch 90/162] avg loss 0.00150637, throughput 9.53998K wps
[Epoch 107 Batch 120/162] avg loss 0.00138928, throughput 9.26287K wps
[Epoch 107 Batch 150/162] avg loss 0.00151119, throughput 9.24345K wps
Begin Testing...
[Epoch 107] train avg loss 0.00146624, dev acc 0.9089, dev avg loss 0.221703, throughput 9.39656K wps
[Epoch 108 Batch 30/162] avg loss 0.00150272, throughput 9.80207K wps
[Epoch 108 Batch 60/162] avg loss 0.00138365, throughput 9.37652K wps
[Epoch 108 Batch 90/162] avg loss 0.00152939, throughput 9.60643K wps
[Epoch 108 Batch 120/162] avg loss 0.00158551, throughput 9.43018K wps
[Epoch 108 Batch 150/162] avg loss 0.00163434, throughput 9.32184K wps
Begin Testing...
[Epoch 108] train avg loss 0.00152277, dev acc 0.9100, dev avg loss 0.222098, throughput 9.49K wps
[Epoch 109 Batch 30/162] avg loss 0.00146405, throughput 9.84896K wps
[Epoch 109 Batch 60/162] avg loss 0.00143059, throughput 9.44212K wps
[Epoch 109 Batch 90/162] avg loss 0.00153194, throughput 9.50758K wps
[Epoch 109 Batch 120/162] avg loss 0.00150643, throughput 9.33685K wps
[Epoch 109 Batch 150/162] avg loss 0.00132447, throughput 9.52343K wps
Begin Testing...
[Epoch 109] train avg loss 0.00146392, dev acc 0.9100, dev avg loss 0.221666, throughput 9.50407K wps
[Epoch 110 Batch 30/162] avg loss 0.00135284, throughput 9.55658K wps
[Epoch 110 Batch 60/162] avg loss 0.00137775, throughput 9.44643K wps
[Epoch 110 Batch 90/162] avg loss 0.00149369, throughput 9.39595K wps
[Epoch 110 Batch 120/162] avg loss 0.00149911, throughput 9.36309K wps
[Epoch 110 Batch 150/162] avg loss 0.00146335, throughput 9.35473K wps
Begin Testing...
[Epoch 110] train avg loss 0.00143569, dev acc 0.9100, dev avg loss 0.222814, throughput 9.43283K wps
[Epoch 111 Batch 30/162] avg loss 0.00138026, throughput 9.64692K wps
[Epoch 111 Batch 60/162] avg loss 0.00139064, throughput 9.43645K wps
[Epoch 111 Batch 90/162] avg loss 0.00149803, throughput 9.56594K wps
[Epoch 111 Batch 120/162] avg loss 0.00145327, throughput 9.34719K wps
[Epoch 111 Batch 150/162] avg loss 0.00141895, throughput 9.52009K wps
Begin Testing...
[Epoch 111] train avg loss 0.00143278, dev acc 0.9122, dev avg loss 0.221731, throughput 9.49545K wps
[Epoch 112 Batch 30/162] avg loss 0.00142091, throughput 9.74521K wps
[Epoch 112 Batch 60/162] avg loss 0.00134002, throughput 9.39291K wps
[Epoch 112 Batch 90/162] avg loss 0.00125917, throughput 9.10316K wps
[Epoch 112 Batch 120/162] avg loss 0.0014779, throughput 9.42114K wps
[Epoch 112 Batch 150/162] avg loss 0.00142531, throughput 9.33229K wps
Begin Testing...
[Epoch 112] train avg loss 0.00141332, dev acc 0.9100, dev avg loss 0.221703, throughput 9.41163K wps
[Epoch 113 Batch 30/162] avg loss 0.00135784, throughput 9.66712K wps
[Epoch 113 Batch 60/162] avg loss 0.00111503, throughput 9.53374K wps
[Epoch 113 Batch 90/162] avg loss 0.00128004, throughput 9.52063K wps
[Epoch 113 Batch 120/162] avg loss 0.00156751, throughput 9.34236K wps
[Epoch 113 Batch 150/162] avg loss 0.0012855, throughput 9.39005K wps
Begin Testing...
[Epoch 113] train avg loss 0.00133562, dev acc 0.9100, dev avg loss 0.221894, throughput 9.46437K wps
[Epoch 114 Batch 30/162] avg loss 0.00140393, throughput 9.69796K wps
[Epoch 114 Batch 60/162] avg loss 0.00136709, throughput 9.5882K wps
[Epoch 114 Batch 90/162] avg loss 0.00143445, throughput 9.24498K wps
[Epoch 114 Batch 120/162] avg loss 0.00123454, throughput 9.59949K wps
[Epoch 114 Batch 150/162] avg loss 0.00141351, throughput 9.50042K wps
Begin Testing...
[Epoch 114] train avg loss 0.00137146, dev acc 0.9089, dev avg loss 0.22262, throughput 9.53255K wps
[Epoch 115 Batch 30/162] avg loss 0.00125641, throughput 9.72745K wps
[Epoch 115 Batch 60/162] avg loss 0.00139346, throughput 9.27458K wps
[Epoch 115 Batch 90/162] avg loss 0.00143287, throughput 9.50376K wps
[Epoch 115 Batch 120/162] avg loss 0.00129611, throughput 9.40424K wps
[Epoch 115 Batch 150/162] avg loss 0.00143006, throughput 9.2337K wps
Begin Testing...
[Epoch 115] train avg loss 0.00135838, dev acc 0.9122, dev avg loss 0.221987, throughput 9.43329K wps
[Epoch 116 Batch 30/162] avg loss 0.00132687, throughput 9.55947K wps
[Epoch 116 Batch 60/162] avg loss 0.00141775, throughput 9.51491K wps
[Epoch 116 Batch 90/162] avg loss 0.00144991, throughput 9.50661K wps
[Epoch 116 Batch 120/162] avg loss 0.00127864, throughput 9.46044K wps
[Epoch 116 Batch 150/162] avg loss 0.00117843, throughput 9.46565K wps
Begin Testing...
[Epoch 116] train avg loss 0.00131752, dev acc 0.9144, dev avg loss 0.22564, throughput 9.49084K wps
Observed Improvement.
Begin Testing...
[Epoch 117 Batch 30/162] avg loss 0.00122943, throughput 9.62231K wps
[Epoch 117 Batch 60/162] avg loss 0.00121805, throughput 9.37403K wps
[Epoch 117 Batch 90/162] avg loss 0.00130586, throughput 9.51335K wps
[Epoch 117 Batch 120/162] avg loss 0.00128913, throughput 9.41015K wps
[Epoch 117 Batch 150/162] avg loss 0.00156928, throughput 9.53793K wps
Begin Testing...
[Epoch 117] train avg loss 0.00131645, dev acc 0.9100, dev avg loss 0.222418, throughput 9.4934K wps
[Epoch 118 Batch 30/162] avg loss 0.00128903, throughput 9.68165K wps
[Epoch 118 Batch 60/162] avg loss 0.00141475, throughput 9.48317K wps
[Epoch 118 Batch 90/162] avg loss 0.00135876, throughput 9.44147K wps
[Epoch 118 Batch 120/162] avg loss 0.00146399, throughput 9.45251K wps
[Epoch 118 Batch 150/162] avg loss 0.00136926, throughput 9.45762K wps
Begin Testing...
