Permalink
Find file Copy path
Fetching contributors…
Cannot retrieve contributors at this time
4381 lines (4380 sloc) 278 KB
Namespace(batch_size=50, data_name='MR', dropout=0.5, epochs=60, gpu=0, log_interval=30, lr=0.0001, model_mode='rand', save_prefix='sa-model')
Use gpu0
2320
56
Done! Tokenizing Time=0.98s, #Sentences=10662
SentimentNet(
(embedding): Embedding(18768 -> 300, float32)
(encoder): ConvolutionalEncoder(
(_convs): HybridConcurrent(
(0): HybridSequential(
(0): Conv1D(300 -> 100, kernel_size=(3,), stride=(1,))
(1): Activation(relu)
(2): HybridLambda(<lambda>)
)
(1): HybridSequential(
(0): Conv1D(300 -> 100, kernel_size=(4,), stride=(1,))
(1): Activation(relu)
(2): HybridLambda(<lambda>)
)
(2): HybridSequential(
(0): Conv1D(300 -> 100, kernel_size=(5,), stride=(1,))
(1): Activation(relu)
(2): HybridLambda(<lambda>)
)
)
)
(output): HybridSequential(
(0): Dropout(p = 0.5, axes=())
(1): Dense(None -> 2, linear)
)
)
9596 1066
[Epoch 0 Batch 30/173] avg loss 0.0138663, throughput 3.72594K wps
[Epoch 0 Batch 60/173] avg loss 0.01384, throughput 6.01219K wps
[Epoch 0 Batch 90/173] avg loss 0.0138427, throughput 6.01463K wps
[Epoch 0 Batch 120/173] avg loss 0.0137913, throughput 6.02111K wps
[Epoch 0 Batch 150/173] avg loss 0.0137712, throughput 6.0258K wps
Begin Testing...
[Epoch 0] train avg loss 0.0138286, test acc 0.5552, test avg loss 0.688479, throughput 5.11903K wps
Observed Improvement.
Begin Testing...
[Epoch 1 Batch 30/173] avg loss 0.0134876, throughput 6.16784K wps
[Epoch 1 Batch 60/173] avg loss 0.0134709, throughput 6.01262K wps
[Epoch 1 Batch 90/173] avg loss 0.0134347, throughput 6.00654K wps
[Epoch 1 Batch 120/173] avg loss 0.0133716, throughput 6.02738K wps
[Epoch 1 Batch 150/173] avg loss 0.0133116, throughput 6.0163K wps
Begin Testing...
[Epoch 1] train avg loss 0.0134216, test acc 0.6427, test avg loss 0.674643, throughput 6.04198K wps
Observed Improvement.
Begin Testing...
[Epoch 2 Batch 30/173] avg loss 0.0129061, throughput 6.16312K wps
[Epoch 2 Batch 60/173] avg loss 0.0128499, throughput 6.00346K wps
[Epoch 2 Batch 90/173] avg loss 0.0126704, throughput 6.0043K wps
[Epoch 2 Batch 120/173] avg loss 0.0125033, throughput 6.01644K wps
[Epoch 2 Batch 150/173] avg loss 0.0123288, throughput 6.01193K wps
Begin Testing...
[Epoch 2] train avg loss 0.012616, test acc 0.6875, test avg loss 0.639405, throughput 6.03539K wps
Observed Improvement.
Begin Testing...
[Epoch 3 Batch 30/173] avg loss 0.0115978, throughput 6.16346K wps
[Epoch 3 Batch 60/173] avg loss 0.011352, throughput 6.0139K wps
[Epoch 3 Batch 90/173] avg loss 0.0109941, throughput 6.00198K wps
[Epoch 3 Batch 120/173] avg loss 0.0107991, throughput 6.02547K wps
[Epoch 3 Batch 150/173] avg loss 0.0104387, throughput 6.02651K wps
Begin Testing...
[Epoch 3] train avg loss 0.0109618, test acc 0.7281, test avg loss 0.573414, throughput 6.0415K wps
Observed Improvement.
Begin Testing...
[Epoch 4 Batch 30/173] avg loss 0.00916487, throughput 6.16518K wps
[Epoch 4 Batch 60/173] avg loss 0.00869746, throughput 6.0297K wps
[Epoch 4 Batch 90/173] avg loss 0.00873478, throughput 6.02154K wps
[Epoch 4 Batch 120/173] avg loss 0.0081594, throughput 6.00497K wps
[Epoch 4 Batch 150/173] avg loss 0.00802919, throughput 6.0007K wps
Begin Testing...
[Epoch 4] train avg loss 0.00849013, test acc 0.7552, test avg loss 0.507353, throughput 6.03927K wps
Observed Improvement.
Begin Testing...
[Epoch 5 Batch 30/173] avg loss 0.00646913, throughput 6.15389K wps
[Epoch 5 Batch 60/173] avg loss 0.00625724, throughput 6.00668K wps
[Epoch 5 Batch 90/173] avg loss 0.00623466, throughput 6.02239K wps
[Epoch 5 Batch 120/173] avg loss 0.00581177, throughput 6.01939K wps
[Epoch 5 Batch 150/173] avg loss 0.0062897, throughput 6.01606K wps
Begin Testing...
[Epoch 5] train avg loss 0.00616796, test acc 0.7635, test avg loss 0.483403, throughput 6.03815K wps
Observed Improvement.
Begin Testing...
[Epoch 6 Batch 30/173] avg loss 0.00432039, throughput 6.16369K wps
[Epoch 6 Batch 60/173] avg loss 0.00444111, throughput 5.99162K wps
[Epoch 6 Batch 90/173] avg loss 0.00420086, throughput 6.00192K wps
[Epoch 6 Batch 120/173] avg loss 0.00417586, throughput 6.00353K wps
[Epoch 6 Batch 150/173] avg loss 0.00437345, throughput 5.99543K wps
Begin Testing...
[Epoch 6] train avg loss 0.0043086, test acc 0.7771, test avg loss 0.476477, throughput 6.02597K wps
Observed Improvement.
Begin Testing...
[Epoch 7 Batch 30/173] avg loss 0.00306025, throughput 6.14647K wps
[Epoch 7 Batch 60/173] avg loss 0.00310503, throughput 6.02526K wps
[Epoch 7 Batch 90/173] avg loss 0.00294841, throughput 6.01618K wps
[Epoch 7 Batch 120/173] avg loss 0.0029336, throughput 6.01038K wps
[Epoch 7 Batch 150/173] avg loss 0.00286519, throughput 6.01374K wps
Begin Testing...
[Epoch 7] train avg loss 0.00301641, test acc 0.7698, test avg loss 0.492394, throughput 6.04025K wps
[Epoch 8 Batch 30/173] avg loss 0.0022519, throughput 6.15191K wps
[Epoch 8 Batch 60/173] avg loss 0.00185816, throughput 5.99804K wps
[Epoch 8 Batch 90/173] avg loss 0.00205194, throughput 6.00739K wps
[Epoch 8 Batch 120/173] avg loss 0.00212332, throughput 6.00117K wps
[Epoch 8 Batch 150/173] avg loss 0.00223871, throughput 6.00703K wps
Begin Testing...
[Epoch 8] train avg loss 0.00212722, test acc 0.7698, test avg loss 0.510844, throughput 6.02841K wps
[Epoch 9 Batch 30/173] avg loss 0.00159047, throughput 6.14107K wps
[Epoch 9 Batch 60/173] avg loss 0.00144383, throughput 5.99596K wps
[Epoch 9 Batch 90/173] avg loss 0.00147419, throughput 5.98307K wps
[Epoch 9 Batch 120/173] avg loss 0.00142353, throughput 6.00105K wps
[Epoch 9 Batch 150/173] avg loss 0.00150474, throughput 6.01144K wps
Begin Testing...
[Epoch 9] train avg loss 0.00149399, test acc 0.7667, test avg loss 0.547173, throughput 6.02534K wps
[Epoch 10 Batch 30/173] avg loss 0.0010377, throughput 6.15476K wps
[Epoch 10 Batch 60/173] avg loss 0.00101173, throughput 5.99297K wps
[Epoch 10 Batch 90/173] avg loss 0.00110735, throughput 6.00052K wps
[Epoch 10 Batch 120/173] avg loss 0.00111579, throughput 5.99242K wps
[Epoch 10 Batch 150/173] avg loss 0.0010183, throughput 5.99364K wps
Begin Testing...
[Epoch 10] train avg loss 0.00105892, test acc 0.7646, test avg loss 0.567115, throughput 6.02343K wps
[Epoch 11 Batch 30/173] avg loss 0.000789393, throughput 6.14418K wps
[Epoch 11 Batch 60/173] avg loss 0.000767772, throughput 6.00039K wps
[Epoch 11 Batch 90/173] avg loss 0.000651349, throughput 6.01644K wps
[Epoch 11 Batch 120/173] avg loss 0.000688339, throughput 6.0174K wps
[Epoch 11 Batch 150/173] avg loss 0.000921849, throughput 5.9967K wps
Begin Testing...
[Epoch 11] train avg loss 0.000777474, test acc 0.7646, test avg loss 0.597798, throughput 6.03072K wps
[Epoch 12 Batch 30/173] avg loss 0.000589077, throughput 6.13724K wps
[Epoch 12 Batch 60/173] avg loss 0.000532233, throughput 5.99594K wps
[Epoch 12 Batch 90/173] avg loss 0.000597025, throughput 5.99486K wps
[Epoch 12 Batch 120/173] avg loss 0.000578025, throughput 5.99978K wps
[Epoch 12 Batch 150/173] avg loss 0.000594435, throughput 5.99363K wps
Begin Testing...
[Epoch 12] train avg loss 0.000575816, test acc 0.7521, test avg loss 0.64178, throughput 6.02077K wps
[Epoch 13 Batch 30/173] avg loss 0.000410284, throughput 6.16032K wps
[Epoch 13 Batch 60/173] avg loss 0.000385119, throughput 6.0065K wps
[Epoch 13 Batch 90/173] avg loss 0.000428699, throughput 5.99205K wps
[Epoch 13 Batch 120/173] avg loss 0.000456351, throughput 5.99612K wps
[Epoch 13 Batch 150/173] avg loss 0.000450211, throughput 5.99372K wps
Begin Testing...
[Epoch 13] train avg loss 0.000428308, test acc 0.7500, test avg loss 0.667999, throughput 6.02574K wps
[Epoch 14 Batch 30/173] avg loss 0.000295147, throughput 6.1433K wps
[Epoch 14 Batch 60/173] avg loss 0.000347313, throughput 5.99611K wps
[Epoch 14 Batch 90/173] avg loss 0.000312657, throughput 5.99535K wps
[Epoch 14 Batch 120/173] avg loss 0.000313198, throughput 6.00747K wps
[Epoch 14 Batch 150/173] avg loss 0.000373344, throughput 6.00662K wps
Begin Testing...
[Epoch 14] train avg loss 0.00033112, test acc 0.7521, test avg loss 0.695875, throughput 6.02506K wps
[Epoch 15 Batch 30/173] avg loss 0.000255418, throughput 5.94246K wps
[Epoch 15 Batch 60/173] avg loss 0.000270975, throughput 5.99191K wps
[Epoch 15 Batch 90/173] avg loss 0.000294063, throughput 5.99127K wps
[Epoch 15 Batch 120/173] avg loss 0.00025151, throughput 5.98894K wps
[Epoch 15 Batch 150/173] avg loss 0.000244447, throughput 6.00218K wps
Begin Testing...
[Epoch 15] train avg loss 0.000273261, test acc 0.7490, test avg loss 0.719342, throughput 5.98556K wps
[Epoch 16 Batch 30/173] avg loss 0.000196549, throughput 6.14136K wps
[Epoch 16 Batch 60/173] avg loss 0.00020926, throughput 5.99647K wps
[Epoch 16 Batch 90/173] avg loss 0.000205823, throughput 5.99412K wps
[Epoch 16 Batch 120/173] avg loss 0.000204346, throughput 5.99045K wps
[Epoch 16 Batch 150/173] avg loss 0.000211202, throughput 5.99098K wps
Begin Testing...
[Epoch 16] train avg loss 0.00020478, test acc 0.7500, test avg loss 0.751002, throughput 6.01726K wps
[Epoch 17 Batch 30/173] avg loss 0.000160783, throughput 6.12904K wps
[Epoch 17 Batch 60/173] avg loss 0.000196753, throughput 5.99286K wps
[Epoch 17 Batch 90/173] avg loss 0.000161594, throughput 5.98712K wps
[Epoch 17 Batch 120/173] avg loss 0.000154192, throughput 5.99226K wps
[Epoch 17 Batch 150/173] avg loss 0.000160415, throughput 5.98006K wps
Begin Testing...
[Epoch 17] train avg loss 0.000165575, test acc 0.7490, test avg loss 0.772001, throughput 6.01361K wps
[Epoch 18 Batch 30/173] avg loss 0.000134419, throughput 6.1405K wps
[Epoch 18 Batch 60/173] avg loss 0.000119324, throughput 5.99279K wps
[Epoch 18 Batch 90/173] avg loss 0.000157689, throughput 5.99241K wps
[Epoch 18 Batch 120/173] avg loss 0.000154025, throughput 5.98233K wps
[Epoch 18 Batch 150/173] avg loss 0.000114976, throughput 5.97739K wps
Begin Testing...
[Epoch 18] train avg loss 0.000134838, test acc 0.7469, test avg loss 0.797028, throughput 6.01282K wps
[Epoch 19 Batch 30/173] avg loss 0.000118597, throughput 6.1356K wps
[Epoch 19 Batch 60/173] avg loss 0.000123039, throughput 5.98507K wps
[Epoch 19 Batch 90/173] avg loss 8.97685e-05, throughput 5.99332K wps
[Epoch 19 Batch 120/173] avg loss 0.000136381, throughput 5.99655K wps
[Epoch 19 Batch 150/173] avg loss 0.000112636, throughput 5.98914K wps
Begin Testing...
[Epoch 19] train avg loss 0.000117975, test acc 0.7479, test avg loss 0.823194, throughput 6.01716K wps
[Epoch 20 Batch 30/173] avg loss 8.54718e-05, throughput 6.14243K wps
[Epoch 20 Batch 60/173] avg loss 0.00010435, throughput 5.98299K wps
[Epoch 20 Batch 90/173] avg loss 0.000100516, throughput 5.99047K wps
[Epoch 20 Batch 120/173] avg loss 9.27294e-05, throughput 5.97629K wps
[Epoch 20 Batch 150/173] avg loss 0.000105203, throughput 5.99707K wps
Begin Testing...
[Epoch 20] train avg loss 9.74868e-05, test acc 0.7458, test avg loss 0.841995, throughput 6.01434K wps
[Epoch 21 Batch 30/173] avg loss 7.49131e-05, throughput 6.12865K wps
[Epoch 21 Batch 60/173] avg loss 8.66698e-05, throughput 5.98095K wps
[Epoch 21 Batch 90/173] avg loss 8.45976e-05, throughput 5.99029K wps
[Epoch 21 Batch 120/173] avg loss 8.08137e-05, throughput 5.99499K wps
[Epoch 21 Batch 150/173] avg loss 7.41076e-05, throughput 5.98549K wps
Begin Testing...
[Epoch 21] train avg loss 8.02525e-05, test acc 0.7458, test avg loss 0.870026, throughput 6.01256K wps
[Epoch 22 Batch 30/173] avg loss 6.5336e-05, throughput 6.1341K wps
[Epoch 22 Batch 60/173] avg loss 5.96374e-05, throughput 5.99215K wps
[Epoch 22 Batch 90/173] avg loss 6.44118e-05, throughput 6.00006K wps
[Epoch 22 Batch 120/173] avg loss 7.46332e-05, throughput 5.98857K wps
[Epoch 22 Batch 150/173] avg loss 6.33642e-05, throughput 5.99134K wps
Begin Testing...
[Epoch 22] train avg loss 6.67311e-05, test acc 0.7427, test avg loss 0.889908, throughput 6.01517K wps
[Epoch 23 Batch 30/173] avg loss 5.48449e-05, throughput 6.13053K wps
[Epoch 23 Batch 60/173] avg loss 5.34753e-05, throughput 5.97648K wps
[Epoch 23 Batch 90/173] avg loss 6.34939e-05, throughput 5.98852K wps
[Epoch 23 Batch 120/173] avg loss 5.0065e-05, throughput 5.99293K wps
[Epoch 23 Batch 150/173] avg loss 7.20023e-05, throughput 5.98679K wps
Begin Testing...