[Epoch 118] train avg loss 0.0013692, dev acc 0.9100, dev avg loss 0.222857, throughput 9.49209K wps
[Epoch 119 Batch 30/162] avg loss 0.00136451, throughput 9.57754K wps
[Epoch 119 Batch 60/162] avg loss 0.00135196, throughput 9.54305K wps
[Epoch 119 Batch 90/162] avg loss 0.00134072, throughput 9.24381K wps
[Epoch 119 Batch 120/162] avg loss 0.00114468, throughput 9.43458K wps
[Epoch 119 Batch 150/162] avg loss 0.00134924, throughput 9.50559K wps
Begin Testing...
[Epoch 119] train avg loss 0.00130256, dev acc 0.9089, dev avg loss 0.22362, throughput 9.46818K wps
[Epoch 120 Batch 30/162] avg loss 0.00127147, throughput 9.47676K wps
[Epoch 120 Batch 60/162] avg loss 0.00114476, throughput 9.55757K wps
[Epoch 120 Batch 90/162] avg loss 0.00125503, throughput 9.61913K wps
[Epoch 120 Batch 120/162] avg loss 0.00119459, throughput 9.4399K wps
[Epoch 120 Batch 150/162] avg loss 0.00140475, throughput 9.40849K wps
Begin Testing...
[Epoch 120] train avg loss 0.00125699, dev acc 0.9133, dev avg loss 0.222498, throughput 9.50432K wps
[Epoch 121 Batch 30/162] avg loss 0.00114119, throughput 9.69159K wps
[Epoch 121 Batch 60/162] avg loss 0.00138387, throughput 9.53339K wps
[Epoch 121 Batch 90/162] avg loss 0.00126564, throughput 9.59052K wps
[Epoch 121 Batch 120/162] avg loss 0.00121337, throughput 9.44418K wps
[Epoch 121 Batch 150/162] avg loss 0.00115387, throughput 9.45679K wps
Begin Testing...
[Epoch 121] train avg loss 0.00123913, dev acc 0.9100, dev avg loss 0.222634, throughput 9.53118K wps
[Epoch 122 Batch 30/162] avg loss 0.00122985, throughput 9.72445K wps
[Epoch 122 Batch 60/162] avg loss 0.00123836, throughput 9.25566K wps
[Epoch 122 Batch 90/162] avg loss 0.0012213, throughput 9.4212K wps
[Epoch 122 Batch 120/162] avg loss 0.00124092, throughput 9.35129K wps
[Epoch 122 Batch 150/162] avg loss 0.00130991, throughput 9.23134K wps
Begin Testing...
[Epoch 122] train avg loss 0.00124978, dev acc 0.9111, dev avg loss 0.224055, throughput 9.39897K wps
[Epoch 123 Batch 30/162] avg loss 0.00116557, throughput 9.3764K wps
[Epoch 123 Batch 60/162] avg loss 0.00118667, throughput 9.30423K wps
[Epoch 123 Batch 90/162] avg loss 0.00125872, throughput 9.57196K wps
[Epoch 123 Batch 120/162] avg loss 0.00126369, throughput 9.53586K wps
[Epoch 123 Batch 150/162] avg loss 0.00131792, throughput 9.58002K wps
Begin Testing...
[Epoch 123] train avg loss 0.00122882, dev acc 0.9089, dev avg loss 0.223071, throughput 9.45453K wps
[Epoch 124 Batch 30/162] avg loss 0.0012521, throughput 9.69596K wps
[Epoch 124 Batch 60/162] avg loss 0.00131599, throughput 9.2816K wps
[Epoch 124 Batch 90/162] avg loss 0.0011898, throughput 9.43416K wps
[Epoch 124 Batch 120/162] avg loss 0.00123039, throughput 9.59377K wps
[Epoch 124 Batch 150/162] avg loss 0.00115976, throughput 9.44675K wps
Begin Testing...
[Epoch 124] train avg loss 0.00122495, dev acc 0.9133, dev avg loss 0.222843, throughput 9.47436K wps
[Epoch 125 Batch 30/162] avg loss 0.00105982, throughput 9.52406K wps
[Epoch 125 Batch 60/162] avg loss 0.00111043, throughput 9.41026K wps
[Epoch 125 Batch 90/162] avg loss 0.00123868, throughput 9.40252K wps
[Epoch 125 Batch 120/162] avg loss 0.00130943, throughput 9.45509K wps
[Epoch 125 Batch 150/162] avg loss 0.00108723, throughput 9.35049K wps
Begin Testing...
[Epoch 125] train avg loss 0.00117095, dev acc 0.9111, dev avg loss 0.222912, throughput 9.429K wps
[Epoch 126 Batch 30/162] avg loss 0.00114934, throughput 9.69774K wps
[Epoch 126 Batch 60/162] avg loss 0.00127205, throughput 9.42149K wps
[Epoch 126 Batch 90/162] avg loss 0.00109735, throughput 9.2696K wps
[Epoch 126 Batch 120/162] avg loss 0.00118558, throughput 9.47362K wps
[Epoch 126 Batch 150/162] avg loss 0.00118299, throughput 9.3863K wps
Begin Testing...
[Epoch 126] train avg loss 0.00117603, dev acc 0.9111, dev avg loss 0.223012, throughput 9.43158K wps
[Epoch 127 Batch 30/162] avg loss 0.00106438, throughput 9.61103K wps
[Epoch 127 Batch 60/162] avg loss 0.00126581, throughput 9.46881K wps
[Epoch 127 Batch 90/162] avg loss 0.00123103, throughput 9.44252K wps
[Epoch 127 Batch 120/162] avg loss 0.00115324, throughput 9.53635K wps
[Epoch 127 Batch 150/162] avg loss 0.0011541, throughput 9.45099K wps
Begin Testing...
[Epoch 127] train avg loss 0.00116692, dev acc 0.9100, dev avg loss 0.223098, throughput 9.48731K wps
[Epoch 128 Batch 30/162] avg loss 0.00125239, throughput 9.77453K wps
[Epoch 128 Batch 60/162] avg loss 0.00114207, throughput 9.33113K wps
[Epoch 128 Batch 90/162] avg loss 0.00127312, throughput 9.21948K wps
[Epoch 128 Batch 120/162] avg loss 0.00107674, throughput 9.4555K wps
[Epoch 128 Batch 150/162] avg loss 0.00122742, throughput 9.34328K wps
Begin Testing...
[Epoch 128] train avg loss 0.00118577, dev acc 0.9122, dev avg loss 0.222512, throughput 9.41094K wps
[Epoch 129 Batch 30/162] avg loss 0.00108751, throughput 9.4553K wps
[Epoch 129 Batch 60/162] avg loss 0.00109865, throughput 9.35297K wps
[Epoch 129 Batch 90/162] avg loss 0.00107189, throughput 9.43756K wps
[Epoch 129 Batch 120/162] avg loss 0.00115176, throughput 9.46549K wps
[Epoch 129 Batch 150/162] avg loss 0.00111142, throughput 9.57869K wps
Begin Testing...
[Epoch 129] train avg loss 0.00111762, dev acc 0.9111, dev avg loss 0.223377, throughput 9.45679K wps
[Epoch 130 Batch 30/162] avg loss 0.00103984, throughput 9.61557K wps
[Epoch 130 Batch 60/162] avg loss 0.00112941, throughput 9.53655K wps
[Epoch 130 Batch 90/162] avg loss 0.00112167, throughput 9.35607K wps
[Epoch 130 Batch 120/162] avg loss 0.00112502, throughput 9.4909K wps
[Epoch 130 Batch 150/162] avg loss 0.000984955, throughput 9.36631K wps
Begin Testing...