[Epoch 23] train avg loss 5.83752e-05, test acc 0.7458, test avg loss 0.91291, throughput 6.01107K wps
[Epoch 24 Batch 30/173] avg loss 4.4815e-05, throughput 6.14247K wps
[Epoch 24 Batch 60/173] avg loss 5.61718e-05, throughput 5.97834K wps
[Epoch 24 Batch 90/173] avg loss 5.23167e-05, throughput 5.99864K wps
[Epoch 24 Batch 120/173] avg loss 5.67558e-05, throughput 5.98673K wps
[Epoch 24 Batch 150/173] avg loss 4.56523e-05, throughput 5.97711K wps
Begin Testing...
[Epoch 24] train avg loss 5.22616e-05, test acc 0.7438, test avg loss 0.940106, throughput 6.0132K wps
[Epoch 25 Batch 30/173] avg loss 4.54875e-05, throughput 6.14166K wps
[Epoch 25 Batch 60/173] avg loss 3.97646e-05, throughput 5.9943K wps
[Epoch 25 Batch 90/173] avg loss 4.99732e-05, throughput 5.98019K wps
[Epoch 25 Batch 120/173] avg loss 3.84397e-05, throughput 5.99197K wps
[Epoch 25 Batch 150/173] avg loss 3.59919e-05, throughput 5.98279K wps
Begin Testing...
[Epoch 25] train avg loss 4.22244e-05, test acc 0.7458, test avg loss 0.954335, throughput 6.01517K wps
[Epoch 26 Batch 30/173] avg loss 3.46288e-05, throughput 6.13666K wps
[Epoch 26 Batch 60/173] avg loss 4.26281e-05, throughput 5.99053K wps
[Epoch 26 Batch 90/173] avg loss 3.75747e-05, throughput 5.99276K wps
[Epoch 26 Batch 120/173] avg loss 3.71634e-05, throughput 5.96541K wps
[Epoch 26 Batch 150/173] avg loss 4.18739e-05, throughput 5.96176K wps
Begin Testing...
[Epoch 26] train avg loss 3.79524e-05, test acc 0.7448, test avg loss 0.981031, throughput 5.99477K wps
[Epoch 27 Batch 30/173] avg loss 4.08792e-05, throughput 6.13457K wps
[Epoch 27 Batch 60/173] avg loss 3.40331e-05, throughput 5.9778K wps
[Epoch 27 Batch 90/173] avg loss 3.18622e-05, throughput 5.99134K wps
[Epoch 27 Batch 120/173] avg loss 4.14976e-05, throughput 5.99145K wps
[Epoch 27 Batch 150/173] avg loss 3.07537e-05, throughput 5.99222K wps
Begin Testing...
[Epoch 27] train avg loss 3.52624e-05, test acc 0.7490, test avg loss 0.999573, throughput 6.0138K wps
[Epoch 28 Batch 30/173] avg loss 3.75177e-05, throughput 6.12763K wps
[Epoch 28 Batch 60/173] avg loss 3.02113e-05, throughput 5.9836K wps
[Epoch 28 Batch 90/173] avg loss 2.64729e-05, throughput 5.98904K wps
[Epoch 28 Batch 120/173] avg loss 3.08645e-05, throughput 5.99245K wps
[Epoch 28 Batch 150/173] avg loss 2.51235e-05, throughput 5.98074K wps
Begin Testing...
[Epoch 28] train avg loss 2.98342e-05, test acc 0.7479, test avg loss 1.01503, throughput 6.01095K wps
[Epoch 29 Batch 30/173] avg loss 2.44871e-05, throughput 6.1323K wps
[Epoch 29 Batch 60/173] avg loss 2.30164e-05, throughput 5.97732K wps
[Epoch 29 Batch 90/173] avg loss 3.34841e-05, throughput 5.99236K wps
[Epoch 29 Batch 120/173] avg loss 2.83044e-05, throughput 5.98899K wps
[Epoch 29 Batch 150/173] avg loss 2.64018e-05, throughput 5.99988K wps
Begin Testing...
[Epoch 29] train avg loss 2.69702e-05, test acc 0.7469, test avg loss 1.03354, throughput 6.01552K wps
[Epoch 30 Batch 30/173] avg loss 1.96031e-05, throughput 6.14553K wps
[Epoch 30 Batch 60/173] avg loss 2.18225e-05, throughput 5.99507K wps
[Epoch 30 Batch 90/173] avg loss 2.66564e-05, throughput 5.98967K wps
[Epoch 30 Batch 120/173] avg loss 3.26055e-05, throughput 5.98175K wps
[Epoch 30 Batch 150/173] avg loss 2.1263e-05, throughput 5.9844K wps
Begin Testing...
[Epoch 30] train avg loss 2.39351e-05, test acc 0.7458, test avg loss 1.04932, throughput 6.01507K wps
[Epoch 31 Batch 30/173] avg loss 2.00904e-05, throughput 6.14671K wps
[Epoch 31 Batch 60/173] avg loss 2.01522e-05, throughput 5.98779K wps
[Epoch 31 Batch 90/173] avg loss 1.78898e-05, throughput 5.99904K wps
[Epoch 31 Batch 120/173] avg loss 1.69605e-05, throughput 5.99599K wps
[Epoch 31 Batch 150/173] avg loss 3.41376e-05, throughput 5.98191K wps
Begin Testing...
[Epoch 31] train avg loss 2.13254e-05, test acc 0.7490, test avg loss 1.06881, throughput 6.01831K wps
[Epoch 32 Batch 30/173] avg loss 2.00622e-05, throughput 6.13696K wps
[Epoch 32 Batch 60/173] avg loss 2.73735e-05, throughput 5.98087K wps
[Epoch 32 Batch 90/173] avg loss 1.82839e-05, throughput 5.97596K wps
[Epoch 32 Batch 120/173] avg loss 1.51228e-05, throughput 5.97982K wps
[Epoch 32 Batch 150/173] avg loss 1.83575e-05, throughput 5.99344K wps
Begin Testing...
[Epoch 32] train avg loss 1.94936e-05, test acc 0.7490, test avg loss 1.08293, throughput 6.01121K wps
[Epoch 33 Batch 30/173] avg loss 1.54905e-05, throughput 6.12511K wps
[Epoch 33 Batch 60/173] avg loss 1.28845e-05, throughput 5.98048K wps
[Epoch 33 Batch 90/173] avg loss 2.3662e-05, throughput 5.98712K wps
[Epoch 33 Batch 120/173] avg loss 1.83802e-05, throughput 5.98452K wps
[Epoch 33 Batch 150/173] avg loss 1.43578e-05, throughput 5.99668K wps
Begin Testing...
[Epoch 33] train avg loss 1.67971e-05, test acc 0.7490, test avg loss 1.09676, throughput 6.00916K wps
[Epoch 34 Batch 30/173] avg loss 1.27501e-05, throughput 6.13956K wps
[Epoch 34 Batch 60/173] avg loss 1.13142e-05, throughput 5.97821K wps
[Epoch 34 Batch 90/173] avg loss 1.17651e-05, throughput 5.9829K wps
[Epoch 34 Batch 120/173] avg loss 1.41691e-05, throughput 5.98851K wps
[Epoch 34 Batch 150/173] avg loss 2.10703e-05, throughput 5.97979K wps
Begin Testing...
[Epoch 34] train avg loss 1.42556e-05, test acc 0.7438, test avg loss 1.11468, throughput 6.00695K wps
[Epoch 35 Batch 30/173] avg loss 1.20684e-05, throughput 6.14344K wps
[Epoch 35 Batch 60/173] avg loss 1.0243e-05, throughput 5.99224K wps
[Epoch 35 Batch 90/173] avg loss 1.42512e-05, throughput 5.98365K wps
[Epoch 35 Batch 120/173] avg loss 2.03445e-05, throughput 5.99527K wps
[Epoch 35 Batch 150/173] avg loss 9.73932e-06, throughput 5.97915K wps
Begin Testing...
[Epoch 35] train avg loss 1.31414e-05, test acc 0.7479, test avg loss 1.1293, throughput 6.0167K wps
[Epoch 36 Batch 30/173] avg loss 1.72159e-05, throughput 6.13615K wps
[Epoch 36 Batch 60/173] avg loss 9.0319e-06, throughput 5.98689K wps
[Epoch 36 Batch 90/173] avg loss 1.15552e-05, throughput 5.98257K wps
[Epoch 36 Batch 120/173] avg loss 1.04341e-05, throughput 5.98594K wps
[Epoch 36 Batch 150/173] avg loss 1.25111e-05, throughput 5.98707K wps
Begin Testing...