[Epoch 130] train avg loss 0.0010796, dev acc 0.9100, dev avg loss 0.2234, throughput 9.46129K wps
[Epoch 131 Batch 30/162] avg loss 0.00113738, throughput 9.80491K wps
[Epoch 131 Batch 60/162] avg loss 0.00116377, throughput 9.43327K wps
[Epoch 131 Batch 90/162] avg loss 0.00118268, throughput 9.39361K wps
[Epoch 131 Batch 120/162] avg loss 0.000993232, throughput 9.28708K wps
[Epoch 131 Batch 150/162] avg loss 0.001163, throughput 9.50423K wps
Begin Testing...
[Epoch 131] train avg loss 0.00114074, dev acc 0.9100, dev avg loss 0.224313, throughput 9.47435K wps
[Epoch 132 Batch 30/162] avg loss 0.00111567, throughput 9.67337K wps
[Epoch 132 Batch 60/162] avg loss 0.00105687, throughput 9.50245K wps
[Epoch 132 Batch 90/162] avg loss 0.00100776, throughput 9.42686K wps
[Epoch 132 Batch 120/162] avg loss 0.00108074, throughput 9.2436K wps
[Epoch 132 Batch 150/162] avg loss 0.00114313, throughput 9.47506K wps
Begin Testing...
[Epoch 132] train avg loss 0.00107568, dev acc 0.9133, dev avg loss 0.223157, throughput 9.46699K wps
[Epoch 133 Batch 30/162] avg loss 0.00111032, throughput 9.54253K wps
[Epoch 133 Batch 60/162] avg loss 0.00114713, throughput 9.34676K wps
[Epoch 133 Batch 90/162] avg loss 0.00102403, throughput 9.41209K wps
[Epoch 133 Batch 120/162] avg loss 0.00105928, throughput 9.45071K wps
[Epoch 133 Batch 150/162] avg loss 0.00101195, throughput 9.29149K wps
Begin Testing...
[Epoch 133] train avg loss 0.00106775, dev acc 0.9111, dev avg loss 0.225729, throughput 9.40193K wps
[Epoch 134 Batch 30/162] avg loss 0.00113335, throughput 9.58842K wps
[Epoch 134 Batch 60/162] avg loss 0.00110991, throughput 9.54862K wps
[Epoch 134 Batch 90/162] avg loss 0.00105711, throughput 9.35097K wps
[Epoch 134 Batch 120/162] avg loss 0.000976467, throughput 9.25573K wps
[Epoch 134 Batch 150/162] avg loss 0.00110767, throughput 9.31084K wps
Begin Testing...
[Epoch 134] train avg loss 0.0011009, dev acc 0.9156, dev avg loss 0.225927, throughput 9.40236K wps
Observed Improvement.
Begin Testing...
[Epoch 135 Batch 30/162] avg loss 0.00116725, throughput 9.61091K wps
[Epoch 135 Batch 60/162] avg loss 0.00109201, throughput 9.32495K wps
[Epoch 135 Batch 90/162] avg loss 0.00104943, throughput 9.37414K wps
[Epoch 135 Batch 120/162] avg loss 0.000947498, throughput 9.48612K wps
[Epoch 135 Batch 150/162] avg loss 0.000960457, throughput 9.53163K wps
Begin Testing...
[Epoch 135] train avg loss 0.001038, dev acc 0.9111, dev avg loss 0.223876, throughput 9.46677K wps
[Epoch 136 Batch 30/162] avg loss 0.00106191, throughput 9.67335K wps
[Epoch 136 Batch 60/162] avg loss 0.00111192, throughput 9.22045K wps
[Epoch 136 Batch 90/162] avg loss 0.00115435, throughput 9.56161K wps
[Epoch 136 Batch 120/162] avg loss 0.00115463, throughput 9.37278K wps
[Epoch 136 Batch 150/162] avg loss 0.00111101, throughput 9.54358K wps
Begin Testing...
[Epoch 136] train avg loss 0.00110367, dev acc 0.9111, dev avg loss 0.222991, throughput 9.47841K wps
[Epoch 137 Batch 30/162] avg loss 0.00106102, throughput 9.62156K wps
[Epoch 137 Batch 60/162] avg loss 0.00104989, throughput 9.25409K wps
[Epoch 137 Batch 90/162] avg loss 0.00105861, throughput 9.20055K wps
[Epoch 137 Batch 120/162] avg loss 0.000905393, throughput 9.64465K wps
[Epoch 137 Batch 150/162] avg loss 0.00112424, throughput 9.36454K wps
Begin Testing...
[Epoch 137] train avg loss 0.00105508, dev acc 0.9133, dev avg loss 0.223415, throughput 9.40346K wps
[Epoch 138 Batch 30/162] avg loss 0.0010362, throughput 9.7074K wps
[Epoch 138 Batch 60/162] avg loss 0.000974917, throughput 9.38257K wps
[Epoch 138 Batch 90/162] avg loss 0.00101066, throughput 9.47824K wps
[Epoch 138 Batch 120/162] avg loss 0.00106326, throughput 9.34612K wps
[Epoch 138 Batch 150/162] avg loss 0.00102337, throughput 9.63837K wps
Begin Testing...
[Epoch 138] train avg loss 0.001022, dev acc 0.9111, dev avg loss 0.223484, throughput 9.49223K wps
[Epoch 139 Batch 30/162] avg loss 0.000945979, throughput 9.52821K wps
[Epoch 139 Batch 60/162] avg loss 0.00102061, throughput 9.33548K wps
[Epoch 139 Batch 90/162] avg loss 0.00104412, throughput 9.53984K wps
[Epoch 139 Batch 120/162] avg loss 0.00112662, throughput 9.45055K wps
[Epoch 139 Batch 150/162] avg loss 0.000977858, throughput 9.35481K wps
Begin Testing...
[Epoch 139] train avg loss 0.00103237, dev acc 0.9133, dev avg loss 0.223587, throughput 9.43519K wps
[Epoch 140 Batch 30/162] avg loss 0.000959292, throughput 9.58615K wps
[Epoch 140 Batch 60/162] avg loss 0.000953605, throughput 9.56158K wps
[Epoch 140 Batch 90/162] avg loss 0.000964439, throughput 9.405K wps
[Epoch 140 Batch 120/162] avg loss 0.00102098, throughput 9.30545K wps
[Epoch 140 Batch 150/162] avg loss 0.00115789, throughput 9.25038K wps
Begin Testing...
[Epoch 140] train avg loss 0.00101178, dev acc 0.9167, dev avg loss 0.226709, throughput 9.40956K wps
Observed Improvement.
Begin Testing...
[Epoch 141 Batch 30/162] avg loss 0.00113057, throughput 9.77214K wps
[Epoch 141 Batch 60/162] avg loss 0.000975399, throughput 9.39649K wps
[Epoch 141 Batch 90/162] avg loss 0.000910517, throughput 9.55068K wps
[Epoch 141 Batch 120/162] avg loss 0.000878768, throughput 9.3249K wps
[Epoch 141 Batch 150/162] avg loss 0.0010695, throughput 9.36212K wps
Begin Testing...
[Epoch 141] train avg loss 0.00100098, dev acc 0.9133, dev avg loss 0.224048, throughput 9.46573K wps
[Epoch 142 Batch 30/162] avg loss 0.00102188, throughput 9.46622K wps
[Epoch 142 Batch 60/162] avg loss 0.00100031, throughput 9.56111K wps
[Epoch 142 Batch 90/162] avg loss 0.000943464, throughput 9.26216K wps
[Epoch 142 Batch 120/162] avg loss 0.000942223, throughput 9.28736K wps
[Epoch 142 Batch 150/162] avg loss 0.000883694, throughput 9.48009K wps
Begin Testing...
[Epoch 142] train avg loss 0.00097274, dev acc 0.9111, dev avg loss 0.225363, throughput 9.40749K wps
[Epoch 143 Batch 30/162] avg loss 0.001082, throughput 9.73294K wps
[Epoch 143 Batch 60/162] avg loss 0.000956779, throughput 9.40779K wps
[Epoch 143 Batch 90/162] avg loss 0.000976916, throughput 9.5224K wps
[Epoch 143 Batch 120/162] avg loss 0.000912033, throughput 9.55396K wps
[Epoch 143 Batch 150/162] avg loss 0.00100607, throughput 9.3939K wps
Begin Testing...
[Epoch 143] train avg loss 0.00100512, dev acc 0.9122, dev avg loss 0.224029, throughput 9.49355K wps
[Epoch 144 Batch 30/162] avg loss 0.000975586, throughput 9.67125K wps
[Epoch 144 Batch 60/162] avg loss 0.000973175, throughput 9.42653K wps
[Epoch 144 Batch 90/162] avg loss 0.000978693, throughput 9.72481K wps
[Epoch 144 Batch 120/162] avg loss 0.00106694, throughput 9.28054K wps
[Epoch 144 Batch 150/162] avg loss 0.000963126, throughput 9.32752K wps
Begin Testing...
[Epoch 144] train avg loss 0.000978409, dev acc 0.9122, dev avg loss 0.223663, throughput 9.4918K wps
[Epoch 145 Batch 30/162] avg loss 0.000981131, throughput 9.64032K wps
[Epoch 145 Batch 60/162] avg loss 0.000973371, throughput 9.44804K wps
[Epoch 145 Batch 90/162] avg loss 0.00109274, throughput 9.57748K wps
[Epoch 145 Batch 120/162] avg loss 0.000957799, throughput 9.53708K wps
[Epoch 145 Batch 150/162] avg loss 0.00101956, throughput 9.29005K wps
Begin Testing...
[Epoch 145] train avg loss 0.000995018, dev acc 0.9111, dev avg loss 0.224103, throughput 9.50179K wps
[Epoch 146 Batch 30/162] avg loss 0.000853733, throughput 9.83061K wps
[Epoch 146 Batch 60/162] avg loss 0.000869478, throughput 9.34752K wps
[Epoch 146 Batch 90/162] avg loss 0.000881037, throughput 9.40477K wps
[Epoch 146 Batch 120/162] avg loss 0.000869568, throughput 9.34224K wps
[Epoch 146 Batch 150/162] avg loss 0.00104685, throughput 9.34322K wps
Begin Testing...
[Epoch 146] train avg loss 0.000903813, dev acc 0.9133, dev avg loss 0.224297, throughput 9.44272K wps
[Epoch 147 Batch 30/162] avg loss 0.000933515, throughput 9.46923K wps
[Epoch 147 Batch 60/162] avg loss 0.000844412, throughput 9.36407K wps
[Epoch 147 Batch 90/162] avg loss 0.000997759, throughput 9.52633K wps
[Epoch 147 Batch 120/162] avg loss 0.000813217, throughput 9.39503K wps
[Epoch 147 Batch 150/162] avg loss 0.000855254, throughput 9.42886K wps
Begin Testing...
[Epoch 147] train avg loss 0.000912873, dev acc 0.9122, dev avg loss 0.224658, throughput 9.4154K wps
[Epoch 148 Batch 30/162] avg loss 0.000867021, throughput 9.59317K wps
[Epoch 148 Batch 60/162] avg loss 0.00102114, throughput 9.47035K wps
[Epoch 148 Batch 90/162] avg loss 0.000803051, throughput 9.39425K wps
[Epoch 148 Batch 120/162] avg loss 0.000933587, throughput 9.39206K wps
[Epoch 148 Batch 150/162] avg loss 0.00098846, throughput 9.32998K wps
Begin Testing...
[Epoch 148] train avg loss 0.000919607, dev acc 0.9133, dev avg loss 0.225566, throughput 9.43026K wps
[Epoch 149 Batch 30/162] avg loss 0.000884526, throughput 9.54649K wps
[Epoch 149 Batch 60/162] avg loss 0.000869229, throughput 9.40305K wps
[Epoch 149 Batch 90/162] avg loss 0.000851869, throughput 9.39807K wps
[Epoch 149 Batch 120/162] avg loss 0.000958465, throughput 9.3761K wps
[Epoch 149 Batch 150/162] avg loss 0.000991292, throughput 9.36985K wps
Begin Testing...
[Epoch 149] train avg loss 0.000913499, dev acc 0.9122, dev avg loss 0.226104, throughput 9.43666K wps
[Epoch 150 Batch 30/162] avg loss 0.00103536, throughput 9.50726K wps
[Epoch 150 Batch 60/162] avg loss 0.000863213, throughput 9.43903K wps
[Epoch 150 Batch 90/162] avg loss 0.000833596, throughput 9.38003K wps
[Epoch 150 Batch 120/162] avg loss 0.000913317, throughput 9.4497K wps
[Epoch 150 Batch 150/162] avg loss 0.000869809, throughput 9.42053K wps
Begin Testing...
[Epoch 150] train avg loss 0.000907328, dev acc 0.9122, dev avg loss 0.224837, throughput 9.42357K wps
[Epoch 151 Batch 30/162] avg loss 0.000929423, throughput 9.62597K wps
[Epoch 151 Batch 60/162] avg loss 0.000879522, throughput 9.43371K wps
[Epoch 151 Batch 90/162] avg loss 0.000882323, throughput 9.3822K wps
[Epoch 151 Batch 120/162] avg loss 0.000859007, throughput 9.3781K wps
[Epoch 151 Batch 150/162] avg loss 0.000976699, throughput 9.55755K wps
Begin Testing...
[Epoch 151] train avg loss 0.000912734, dev acc 0.9111, dev avg loss 0.224626, throughput 9.48003K wps
[Epoch 152 Batch 30/162] avg loss 0.000891258, throughput 9.81495K wps
[Epoch 152 Batch 60/162] avg loss 0.00099172, throughput 9.32078K wps
[Epoch 152 Batch 90/162] avg loss 0.00097521, throughput 9.50553K wps
[Epoch 152 Batch 120/162] avg loss 0.000842278, throughput 9.51554K wps
[Epoch 152 Batch 150/162] avg loss 0.000913899, throughput 9.42868K wps
Begin Testing...
[Epoch 152] train avg loss 0.000909668, dev acc 0.9133, dev avg loss 0.224839, throughput 9.52634K wps
[Epoch 153 Batch 30/162] avg loss 0.000921225, throughput 9.69245K wps
[Epoch 153 Batch 60/162] avg loss 0.00074156, throughput 9.50883K wps
[Epoch 153 Batch 90/162] avg loss 0.000924305, throughput 9.5557K wps
[Epoch 153 Batch 120/162] avg loss 0.000973678, throughput 9.52476K wps
[Epoch 153 Batch 150/162] avg loss 0.000898617, throughput 9.57085K wps
Begin Testing...