[Epoch 36] train avg loss 1.23341e-05, test acc 0.7448, test avg loss 1.14704, throughput 6.01132K wps
[Epoch 37 Batch 30/173] avg loss 1.75568e-05, throughput 6.13712K wps
[Epoch 37 Batch 60/173] avg loss 8.81866e-06, throughput 5.98489K wps
[Epoch 37 Batch 90/173] avg loss 9.02484e-06, throughput 5.98585K wps
[Epoch 37 Batch 120/173] avg loss 1.08091e-05, throughput 5.99262K wps
[Epoch 37 Batch 150/173] avg loss 8.16078e-06, throughput 5.98927K wps
Begin Testing...
[Epoch 37] train avg loss 1.06151e-05, test acc 0.7438, test avg loss 1.15332, throughput 6.01335K wps
[Epoch 38 Batch 30/173] avg loss 6.76838e-06, throughput 6.1393K wps
[Epoch 38 Batch 60/173] avg loss 9.0836e-06, throughput 5.99265K wps
[Epoch 38 Batch 90/173] avg loss 7.58223e-06, throughput 5.97881K wps
[Epoch 38 Batch 120/173] avg loss 1.63722e-05, throughput 5.97448K wps
[Epoch 38 Batch 150/173] avg loss 1.1754e-05, throughput 5.98571K wps
Begin Testing...
[Epoch 38] train avg loss 9.90776e-06, test acc 0.7469, test avg loss 1.16439, throughput 6.01099K wps
[Epoch 39 Batch 30/173] avg loss 7.32741e-06, throughput 6.15154K wps
[Epoch 39 Batch 60/173] avg loss 4.99044e-06, throughput 5.99296K wps
[Epoch 39 Batch 90/173] avg loss 6.69811e-06, throughput 5.97248K wps
[Epoch 39 Batch 120/173] avg loss 9.38529e-06, throughput 5.97217K wps
[Epoch 39 Batch 150/173] avg loss 1.49208e-05, throughput 5.97843K wps
Begin Testing...
[Epoch 39] train avg loss 9.55714e-06, test acc 0.7500, test avg loss 1.19866, throughput 6.01136K wps
[Epoch 40 Batch 30/173] avg loss 6.54827e-06, throughput 6.13682K wps
[Epoch 40 Batch 60/173] avg loss 7.24319e-06, throughput 5.99283K wps
[Epoch 40 Batch 90/173] avg loss 6.36225e-06, throughput 5.98057K wps
[Epoch 40 Batch 120/173] avg loss 9.02621e-06, throughput 5.97303K wps
[Epoch 40 Batch 150/173] avg loss 7.1238e-06, throughput 5.99116K wps
Begin Testing...
[Epoch 40] train avg loss 8.72178e-06, test acc 0.7490, test avg loss 1.20631, throughput 6.01174K wps
[Epoch 41 Batch 30/173] avg loss 4.96356e-06, throughput 6.14209K wps
[Epoch 41 Batch 60/173] avg loss 7.03632e-06, throughput 5.98148K wps
[Epoch 41 Batch 90/173] avg loss 4.72731e-06, throughput 5.9886K wps
[Epoch 41 Batch 120/173] avg loss 1.63185e-05, throughput 5.98643K wps
[Epoch 41 Batch 150/173] avg loss 5.78836e-06, throughput 5.98643K wps
Begin Testing...
[Epoch 41] train avg loss 7.48619e-06, test acc 0.7469, test avg loss 1.22978, throughput 6.01128K wps
[Epoch 42 Batch 30/173] avg loss 4.73511e-06, throughput 6.12259K wps
[Epoch 42 Batch 60/173] avg loss 5.94757e-06, throughput 5.97701K wps
[Epoch 42 Batch 90/173] avg loss 7.4258e-06, throughput 5.98503K wps
[Epoch 42 Batch 120/173] avg loss 7.99743e-06, throughput 5.98442K wps
[Epoch 42 Batch 150/173] avg loss 4.1586e-06, throughput 5.97878K wps
Begin Testing...
[Epoch 42] train avg loss 7.52771e-06, test acc 0.7521, test avg loss 1.24944, throughput 6.00543K wps
[Epoch 43 Batch 30/173] avg loss 5.00494e-06, throughput 6.12803K wps
[Epoch 43 Batch 60/173] avg loss 3.96408e-06, throughput 5.98754K wps
[Epoch 43 Batch 90/173] avg loss 1.18437e-05, throughput 5.98504K wps
[Epoch 43 Batch 120/173] avg loss 4.90582e-06, throughput 5.98214K wps
[Epoch 43 Batch 150/173] avg loss 6.36077e-06, throughput 6.0031K wps
Begin Testing...
[Epoch 43] train avg loss 6.04133e-06, test acc 0.7500, test avg loss 1.26532, throughput 6.01263K wps
[Epoch 44 Batch 30/173] avg loss 3.71024e-06, throughput 6.12279K wps
[Epoch 44 Batch 60/173] avg loss 6.36115e-06, throughput 5.98305K wps
[Epoch 44 Batch 90/173] avg loss 1.28621e-05, throughput 5.97677K wps
[Epoch 44 Batch 120/173] avg loss 5.40084e-06, throughput 5.97733K wps
[Epoch 44 Batch 150/173] avg loss 3.55254e-06, throughput 5.99895K wps
Begin Testing...
[Epoch 44] train avg loss 6.11191e-06, test acc 0.7500, test avg loss 1.27526, throughput 6.00833K wps
[Epoch 45 Batch 30/173] avg loss 1.08016e-05, throughput 6.14514K wps
[Epoch 45 Batch 60/173] avg loss 4.06711e-06, throughput 5.98228K wps
[Epoch 45 Batch 90/173] avg loss 3.10493e-06, throughput 5.97284K wps
[Epoch 45 Batch 120/173] avg loss 3.40326e-06, throughput 5.98673K wps
[Epoch 45 Batch 150/173] avg loss 4.08086e-06, throughput 5.98868K wps
Begin Testing...
[Epoch 45] train avg loss 5.01335e-06, test acc 0.7479, test avg loss 1.29468, throughput 6.00981K wps
[Epoch 46 Batch 30/173] avg loss 3.04941e-06, throughput 6.13355K wps
[Epoch 46 Batch 60/173] avg loss 2.67768e-06, throughput 5.98304K wps
[Epoch 46 Batch 90/173] avg loss 3.18853e-06, throughput 5.99509K wps
[Epoch 46 Batch 120/173] avg loss 4.0712e-06, throughput 5.99674K wps
[Epoch 46 Batch 150/173] avg loss 3.30143e-06, throughput 5.99355K wps
Begin Testing...
[Epoch 46] train avg loss 4.3813e-06, test acc 0.7469, test avg loss 1.30907, throughput 6.01606K wps
[Epoch 47 Batch 30/173] avg loss 1.01816e-05, throughput 6.13334K wps
[Epoch 47 Batch 60/173] avg loss 3.3491e-06, throughput 5.97613K wps
[Epoch 47 Batch 90/173] avg loss 3.01495e-06, throughput 5.98153K wps
[Epoch 47 Batch 120/173] avg loss 2.26629e-06, throughput 5.98101K wps
[Epoch 47 Batch 150/173] avg loss 2.92934e-06, throughput 5.9823K wps
Begin Testing...
[Epoch 47] train avg loss 4.16441e-06, test acc 0.7479, test avg loss 1.32519, throughput 6.00661K wps
[Epoch 48 Batch 30/173] avg loss 9.4578e-06, throughput 6.14984K wps
[Epoch 48 Batch 60/173] avg loss 2.3074e-06, throughput 5.994K wps
[Epoch 48 Batch 90/173] avg loss 5.64358e-06, throughput 5.99033K wps
[Epoch 48 Batch 120/173] avg loss 3.40239e-06, throughput 5.97849K wps
[Epoch 48 Batch 150/173] avg loss 2.77191e-06, throughput 5.97673K wps
Begin Testing...