[Epoch 153] train avg loss 0.000885351, dev acc 0.9111, dev avg loss 0.224943, throughput 9.57112K wps
[Epoch 154 Batch 30/162] avg loss 0.000847902, throughput 9.63856K wps
[Epoch 154 Batch 60/162] avg loss 0.000861289, throughput 9.34341K wps
[Epoch 154 Batch 90/162] avg loss 0.000905562, throughput 9.51787K wps
[Epoch 154 Batch 120/162] avg loss 0.000806761, throughput 9.19985K wps
[Epoch 154 Batch 150/162] avg loss 0.000889041, throughput 9.50481K wps
Begin Testing...
[Epoch 154] train avg loss 0.00085996, dev acc 0.9122, dev avg loss 0.225457, throughput 9.44672K wps
[Epoch 155 Batch 30/162] avg loss 0.000811031, throughput 9.64314K wps
[Epoch 155 Batch 60/162] avg loss 0.000832428, throughput 9.43204K wps
[Epoch 155 Batch 90/162] avg loss 0.000813322, throughput 9.55769K wps
[Epoch 155 Batch 120/162] avg loss 0.000903611, throughput 9.47007K wps
[Epoch 155 Batch 150/162] avg loss 0.000891055, throughput 9.43069K wps
Begin Testing...
[Epoch 155] train avg loss 0.000852901, dev acc 0.9133, dev avg loss 0.225183, throughput 9.51721K wps
[Epoch 156 Batch 30/162] avg loss 0.000835327, throughput 9.70701K wps
[Epoch 156 Batch 60/162] avg loss 0.000738526, throughput 9.24208K wps
[Epoch 156 Batch 90/162] avg loss 0.000847573, throughput 9.42269K wps
[Epoch 156 Batch 120/162] avg loss 0.000822148, throughput 9.36582K wps
[Epoch 156 Batch 150/162] avg loss 0.000799497, throughput 9.28875K wps
Begin Testing...
[Epoch 156] train avg loss 0.000815921, dev acc 0.9111, dev avg loss 0.225875, throughput 9.39624K wps
[Epoch 157 Batch 30/162] avg loss 0.000810471, throughput 9.5175K wps
[Epoch 157 Batch 60/162] avg loss 0.000884315, throughput 9.33332K wps
[Epoch 157 Batch 90/162] avg loss 0.000906803, throughput 9.52434K wps
[Epoch 157 Batch 120/162] avg loss 0.000836198, throughput 9.51614K wps
[Epoch 157 Batch 150/162] avg loss 0.000734986, throughput 9.24872K wps
Begin Testing...
[Epoch 157] train avg loss 0.000851557, dev acc 0.9144, dev avg loss 0.22564, throughput 9.42881K wps
[Epoch 158 Batch 30/162] avg loss 0.000784531, throughput 9.8125K wps
[Epoch 158 Batch 60/162] avg loss 0.000842972, throughput 9.39657K wps
[Epoch 158 Batch 90/162] avg loss 0.000911293, throughput 9.59778K wps
[Epoch 158 Batch 120/162] avg loss 0.000859603, throughput 9.4819K wps
[Epoch 158 Batch 150/162] avg loss 0.00089083, throughput 9.63608K wps
Begin Testing...
[Epoch 158] train avg loss 0.000852318, dev acc 0.9133, dev avg loss 0.227462, throughput 9.55591K wps
[Epoch 159 Batch 30/162] avg loss 0.000855801, throughput 9.49532K wps
[Epoch 159 Batch 60/162] avg loss 0.000771537, throughput 9.34354K wps
[Epoch 159 Batch 90/162] avg loss 0.00081932, throughput 9.48587K wps
[Epoch 159 Batch 120/162] avg loss 0.000827659, throughput 9.41467K wps
[Epoch 159 Batch 150/162] avg loss 0.000862717, throughput 9.46331K wps
Begin Testing...
[Epoch 159] train avg loss 0.000831797, dev acc 0.9144, dev avg loss 0.226697, throughput 9.45324K wps
[Epoch 160 Batch 30/162] avg loss 0.00083123, throughput 9.6357K wps
[Epoch 160 Batch 60/162] avg loss 0.000803236, throughput 9.34554K wps
[Epoch 160 Batch 90/162] avg loss 0.000787448, throughput 9.43785K wps
[Epoch 160 Batch 120/162] avg loss 0.00085988, throughput 9.6958K wps
[Epoch 160 Batch 150/162] avg loss 0.000768387, throughput 9.47545K wps
Begin Testing...
[Epoch 160] train avg loss 0.000803617, dev acc 0.9122, dev avg loss 0.226279, throughput 9.52291K wps
[Epoch 161 Batch 30/162] avg loss 0.000779754, throughput 9.70696K wps
[Epoch 161 Batch 60/162] avg loss 0.000745472, throughput 9.48743K wps
[Epoch 161 Batch 90/162] avg loss 0.000796255, throughput 9.33892K wps
[Epoch 161 Batch 120/162] avg loss 0.00085385, throughput 9.5399K wps
[Epoch 161 Batch 150/162] avg loss 0.000898534, throughput 9.63374K wps
Begin Testing...
[Epoch 161] train avg loss 0.000813704, dev acc 0.9133, dev avg loss 0.226121, throughput 9.52013K wps
[Epoch 162 Batch 30/162] avg loss 0.000831688, throughput 9.58897K wps
[Epoch 162 Batch 60/162] avg loss 0.000787387, throughput 9.31192K wps
[Epoch 162 Batch 90/162] avg loss 0.000841297, throughput 9.37468K wps
[Epoch 162 Batch 120/162] avg loss 0.000735517, throughput 9.26154K wps
[Epoch 162 Batch 150/162] avg loss 0.000808323, throughput 9.42642K wps
Begin Testing...
[Epoch 162] train avg loss 0.000797242, dev acc 0.9122, dev avg loss 0.226307, throughput 9.39016K wps
[Epoch 163 Batch 30/162] avg loss 0.000768602, throughput 9.47395K wps
[Epoch 163 Batch 60/162] avg loss 0.000769082, throughput 9.43322K wps
[Epoch 163 Batch 90/162] avg loss 0.000741773, throughput 9.32792K wps
[Epoch 163 Batch 120/162] avg loss 0.000755426, throughput 9.49844K wps
[Epoch 163 Batch 150/162] avg loss 0.000772143, throughput 9.38931K wps
Begin Testing...
[Epoch 163] train avg loss 0.000769328, dev acc 0.9122, dev avg loss 0.226606, throughput 9.42387K wps
[Epoch 164 Batch 30/162] avg loss 0.000750479, throughput 9.53318K wps
[Epoch 164 Batch 60/162] avg loss 0.000786015, throughput 9.34163K wps
[Epoch 164 Batch 90/162] avg loss 0.000782792, throughput 9.64076K wps
[Epoch 164 Batch 120/162] avg loss 0.000781725, throughput 9.52478K wps
[Epoch 164 Batch 150/162] avg loss 0.000810021, throughput 9.26919K wps
Begin Testing...
[Epoch 164] train avg loss 0.000784969, dev acc 0.9122, dev avg loss 0.227337, throughput 9.4452K wps
[Epoch 165 Batch 30/162] avg loss 0.000841442, throughput 9.50454K wps
[Epoch 165 Batch 60/162] avg loss 0.000783675, throughput 9.44311K wps
[Epoch 165 Batch 90/162] avg loss 0.000775634, throughput 9.28703K wps
[Epoch 165 Batch 120/162] avg loss 0.000723225, throughput 9.33948K wps
[Epoch 165 Batch 150/162] avg loss 0.000684097, throughput 9.46626K wps
Begin Testing...
[Epoch 165] train avg loss 0.000761945, dev acc 0.9156, dev avg loss 0.227343, throughput 9.41391K wps
[Epoch 166 Batch 30/162] avg loss 0.000774307, throughput 9.45839K wps
[Epoch 166 Batch 60/162] avg loss 0.000729415, throughput 9.5607K wps
[Epoch 166 Batch 90/162] avg loss 0.00075124, throughput 9.32942K wps
[Epoch 166 Batch 120/162] avg loss 0.000766863, throughput 9.33272K wps
[Epoch 166 Batch 150/162] avg loss 0.000738584, throughput 9.37509K wps
Begin Testing...
[Epoch 166] train avg loss 0.00076355, dev acc 0.9122, dev avg loss 0.227278, throughput 9.38712K wps
[Epoch 167 Batch 30/162] avg loss 0.000692793, throughput 9.78161K wps
[Epoch 167 Batch 60/162] avg loss 0.000739576, throughput 9.31059K wps
[Epoch 167 Batch 90/162] avg loss 0.000722249, throughput 9.37463K wps
[Epoch 167 Batch 120/162] avg loss 0.000676367, throughput 9.64143K wps
[Epoch 167 Batch 150/162] avg loss 0.00081189, throughput 9.38279K wps
Begin Testing...
[Epoch 167] train avg loss 0.000729666, dev acc 0.9122, dev avg loss 0.227464, throughput 9.49125K wps
[Epoch 168 Batch 30/162] avg loss 0.000740285, throughput 9.63092K wps
[Epoch 168 Batch 60/162] avg loss 0.000691745, throughput 9.41775K wps
[Epoch 168 Batch 90/162] avg loss 0.000667764, throughput 9.43484K wps
[Epoch 168 Batch 120/162] avg loss 0.000741871, throughput 9.46505K wps
[Epoch 168 Batch 150/162] avg loss 0.000763664, throughput 9.38897K wps
Begin Testing...
[Epoch 168] train avg loss 0.000725041, dev acc 0.9144, dev avg loss 0.22824, throughput 9.47697K wps
[Epoch 169 Batch 30/162] avg loss 0.000741372, throughput 9.60909K wps
[Epoch 169 Batch 60/162] avg loss 0.000770346, throughput 9.30968K wps
[Epoch 169 Batch 90/162] avg loss 0.000809856, throughput 9.63461K wps
[Epoch 169 Batch 120/162] avg loss 0.0007665, throughput 9.4541K wps
[Epoch 169 Batch 150/162] avg loss 0.000733412, throughput 9.33823K wps
Begin Testing...
[Epoch 169] train avg loss 0.000760422, dev acc 0.9122, dev avg loss 0.227926, throughput 9.47509K wps
[Epoch 170 Batch 30/162] avg loss 0.000742853, throughput 9.49436K wps
[Epoch 170 Batch 60/162] avg loss 0.000713764, throughput 9.40827K wps
[Epoch 170 Batch 90/162] avg loss 0.0008823, throughput 9.38216K wps
[Epoch 170 Batch 120/162] avg loss 0.00073712, throughput 9.45686K wps
[Epoch 170 Batch 150/162] avg loss 0.000789328, throughput 9.4434K wps
Begin Testing...
[Epoch 170] train avg loss 0.000768808, dev acc 0.9122, dev avg loss 0.227669, throughput 9.4331K wps
[Epoch 171 Batch 30/162] avg loss 0.000674199, throughput 9.65296K wps
[Epoch 171 Batch 60/162] avg loss 0.000708372, throughput 9.21592K wps
[Epoch 171 Batch 90/162] avg loss 0.000741139, throughput 9.31535K wps
[Epoch 171 Batch 120/162] avg loss 0.0006861, throughput 9.4698K wps
[Epoch 171 Batch 150/162] avg loss 0.000838966, throughput 9.47935K wps
Begin Testing...
[Epoch 171] train avg loss 0.000732141, dev acc 0.9156, dev avg loss 0.22881, throughput 9.44764K wps
[Epoch 172 Batch 30/162] avg loss 0.000761718, throughput 9.72478K wps
[Epoch 172 Batch 60/162] avg loss 0.000675539, throughput 9.42146K wps
[Epoch 172 Batch 90/162] avg loss 0.000802364, throughput 9.69679K wps
[Epoch 172 Batch 120/162] avg loss 0.000759266, throughput 9.33009K wps
[Epoch 172 Batch 150/162] avg loss 0.000622553, throughput 9.51158K wps
Begin Testing...
[Epoch 172] train avg loss 0.000726559, dev acc 0.9122, dev avg loss 0.227729, throughput 9.53151K wps
[Epoch 173 Batch 30/162] avg loss 0.00072192, throughput 9.72198K wps
[Epoch 173 Batch 60/162] avg loss 0.00077425, throughput 9.44969K wps
[Epoch 173 Batch 90/162] avg loss 0.000707586, throughput 9.35631K wps
[Epoch 173 Batch 120/162] avg loss 0.000730592, throughput 9.3166K wps
[Epoch 173 Batch 150/162] avg loss 0.000659608, throughput 9.4361K wps
Begin Testing...
[Epoch 173] train avg loss 0.000722037, dev acc 0.9144, dev avg loss 0.22899, throughput 9.44125K wps
[Epoch 174 Batch 30/162] avg loss 0.000712421, throughput 9.49832K wps
[Epoch 174 Batch 60/162] avg loss 0.000725642, throughput 9.31665K wps
[Epoch 174 Batch 90/162] avg loss 0.0006417, throughput 9.40774K wps
[Epoch 174 Batch 120/162] avg loss 0.000755112, throughput 9.55306K wps
[Epoch 174 Batch 150/162] avg loss 0.000637733, throughput 9.47605K wps
Begin Testing...
[Epoch 174] train avg loss 0.000706044, dev acc 0.9144, dev avg loss 0.228319, throughput 9.46272K wps
[Epoch 175 Batch 30/162] avg loss 0.000758954, throughput 9.51266K wps
[Epoch 175 Batch 60/162] avg loss 0.000729812, throughput 9.3695K wps
[Epoch 175 Batch 90/162] avg loss 0.000728059, throughput 9.46442K wps
[Epoch 175 Batch 120/162] avg loss 0.000641145, throughput 9.60393K wps
[Epoch 175 Batch 150/162] avg loss 0.000765971, throughput 9.45877K wps
Begin Testing...
[Epoch 175] train avg loss 0.000721681, dev acc 0.9122, dev avg loss 0.228487, throughput 9.48951K wps
[Epoch 176 Batch 30/162] avg loss 0.00066101, throughput 9.78894K wps
[Epoch 176 Batch 60/162] avg loss 0.000683194, throughput 9.39529K wps
[Epoch 176 Batch 90/162] avg loss 0.00066506, throughput 9.36412K wps
[Epoch 176 Batch 120/162] avg loss 0.000724563, throughput 9.46776K wps
[Epoch 176 Batch 150/162] avg loss 0.000667924, throughput 9.53676K wps
Begin Testing...