[Epoch 48] train avg loss 4.4599e-06, test acc 0.7500, test avg loss 1.3366, throughput 6.0135K wps
[Epoch 49 Batch 30/173] avg loss 2.41548e-06, throughput 6.12255K wps
[Epoch 49 Batch 60/173] avg loss 3.25016e-06, throughput 5.99407K wps
[Epoch 49 Batch 90/173] avg loss 1.1482e-05, throughput 5.99253K wps
[Epoch 49 Batch 120/173] avg loss 6.2421e-06, throughput 5.98656K wps
[Epoch 49 Batch 150/173] avg loss 4.97433e-06, throughput 5.98445K wps
Begin Testing...
[Epoch 49] train avg loss 5.19313e-06, test acc 0.7417, test avg loss 1.36764, throughput 6.01408K wps
[Epoch 50 Batch 30/173] avg loss 9.38843e-06, throughput 6.13792K wps
[Epoch 50 Batch 60/173] avg loss 1.84576e-06, throughput 5.98415K wps
[Epoch 50 Batch 90/173] avg loss 3.85816e-06, throughput 5.97015K wps
[Epoch 50 Batch 120/173] avg loss 2.39951e-06, throughput 5.98827K wps
[Epoch 50 Batch 150/173] avg loss 3.56237e-06, throughput 5.98162K wps
Begin Testing...
[Epoch 50] train avg loss 4.04949e-06, test acc 0.7448, test avg loss 1.3806, throughput 6.00816K wps
[Epoch 51 Batch 30/173] avg loss 2.24444e-06, throughput 6.12789K wps
[Epoch 51 Batch 60/173] avg loss 1.40408e-06, throughput 5.9796K wps
[Epoch 51 Batch 90/173] avg loss 2.95853e-05, throughput 5.96564K wps
[Epoch 51 Batch 120/173] avg loss 3.34308e-06, throughput 5.98235K wps
[Epoch 51 Batch 150/173] avg loss 2.73853e-06, throughput 5.97511K wps
Begin Testing...
[Epoch 51] train avg loss 8.29432e-06, test acc 0.7510, test avg loss 1.37754, throughput 6.00303K wps
[Epoch 52 Batch 30/173] avg loss 8.58863e-06, throughput 6.13225K wps
[Epoch 52 Batch 60/173] avg loss 2.68751e-06, throughput 5.99511K wps
[Epoch 52 Batch 90/173] avg loss 2.66934e-06, throughput 5.98641K wps
[Epoch 52 Batch 120/173] avg loss 2.63656e-06, throughput 5.97918K wps
[Epoch 52 Batch 150/173] avg loss 1.67469e-06, throughput 5.97545K wps
Begin Testing...
[Epoch 52] train avg loss 3.51543e-06, test acc 0.7500, test avg loss 1.39068, throughput 6.00739K wps
[Epoch 53 Batch 30/173] avg loss 1.59171e-06, throughput 6.12156K wps
[Epoch 53 Batch 60/173] avg loss 1.86776e-06, throughput 5.97762K wps
[Epoch 53 Batch 90/173] avg loss 1.8715e-06, throughput 5.97782K wps
[Epoch 53 Batch 120/173] avg loss 8.04008e-06, throughput 5.98333K wps
[Epoch 53 Batch 150/173] avg loss 1.69925e-06, throughput 5.97769K wps
Begin Testing...
[Epoch 53] train avg loss 2.893e-06, test acc 0.7500, test avg loss 1.40544, throughput 6.0066K wps
[Epoch 54 Batch 30/173] avg loss 1.24346e-06, throughput 6.12625K wps
[Epoch 54 Batch 60/173] avg loss 4.4879e-06, throughput 5.99076K wps
[Epoch 54 Batch 90/173] avg loss 1.80807e-06, throughput 5.9813K wps
[Epoch 54 Batch 120/173] avg loss 1.76172e-06, throughput 5.98841K wps
[Epoch 54 Batch 150/173] avg loss 7.73739e-06, throughput 5.98888K wps
Begin Testing...
[Epoch 54] train avg loss 3.18637e-06, test acc 0.7490, test avg loss 1.42766, throughput 6.01097K wps
[Epoch 55 Batch 30/173] avg loss 1.4566e-06, throughput 6.13508K wps
[Epoch 55 Batch 60/173] avg loss 9.54089e-06, throughput 5.98703K wps
[Epoch 55 Batch 90/173] avg loss 2.09192e-06, throughput 5.98513K wps
[Epoch 55 Batch 120/173] avg loss 1.41592e-06, throughput 5.97957K wps
[Epoch 55 Batch 150/173] avg loss 1.87523e-06, throughput 5.96923K wps
Begin Testing...
[Epoch 55] train avg loss 3.14327e-06, test acc 0.7521, test avg loss 1.42874, throughput 6.00759K wps
[Epoch 56 Batch 30/173] avg loss 1.37204e-06, throughput 6.13795K wps
[Epoch 56 Batch 60/173] avg loss 5.97067e-06, throughput 5.97552K wps
[Epoch 56 Batch 90/173] avg loss 1.232e-06, throughput 5.9758K wps
[Epoch 56 Batch 120/173] avg loss 1.45984e-06, throughput 5.97159K wps
[Epoch 56 Batch 150/173] avg loss 1.7238e-06, throughput 5.97694K wps
Begin Testing...
[Epoch 56] train avg loss 2.23091e-06, test acc 0.7521, test avg loss 1.43861, throughput 5.98949K wps
[Epoch 57 Batch 30/173] avg loss 1.04667e-06, throughput 6.08848K wps
[Epoch 57 Batch 60/173] avg loss 1.17196e-06, throughput 5.98455K wps
[Epoch 57 Batch 90/173] avg loss 9.10633e-06, throughput 5.96643K wps
[Epoch 57 Batch 120/173] avg loss 1.51551e-06, throughput 5.98886K wps
[Epoch 57 Batch 150/173] avg loss 1.52303e-06, throughput 5.98348K wps
Begin Testing...
[Epoch 57] train avg loss 2.79604e-06, test acc 0.7521, test avg loss 1.45257, throughput 6.00143K wps
[Epoch 58 Batch 30/173] avg loss 3.29785e-06, throughput 6.13946K wps
[Epoch 58 Batch 60/173] avg loss 1.1843e-06, throughput 5.96992K wps
[Epoch 58 Batch 90/173] avg loss 1.13957e-06, throughput 5.97429K wps
[Epoch 58 Batch 120/173] avg loss 1.27041e-06, throughput 5.98191K wps
[Epoch 58 Batch 150/173] avg loss 1.70323e-06, throughput 5.98638K wps
Begin Testing...
[Epoch 58] train avg loss 2.55667e-06, test acc 0.7521, test avg loss 1.46429, throughput 6.00663K wps
[Epoch 59 Batch 30/173] avg loss 9.41738e-07, throughput 6.1158K wps
[Epoch 59 Batch 60/173] avg loss 1.58796e-06, throughput 5.98126K wps
[Epoch 59 Batch 90/173] avg loss 1.31043e-06, throughput 5.98781K wps
[Epoch 59 Batch 120/173] avg loss 1.3425e-06, throughput 5.97439K wps
[Epoch 59 Batch 150/173] avg loss 1.19255e-06, throughput 5.9873K wps
Begin Testing...
[Epoch 59] train avg loss 2.18131e-06, test acc 0.7500, test avg loss 1.47351, throughput 6.0071K wps
Test loss 0.492248, test acc 0.7730
Total time cost 361.46s
9596 1066
[Epoch 0 Batch 30/173] avg loss 0.0138497, throughput 5.75967K wps
[Epoch 0 Batch 60/173] avg loss 0.0138503, throughput 5.97521K wps
[Epoch 0 Batch 90/173] avg loss 0.013809, throughput 5.97875K wps
[Epoch 0 Batch 120/173] avg loss 0.0138004, throughput 5.9776K wps
[Epoch 0 Batch 150/173] avg loss 0.0137621, throughput 5.97893K wps
Begin Testing...
[Epoch 0] train avg loss 0.0138281, test acc 0.6542, test avg loss 0.686231, throughput 5.94124K wps
Observed Improvement.
Begin Testing...