[Epoch 176] train avg loss 0.000686892, dev acc 0.9133, dev avg loss 0.22849, throughput 9.4951K wps
[Epoch 177 Batch 30/162] avg loss 0.00066356, throughput 9.49517K wps
[Epoch 177 Batch 60/162] avg loss 0.000683949, throughput 9.39788K wps
[Epoch 177 Batch 90/162] avg loss 0.00061206, throughput 9.49784K wps
[Epoch 177 Batch 120/162] avg loss 0.000694638, throughput 9.31057K wps
[Epoch 177 Batch 150/162] avg loss 0.000775592, throughput 9.43885K wps
Begin Testing...
[Epoch 177] train avg loss 0.000699537, dev acc 0.9133, dev avg loss 0.228493, throughput 9.43813K wps
[Epoch 178 Batch 30/162] avg loss 0.000687203, throughput 9.73482K wps
[Epoch 178 Batch 60/162] avg loss 0.000734681, throughput 9.42476K wps
[Epoch 178 Batch 90/162] avg loss 0.000669249, throughput 9.40832K wps
[Epoch 178 Batch 120/162] avg loss 0.000611365, throughput 9.29011K wps
[Epoch 178 Batch 150/162] avg loss 0.000647053, throughput 9.58579K wps
Begin Testing...
[Epoch 178] train avg loss 0.000677725, dev acc 0.9156, dev avg loss 0.229111, throughput 9.46132K wps
[Epoch 179 Batch 30/162] avg loss 0.000752541, throughput 9.6688K wps
[Epoch 179 Batch 60/162] avg loss 0.000707151, throughput 9.24902K wps
[Epoch 179 Batch 90/162] avg loss 0.000677577, throughput 9.41248K wps
[Epoch 179 Batch 120/162] avg loss 0.000617344, throughput 9.46066K wps
[Epoch 179 Batch 150/162] avg loss 0.000724181, throughput 9.42849K wps
Begin Testing...
[Epoch 179] train avg loss 0.000702508, dev acc 0.9133, dev avg loss 0.228933, throughput 9.43449K wps
[Epoch 180 Batch 30/162] avg loss 0.000741006, throughput 9.65782K wps
[Epoch 180 Batch 60/162] avg loss 0.000684719, throughput 9.49201K wps
[Epoch 180 Batch 90/162] avg loss 0.000707943, throughput 9.31844K wps
[Epoch 180 Batch 120/162] avg loss 0.000617268, throughput 9.51339K wps
[Epoch 180 Batch 150/162] avg loss 0.000756673, throughput 9.44664K wps
Begin Testing...
[Epoch 180] train avg loss 0.000692751, dev acc 0.9167, dev avg loss 0.230026, throughput 9.46155K wps
Observed Improvement.
Begin Testing...
[Epoch 181 Batch 30/162] avg loss 0.000732699, throughput 9.83884K wps
[Epoch 181 Batch 60/162] avg loss 0.000674191, throughput 9.58113K wps
[Epoch 181 Batch 90/162] avg loss 0.000691134, throughput 9.59453K wps
[Epoch 181 Batch 120/162] avg loss 0.000636268, throughput 9.39896K wps
[Epoch 181 Batch 150/162] avg loss 0.000679339, throughput 9.61283K wps
Begin Testing...
[Epoch 181] train avg loss 0.000671888, dev acc 0.9144, dev avg loss 0.229258, throughput 9.57805K wps
[Epoch 182 Batch 30/162] avg loss 0.000634608, throughput 9.80659K wps
[Epoch 182 Batch 60/162] avg loss 0.000622103, throughput 9.31613K wps
[Epoch 182 Batch 90/162] avg loss 0.000565127, throughput 9.37557K wps
[Epoch 182 Batch 120/162] avg loss 0.00068467, throughput 9.37234K wps
[Epoch 182 Batch 150/162] avg loss 0.000694462, throughput 9.25642K wps
Begin Testing...
[Epoch 182] train avg loss 0.000638364, dev acc 0.9133, dev avg loss 0.229372, throughput 9.41447K wps
[Epoch 183 Batch 30/162] avg loss 0.000600163, throughput 9.63877K wps
[Epoch 183 Batch 60/162] avg loss 0.000653543, throughput 9.37157K wps
[Epoch 183 Batch 90/162] avg loss 0.000765288, throughput 9.47476K wps
[Epoch 183 Batch 120/162] avg loss 0.000686804, throughput 9.34099K wps
[Epoch 183 Batch 150/162] avg loss 0.000602954, throughput 9.5167K wps
Begin Testing...
[Epoch 183] train avg loss 0.000663596, dev acc 0.9133, dev avg loss 0.229554, throughput 9.47744K wps
[Epoch 184 Batch 30/162] avg loss 0.000632299, throughput 9.68356K wps
[Epoch 184 Batch 60/162] avg loss 0.000729371, throughput 9.53563K wps
[Epoch 184 Batch 90/162] avg loss 0.000609153, throughput 9.49795K wps
[Epoch 184 Batch 120/162] avg loss 0.000635616, throughput 9.55239K wps
[Epoch 184 Batch 150/162] avg loss 0.00066083, throughput 9.5798K wps
Begin Testing...
[Epoch 184] train avg loss 0.000652981, dev acc 0.9122, dev avg loss 0.229764, throughput 9.57005K wps
[Epoch 185 Batch 30/162] avg loss 0.000642287, throughput 9.633K wps
[Epoch 185 Batch 60/162] avg loss 0.000588259, throughput 9.34664K wps
[Epoch 185 Batch 90/162] avg loss 0.000639723, throughput 9.26593K wps
[Epoch 185 Batch 120/162] avg loss 0.000640973, throughput 9.4489K wps
[Epoch 185 Batch 150/162] avg loss 0.000641883, throughput 9.49362K wps
Begin Testing...
[Epoch 185] train avg loss 0.000635775, dev acc 0.9133, dev avg loss 0.229634, throughput 9.42209K wps
[Epoch 186 Batch 30/162] avg loss 0.000571754, throughput 9.67411K wps
[Epoch 186 Batch 60/162] avg loss 0.000704375, throughput 9.54823K wps
[Epoch 186 Batch 90/162] avg loss 0.000617343, throughput 9.62624K wps
[Epoch 186 Batch 120/162] avg loss 0.00066244, throughput 9.45851K wps
[Epoch 186 Batch 150/162] avg loss 0.000681854, throughput 9.5069K wps
Begin Testing...
[Epoch 186] train avg loss 0.000642316, dev acc 0.9133, dev avg loss 0.229183, throughput 9.54235K wps
[Epoch 187 Batch 30/162] avg loss 0.000621153, throughput 9.40676K wps
[Epoch 187 Batch 60/162] avg loss 0.000697022, throughput 9.45632K wps
[Epoch 187 Batch 90/162] avg loss 0.000606554, throughput 9.56706K wps
[Epoch 187 Batch 120/162] avg loss 0.00064045, throughput 9.56673K wps
[Epoch 187 Batch 150/162] avg loss 0.000632558, throughput 9.29649K wps
Begin Testing...
[Epoch 187] train avg loss 0.000652098, dev acc 0.9156, dev avg loss 0.230059, throughput 9.459K wps
[Epoch 188 Batch 30/162] avg loss 0.000626905, throughput 9.47745K wps
[Epoch 188 Batch 60/162] avg loss 0.000636277, throughput 9.47388K wps
[Epoch 188 Batch 90/162] avg loss 0.00054604, throughput 9.34969K wps
[Epoch 188 Batch 120/162] avg loss 0.000704845, throughput 9.46541K wps
[Epoch 188 Batch 150/162] avg loss 0.000636537, throughput 9.48729K wps
Begin Testing...