[Epoch 1 Batch 30/173] avg loss 0.0134944, throughput 6.13394K wps
[Epoch 1 Batch 60/173] avg loss 0.0134781, throughput 5.98525K wps
[Epoch 1 Batch 90/173] avg loss 0.0134635, throughput 5.98705K wps
[Epoch 1 Batch 120/173] avg loss 0.0133645, throughput 5.98388K wps
[Epoch 1 Batch 150/173] avg loss 0.0133371, throughput 5.97942K wps
Begin Testing...
[Epoch 1] train avg loss 0.0134204, test acc 0.6500, test avg loss 0.671684, throughput 6.01004K wps
[Epoch 2 Batch 30/173] avg loss 0.0129501, throughput 6.13189K wps
[Epoch 2 Batch 60/173] avg loss 0.0128152, throughput 5.97935K wps
[Epoch 2 Batch 90/173] avg loss 0.0126632, throughput 5.9752K wps
[Epoch 2 Batch 120/173] avg loss 0.0125892, throughput 5.98022K wps
[Epoch 2 Batch 150/173] avg loss 0.012355, throughput 5.97998K wps
Begin Testing...
[Epoch 2] train avg loss 0.0126371, test acc 0.7031, test avg loss 0.632339, throughput 6.00476K wps
Observed Improvement.
Begin Testing...
[Epoch 3 Batch 30/173] avg loss 0.0116296, throughput 6.14705K wps
[Epoch 3 Batch 60/173] avg loss 0.0112714, throughput 6.00026K wps
[Epoch 3 Batch 90/173] avg loss 0.0109541, throughput 5.99491K wps
[Epoch 3 Batch 120/173] avg loss 0.0106844, throughput 5.96791K wps
[Epoch 3 Batch 150/173] avg loss 0.010394, throughput 5.98284K wps
Begin Testing...
[Epoch 3] train avg loss 0.0109223, test acc 0.7229, test avg loss 0.564962, throughput 6.01448K wps
Observed Improvement.
Begin Testing...
[Epoch 4 Batch 30/173] avg loss 0.00905748, throughput 6.12711K wps
[Epoch 4 Batch 60/173] avg loss 0.00890728, throughput 5.97739K wps
[Epoch 4 Batch 90/173] avg loss 0.00838692, throughput 5.9736K wps
[Epoch 4 Batch 120/173] avg loss 0.00821291, throughput 5.97886K wps
[Epoch 4 Batch 150/173] avg loss 0.00815789, throughput 5.98805K wps
Begin Testing...
[Epoch 4] train avg loss 0.00842166, test acc 0.7469, test avg loss 0.509922, throughput 6.00489K wps
Observed Improvement.
Begin Testing...
[Epoch 5 Batch 30/173] avg loss 0.00628193, throughput 6.12549K wps
[Epoch 5 Batch 60/173] avg loss 0.00601742, throughput 5.98503K wps
[Epoch 5 Batch 90/173] avg loss 0.00619909, throughput 5.98413K wps
[Epoch 5 Batch 120/173] avg loss 0.0062071, throughput 5.98165K wps
[Epoch 5 Batch 150/173] avg loss 0.00610728, throughput 5.98117K wps
Begin Testing...
[Epoch 5] train avg loss 0.00608847, test acc 0.7438, test avg loss 0.495995, throughput 6.00808K wps
[Epoch 6 Batch 30/173] avg loss 0.00430916, throughput 6.12097K wps
[Epoch 6 Batch 60/173] avg loss 0.0043577, throughput 5.97827K wps
[Epoch 6 Batch 90/173] avg loss 0.00424033, throughput 5.97869K wps
[Epoch 6 Batch 120/173] avg loss 0.00431694, throughput 5.98128K wps
[Epoch 6 Batch 150/173] avg loss 0.0043412, throughput 5.98986K wps
Begin Testing...
[Epoch 6] train avg loss 0.00429895, test acc 0.7385, test avg loss 0.509128, throughput 6.00709K wps
[Epoch 7 Batch 30/173] avg loss 0.00328145, throughput 6.12236K wps
[Epoch 7 Batch 60/173] avg loss 0.00294068, throughput 5.97945K wps
[Epoch 7 Batch 90/173] avg loss 0.00299622, throughput 5.9807K wps
[Epoch 7 Batch 120/173] avg loss 0.00288846, throughput 5.9856K wps
[Epoch 7 Batch 150/173] avg loss 0.00292593, throughput 5.98259K wps
Begin Testing...
[Epoch 7] train avg loss 0.00302064, test acc 0.7396, test avg loss 0.54127, throughput 6.00643K wps
[Epoch 8 Batch 30/173] avg loss 0.00216755, throughput 6.14981K wps
[Epoch 8 Batch 60/173] avg loss 0.00214974, throughput 5.98234K wps
[Epoch 8 Batch 90/173] avg loss 0.00199459, throughput 5.99608K wps
[Epoch 8 Batch 120/173] avg loss 0.00216194, throughput 5.97845K wps
[Epoch 8 Batch 150/173] avg loss 0.00225416, throughput 5.98146K wps
Begin Testing...
[Epoch 8] train avg loss 0.00211601, test acc 0.7427, test avg loss 0.585057, throughput 6.01316K wps
[Epoch 9 Batch 30/173] avg loss 0.00143152, throughput 6.15112K wps
[Epoch 9 Batch 60/173] avg loss 0.00148517, throughput 6.00168K wps
[Epoch 9 Batch 90/173] avg loss 0.0014778, throughput 5.99078K wps
[Epoch 9 Batch 120/173] avg loss 0.00140903, throughput 5.99358K wps
[Epoch 9 Batch 150/173] avg loss 0.00146109, throughput 5.98766K wps
Begin Testing...
[Epoch 9] train avg loss 0.00148485, test acc 0.7406, test avg loss 0.638184, throughput 6.02144K wps
[Epoch 10 Batch 30/173] avg loss 0.00102647, throughput 6.1393K wps
[Epoch 10 Batch 60/173] avg loss 0.00103322, throughput 5.99142K wps
[Epoch 10 Batch 90/173] avg loss 0.00110706, throughput 5.99061K wps
[Epoch 10 Batch 120/173] avg loss 0.00110874, throughput 5.99572K wps
[Epoch 10 Batch 150/173] avg loss 0.0009941, throughput 5.99054K wps
Begin Testing...
[Epoch 10] train avg loss 0.00105491, test acc 0.7302, test avg loss 0.687299, throughput 6.01833K wps
[Epoch 11 Batch 30/173] avg loss 0.000728633, throughput 6.13651K wps
[Epoch 11 Batch 60/173] avg loss 0.000668671, throughput 5.99151K wps
[Epoch 11 Batch 90/173] avg loss 0.000726249, throughput 5.99444K wps
[Epoch 11 Batch 120/173] avg loss 0.000765382, throughput 5.9845K wps
[Epoch 11 Batch 150/173] avg loss 0.000919214, throughput 5.97804K wps
Begin Testing...
[Epoch 11] train avg loss 0.000775441, test acc 0.7302, test avg loss 0.736621, throughput 6.01163K wps
[Epoch 12 Batch 30/173] avg loss 0.000562624, throughput 6.137K wps
[Epoch 12 Batch 60/173] avg loss 0.000637525, throughput 5.98619K wps
[Epoch 12 Batch 90/173] avg loss 0.000571319, throughput 5.99005K wps
[Epoch 12 Batch 120/173] avg loss 0.000590993, throughput 5.98613K wps
[Epoch 12 Batch 150/173] avg loss 0.000553301, throughput 5.99117K wps
Begin Testing...
[Epoch 12] train avg loss 0.000584562, test acc 0.7333, test avg loss 0.778313, throughput 6.01481K wps
[Epoch 13 Batch 30/173] avg loss 0.000475012, throughput 6.12886K wps
[Epoch 13 Batch 60/173] avg loss 0.000485105, throughput 5.99995K wps
[Epoch 13 Batch 90/173] avg loss 0.000382215, throughput 5.98252K wps
[Epoch 13 Batch 120/173] avg loss 0.000438795, throughput 5.99191K wps
[Epoch 13 Batch 150/173] avg loss 0.000440286, throughput 5.98532K wps
Begin Testing...
[Epoch 13] train avg loss 0.000443782, test acc 0.7323, test avg loss 0.824109, throughput 6.01518K wps
[Epoch 14 Batch 30/173] avg loss 0.000310817, throughput 6.14863K wps
[Epoch 14 Batch 60/173] avg loss 0.000357573, throughput 5.97799K wps
[Epoch 14 Batch 90/173] avg loss 0.000336358, throughput 5.98602K wps
[Epoch 14 Batch 120/173] avg loss 0.000364011, throughput 5.99763K wps
[Epoch 14 Batch 150/173] avg loss 0.000308704, throughput 5.98238K wps
Begin Testing...
[Epoch 14] train avg loss 0.000347225, test acc 0.7271, test avg loss 0.86796, throughput 6.01554K wps
[Epoch 15 Batch 30/173] avg loss 0.000282861, throughput 6.14989K wps
[Epoch 15 Batch 60/173] avg loss 0.000245637, throughput 5.99672K wps
[Epoch 15 Batch 90/173] avg loss 0.000310277, throughput 5.98419K wps
[Epoch 15 Batch 120/173] avg loss 0.000289963, throughput 5.98847K wps
[Epoch 15 Batch 150/173] avg loss 0.000251699, throughput 5.97889K wps
Begin Testing...
[Epoch 15] train avg loss 0.000272526, test acc 0.7281, test avg loss 0.905812, throughput 6.01617K wps
[Epoch 16 Batch 30/173] avg loss 0.000245069, throughput 6.1388K wps
[Epoch 16 Batch 60/173] avg loss 0.000181111, throughput 6.0006K wps
[Epoch 16 Batch 90/173] avg loss 0.00018274, throughput 5.9832K wps
[Epoch 16 Batch 120/173] avg loss 0.000222955, throughput 5.98495K wps
[Epoch 16 Batch 150/173] avg loss 0.000249298, throughput 6.00178K wps
Begin Testing...
[Epoch 16] train avg loss 0.000217999, test acc 0.7219, test avg loss 0.94571, throughput 6.01758K wps
[Epoch 17 Batch 30/173] avg loss 0.000199997, throughput 6.13505K wps
[Epoch 17 Batch 60/173] avg loss 0.000156833, throughput 5.98498K wps
[Epoch 17 Batch 90/173] avg loss 0.000157351, throughput 5.98288K wps
[Epoch 17 Batch 120/173] avg loss 0.000165751, throughput 5.98803K wps
[Epoch 17 Batch 150/173] avg loss 0.000215449, throughput 5.98751K wps
Begin Testing...
[Epoch 17] train avg loss 0.000174905, test acc 0.7208, test avg loss 0.979591, throughput 6.0128K wps
[Epoch 18 Batch 30/173] avg loss 0.000140112, throughput 6.1288K wps
[Epoch 18 Batch 60/173] avg loss 0.000128559, throughput 5.98438K wps
[Epoch 18 Batch 90/173] avg loss 0.000134725, throughput 5.98854K wps
[Epoch 18 Batch 120/173] avg loss 0.000183399, throughput 5.96985K wps
[Epoch 18 Batch 150/173] avg loss 0.000164627, throughput 5.97658K wps
Begin Testing...
[Epoch 18] train avg loss 0.000149337, test acc 0.7198, test avg loss 1.01996, throughput 6.0067K wps
[Epoch 19 Batch 30/173] avg loss 0.000106895, throughput 6.13261K wps
[Epoch 19 Batch 60/173] avg loss 0.000115767, throughput 5.97932K wps
[Epoch 19 Batch 90/173] avg loss 0.000148665, throughput 5.98575K wps
[Epoch 19 Batch 120/173] avg loss 0.000117304, throughput 5.98444K wps
[Epoch 19 Batch 150/173] avg loss 0.000117333, throughput 5.98568K wps
Begin Testing...
[Epoch 19] train avg loss 0.000122448, test acc 0.7177, test avg loss 1.05352, throughput 6.0103K wps
[Epoch 20 Batch 30/173] avg loss 9.96277e-05, throughput 6.1237K wps
[Epoch 20 Batch 60/173] avg loss 9.61077e-05, throughput 5.98071K wps
[Epoch 20 Batch 90/173] avg loss 0.000114775, throughput 5.98335K wps
[Epoch 20 Batch 120/173] avg loss 0.00012564, throughput 5.99603K wps
[Epoch 20 Batch 150/173] avg loss 9.55735e-05, throughput 5.98712K wps
Begin Testing...
[Epoch 20] train avg loss 0.000105127, test acc 0.7135, test avg loss 1.08265, throughput 6.01077K wps
[Epoch 21 Batch 30/173] avg loss 7.74951e-05, throughput 6.12973K wps
[Epoch 21 Batch 60/173] avg loss 7.23867e-05, throughput 5.98996K wps
[Epoch 21 Batch 90/173] avg loss 9.53474e-05, throughput 5.99775K wps
[Epoch 21 Batch 120/173] avg loss 8.21856e-05, throughput 5.97978K wps
[Epoch 21 Batch 150/173] avg loss 8.17259e-05, throughput 5.99404K wps
Begin Testing...
[Epoch 21] train avg loss 8.50889e-05, test acc 0.7115, test avg loss 1.11008, throughput 6.0152K wps
[Epoch 22 Batch 30/173] avg loss 6.80358e-05, throughput 6.13431K wps
[Epoch 22 Batch 60/173] avg loss 8.18884e-05, throughput 5.99326K wps
[Epoch 22 Batch 90/173] avg loss 7.03804e-05, throughput 6.00633K wps
[Epoch 22 Batch 120/173] avg loss 8.5094e-05, throughput 5.98877K wps
[Epoch 22 Batch 150/173] avg loss 8.41785e-05, throughput 5.98676K wps
Begin Testing...
[Epoch 22] train avg loss 7.65247e-05, test acc 0.7115, test avg loss 1.14557, throughput 6.01762K wps
[Epoch 23 Batch 30/173] avg loss 5.67503e-05, throughput 6.14581K wps
[Epoch 23 Batch 60/173] avg loss 6.51665e-05, throughput 5.99164K wps
[Epoch 23 Batch 90/173] avg loss 6.01543e-05, throughput 5.99239K wps
[Epoch 23 Batch 120/173] avg loss 6.3436e-05, throughput 5.98549K wps
[Epoch 23 Batch 150/173] avg loss 7.10662e-05, throughput 5.99307K wps
Begin Testing...
[Epoch 23] train avg loss 6.33539e-05, test acc 0.7073, test avg loss 1.17034, throughput 6.0188K wps
[Epoch 24 Batch 30/173] avg loss 5.12065e-05, throughput 6.1421K wps
[Epoch 24 Batch 60/173] avg loss 5.80633e-05, throughput 5.99329K wps
[Epoch 24 Batch 90/173] avg loss 5.45484e-05, throughput 6.00006K wps
[Epoch 24 Batch 120/173] avg loss 4.06037e-05, throughput 5.98236K wps
[Epoch 24 Batch 150/173] avg loss 5.5051e-05, throughput 5.99857K wps
Begin Testing...
[Epoch 24] train avg loss 5.36773e-05, test acc 0.7083, test avg loss 1.19436, throughput 6.02069K wps
[Epoch 25 Batch 30/173] avg loss 4.3029e-05, throughput 6.13889K wps
[Epoch 25 Batch 60/173] avg loss 4.59243e-05, throughput 6.00027K wps
[Epoch 25 Batch 90/173] avg loss 3.86022e-05, throughput 5.9888K wps
[Epoch 25 Batch 120/173] avg loss 6.25133e-05, throughput 5.98628K wps
[Epoch 25 Batch 150/173] avg loss 6.28011e-05, throughput 6.0032K wps
Begin Testing...
[Epoch 25] train avg loss 5.02749e-05, test acc 0.7052, test avg loss 1.22928, throughput 6.01916K wps
[Epoch 26 Batch 30/173] avg loss 3.36861e-05, throughput 6.12834K wps
[Epoch 26 Batch 60/173] avg loss 6.10095e-05, throughput 5.99132K wps
[Epoch 26 Batch 90/173] avg loss 4.22633e-05, throughput 5.98568K wps
[Epoch 26 Batch 120/173] avg loss 4.23461e-05, throughput 5.98951K wps
[Epoch 26 Batch 150/173] avg loss 3.93911e-05, throughput 5.97378K wps
Begin Testing...