[Epoch 188] train avg loss 0.000627965, dev acc 0.9144, dev avg loss 0.231693, throughput 9.43137K wps
[Epoch 189 Batch 30/162] avg loss 0.000600711, throughput 9.48535K wps
[Epoch 189 Batch 60/162] avg loss 0.000611298, throughput 9.56368K wps
[Epoch 189 Batch 90/162] avg loss 0.000626991, throughput 9.52525K wps
[Epoch 189 Batch 120/162] avg loss 0.000621029, throughput 9.2231K wps
[Epoch 189 Batch 150/162] avg loss 0.000637896, throughput 9.35668K wps
Begin Testing...
[Epoch 189] train avg loss 0.000618758, dev acc 0.9167, dev avg loss 0.230639, throughput 9.42505K wps
Observed Improvement.
Begin Testing...
[Epoch 190 Batch 30/162] avg loss 0.000624874, throughput 9.54183K wps
[Epoch 190 Batch 60/162] avg loss 0.000560513, throughput 9.40575K wps
[Epoch 190 Batch 90/162] avg loss 0.000652188, throughput 9.51431K wps
[Epoch 190 Batch 120/162] avg loss 0.000584405, throughput 9.4141K wps
[Epoch 190 Batch 150/162] avg loss 0.000712653, throughput 9.44647K wps
Begin Testing...
[Epoch 190] train avg loss 0.000629605, dev acc 0.9133, dev avg loss 0.229564, throughput 9.44716K wps
[Epoch 191 Batch 30/162] avg loss 0.000620245, throughput 9.60984K wps
[Epoch 191 Batch 60/162] avg loss 0.000516378, throughput 9.58168K wps
[Epoch 191 Batch 90/162] avg loss 0.000680764, throughput 9.4097K wps
[Epoch 191 Batch 120/162] avg loss 0.000596497, throughput 9.57788K wps
[Epoch 191 Batch 150/162] avg loss 0.000620525, throughput 9.29172K wps
Begin Testing...
[Epoch 191] train avg loss 0.000608175, dev acc 0.9144, dev avg loss 0.229937, throughput 9.47546K wps
[Epoch 192 Batch 30/162] avg loss 0.00060438, throughput 9.48986K wps
[Epoch 192 Batch 60/162] avg loss 0.000596348, throughput 9.37512K wps
[Epoch 192 Batch 90/162] avg loss 0.000547304, throughput 9.54061K wps
[Epoch 192 Batch 120/162] avg loss 0.00063219, throughput 9.56818K wps
[Epoch 192 Batch 150/162] avg loss 0.000675788, throughput 9.49593K wps
Begin Testing...
[Epoch 192] train avg loss 0.000610182, dev acc 0.9133, dev avg loss 0.230462, throughput 9.50754K wps
[Epoch 193 Batch 30/162] avg loss 0.00059349, throughput 9.7235K wps
[Epoch 193 Batch 60/162] avg loss 0.000640229, throughput 9.38649K wps
[Epoch 193 Batch 90/162] avg loss 0.00060801, throughput 9.34794K wps
[Epoch 193 Batch 120/162] avg loss 0.000617473, throughput 9.50518K wps
[Epoch 193 Batch 150/162] avg loss 0.000631303, throughput 9.24147K wps
Begin Testing...
[Epoch 193] train avg loss 0.000617701, dev acc 0.9133, dev avg loss 0.23049, throughput 9.45333K wps
[Epoch 194 Batch 30/162] avg loss 0.000614025, throughput 9.46173K wps
[Epoch 194 Batch 60/162] avg loss 0.000560478, throughput 9.35267K wps
[Epoch 194 Batch 90/162] avg loss 0.000584585, throughput 9.40261K wps
[Epoch 194 Batch 120/162] avg loss 0.00070845, throughput 9.55765K wps
[Epoch 194 Batch 150/162] avg loss 0.000635731, throughput 9.32915K wps
Begin Testing...
[Epoch 194] train avg loss 0.00062067, dev acc 0.9144, dev avg loss 0.230921, throughput 9.43586K wps
[Epoch 195 Batch 30/162] avg loss 0.000529091, throughput 9.72206K wps
[Epoch 195 Batch 60/162] avg loss 0.0005817, throughput 9.50183K wps
[Epoch 195 Batch 90/162] avg loss 0.000544246, throughput 9.555K wps
[Epoch 195 Batch 120/162] avg loss 0.000610568, throughput 9.34083K wps
[Epoch 195 Batch 150/162] avg loss 0.000659577, throughput 9.50525K wps
Begin Testing...
[Epoch 195] train avg loss 0.000587885, dev acc 0.9144, dev avg loss 0.231731, throughput 9.51089K wps
[Epoch 196 Batch 30/162] avg loss 0.000585364, throughput 9.7785K wps
[Epoch 196 Batch 60/162] avg loss 0.000572929, throughput 9.4963K wps
[Epoch 196 Batch 90/162] avg loss 0.000661969, throughput 9.66917K wps
[Epoch 196 Batch 120/162] avg loss 0.000586805, throughput 9.37868K wps
[Epoch 196 Batch 150/162] avg loss 0.000574936, throughput 9.62535K wps
Begin Testing...
[Epoch 196] train avg loss 0.000597396, dev acc 0.9156, dev avg loss 0.231631, throughput 9.57744K wps
[Epoch 197 Batch 30/162] avg loss 0.000605318, throughput 9.64442K wps
[Epoch 197 Batch 60/162] avg loss 0.000623136, throughput 9.22174K wps
[Epoch 197 Batch 90/162] avg loss 0.000598589, throughput 9.31682K wps
[Epoch 197 Batch 120/162] avg loss 0.000504173, throughput 9.29053K wps
[Epoch 197 Batch 150/162] avg loss 0.000576957, throughput 9.43011K wps
Begin Testing...
[Epoch 197] train avg loss 0.00058162, dev acc 0.9144, dev avg loss 0.231824, throughput 9.37971K wps
[Epoch 198 Batch 30/162] avg loss 0.00058581, throughput 9.58231K wps
[Epoch 198 Batch 60/162] avg loss 0.000610888, throughput 9.44307K wps
[Epoch 198 Batch 90/162] avg loss 0.000584477, throughput 9.64912K wps
[Epoch 198 Batch 120/162] avg loss 0.000529232, throughput 9.53674K wps
[Epoch 198 Batch 150/162] avg loss 0.000556629, throughput 9.39602K wps
Begin Testing...
[Epoch 198] train avg loss 0.000572314, dev acc 0.9144, dev avg loss 0.231722, throughput 9.51045K wps
[Epoch 199 Batch 30/162] avg loss 0.000616678, throughput 9.47371K wps
[Epoch 199 Batch 60/162] avg loss 0.000555414, throughput 9.29538K wps
[Epoch 199 Batch 90/162] avg loss 0.000632364, throughput 9.26944K wps
[Epoch 199 Batch 120/162] avg loss 0.00055193, throughput 9.47122K wps
[Epoch 199 Batch 150/162] avg loss 0.000623709, throughput 9.3037K wps
Begin Testing...
[Epoch 199] train avg loss 0.000594277, dev acc 0.9167, dev avg loss 0.232569, throughput 9.37214K wps
Observed Improvement.
Begin Testing...
Test loss 0.231089, test acc 0.9090
Total time cost 440.40s
K-fold cross valid avg acc 0.9176