[Epoch 26] train avg loss 4.32277e-05, test acc 0.7094, test avg loss 1.25042, throughput 6.01053K wps
[Epoch 27 Batch 30/173] avg loss 3.36572e-05, throughput 6.09905K wps
[Epoch 27 Batch 60/173] avg loss 3.79812e-05, throughput 5.95211K wps
[Epoch 27 Batch 90/173] avg loss 3.45062e-05, throughput 5.99762K wps
[Epoch 27 Batch 120/173] avg loss 4.61716e-05, throughput 5.98407K wps
[Epoch 27 Batch 150/173] avg loss 5.17502e-05, throughput 5.98773K wps
Begin Testing...
[Epoch 27] train avg loss 3.99783e-05, test acc 0.7063, test avg loss 1.27586, throughput 6.00337K wps
[Epoch 28 Batch 30/173] avg loss 3.03751e-05, throughput 6.13094K wps
[Epoch 28 Batch 60/173] avg loss 3.22089e-05, throughput 5.97865K wps
[Epoch 28 Batch 90/173] avg loss 3.04886e-05, throughput 5.98816K wps
[Epoch 28 Batch 120/173] avg loss 2.78199e-05, throughput 5.98079K wps
[Epoch 28 Batch 150/173] avg loss 3.25845e-05, throughput 5.98726K wps
Begin Testing...
[Epoch 28] train avg loss 3.10381e-05, test acc 0.7042, test avg loss 1.29971, throughput 6.01045K wps
[Epoch 29 Batch 30/173] avg loss 2.50633e-05, throughput 6.13461K wps
[Epoch 29 Batch 60/173] avg loss 2.60005e-05, throughput 5.98538K wps
[Epoch 29 Batch 90/173] avg loss 3.81802e-05, throughput 6.01421K wps
[Epoch 29 Batch 120/173] avg loss 2.15347e-05, throughput 5.98946K wps
[Epoch 29 Batch 150/173] avg loss 2.13837e-05, throughput 5.97434K wps
Begin Testing...
[Epoch 29] train avg loss 2.72337e-05, test acc 0.7083, test avg loss 1.32314, throughput 6.01514K wps
[Epoch 30 Batch 30/173] avg loss 2.34498e-05, throughput 6.12824K wps
[Epoch 30 Batch 60/173] avg loss 2.70758e-05, throughput 5.98414K wps
[Epoch 30 Batch 90/173] avg loss 2.28641e-05, throughput 5.98561K wps
[Epoch 30 Batch 120/173] avg loss 1.90581e-05, throughput 5.98247K wps
[Epoch 30 Batch 150/173] avg loss 2.03275e-05, throughput 5.9804K wps
Begin Testing...
[Epoch 30] train avg loss 2.57939e-05, test acc 0.7094, test avg loss 1.34953, throughput 6.00789K wps
[Epoch 31 Batch 30/173] avg loss 1.6646e-05, throughput 6.13792K wps
[Epoch 31 Batch 60/173] avg loss 2.48269e-05, throughput 5.98552K wps
[Epoch 31 Batch 90/173] avg loss 1.40808e-05, throughput 5.97855K wps
[Epoch 31 Batch 120/173] avg loss 3.51252e-05, throughput 5.98722K wps
[Epoch 31 Batch 150/173] avg loss 2.11309e-05, throughput 5.9876K wps
Begin Testing...
[Epoch 31] train avg loss 2.14553e-05, test acc 0.7083, test avg loss 1.3793, throughput 6.01147K wps
[Epoch 32 Batch 30/173] avg loss 1.5165e-05, throughput 6.1452K wps
[Epoch 32 Batch 60/173] avg loss 1.72342e-05, throughput 5.99417K wps
[Epoch 32 Batch 90/173] avg loss 2.31266e-05, throughput 5.99504K wps
[Epoch 32 Batch 120/173] avg loss 1.3141e-05, throughput 5.98755K wps
[Epoch 32 Batch 150/173] avg loss 2.91853e-05, throughput 5.97652K wps
Begin Testing...
[Epoch 32] train avg loss 2.04782e-05, test acc 0.7083, test avg loss 1.40204, throughput 6.01852K wps
[Epoch 33 Batch 30/173] avg loss 2.11336e-05, throughput 6.14794K wps
[Epoch 33 Batch 60/173] avg loss 1.31545e-05, throughput 5.99088K wps
[Epoch 33 Batch 90/173] avg loss 2.56666e-05, throughput 5.99024K wps
[Epoch 33 Batch 120/173] avg loss 1.75775e-05, throughput 5.99842K wps
[Epoch 33 Batch 150/173] avg loss 1.38251e-05, throughput 5.99206K wps
Begin Testing...
[Epoch 33] train avg loss 1.76047e-05, test acc 0.7021, test avg loss 1.42706, throughput 6.02057K wps
[Epoch 34 Batch 30/173] avg loss 1.33252e-05, throughput 6.1462K wps
[Epoch 34 Batch 60/173] avg loss 1.43453e-05, throughput 5.99476K wps
[Epoch 34 Batch 90/173] avg loss 2.28495e-05, throughput 5.99144K wps
[Epoch 34 Batch 120/173] avg loss 1.04703e-05, throughput 6.00328K wps
[Epoch 34 Batch 150/173] avg loss 2.28933e-05, throughput 5.99005K wps
Begin Testing...
[Epoch 34] train avg loss 1.63191e-05, test acc 0.7031, test avg loss 1.44248, throughput 6.02158K wps
[Epoch 35 Batch 30/173] avg loss 1.06632e-05, throughput 6.13085K wps
[Epoch 35 Batch 60/173] avg loss 1.10993e-05, throughput 5.99803K wps
[Epoch 35 Batch 90/173] avg loss 1.41719e-05, throughput 5.99484K wps
[Epoch 35 Batch 120/173] avg loss 1.93859e-05, throughput 5.99134K wps
[Epoch 35 Batch 150/173] avg loss 2.24627e-05, throughput 5.99476K wps
Begin Testing...
[Epoch 35] train avg loss 1.5037e-05, test acc 0.7073, test avg loss 1.46978, throughput 6.01917K wps
[Epoch 36 Batch 30/173] avg loss 8.90646e-06, throughput 6.14759K wps
[Epoch 36 Batch 60/173] avg loss 1.01439e-05, throughput 5.99527K wps
[Epoch 36 Batch 90/173] avg loss 1.03329e-05, throughput 5.99051K wps
[Epoch 36 Batch 120/173] avg loss 1.59532e-05, throughput 5.9985K wps
[Epoch 36 Batch 150/173] avg loss 1.06893e-05, throughput 5.97885K wps
Begin Testing...
[Epoch 36] train avg loss 1.3034e-05, test acc 0.7073, test avg loss 1.49051, throughput 6.01909K wps
[Epoch 37 Batch 30/173] avg loss 1.13834e-05, throughput 6.13798K wps
[Epoch 37 Batch 60/173] avg loss 1.35139e-05, throughput 5.98019K wps
[Epoch 37 Batch 90/173] avg loss 9.90722e-06, throughput 5.98875K wps
[Epoch 37 Batch 120/173] avg loss 1.80749e-05, throughput 5.99021K wps
[Epoch 37 Batch 150/173] avg loss 1.80392e-05, throughput 5.99389K wps
Begin Testing...
[Epoch 37] train avg loss 1.35942e-05, test acc 0.7083, test avg loss 1.51053, throughput 6.01533K wps
[Epoch 38 Batch 30/173] avg loss 7.71877e-06, throughput 6.14522K wps
[Epoch 38 Batch 60/173] avg loss 1.52379e-05, throughput 5.98536K wps
[Epoch 38 Batch 90/173] avg loss 7.38574e-06, throughput 5.98119K wps
[Epoch 38 Batch 120/173] avg loss 8.46542e-06, throughput 5.97945K wps
[Epoch 38 Batch 150/173] avg loss 9.19654e-06, throughput 5.98607K wps
Begin Testing...
[Epoch 38] train avg loss 1.12245e-05, test acc 0.7042, test avg loss 1.526, throughput 6.01056K wps
[Epoch 39 Batch 30/173] avg loss 6.28701e-06, throughput 6.13352K wps
[Epoch 39 Batch 60/173] avg loss 1.35286e-05, throughput 5.9859K wps
[Epoch 39 Batch 90/173] avg loss 6.83627e-06, throughput 5.98355K wps
[Epoch 39 Batch 120/173] avg loss 8.3789e-06, throughput 5.98627K wps
[Epoch 39 Batch 150/173] avg loss 1.63894e-05, throughput 5.985K wps
Begin Testing...
[Epoch 39] train avg loss 9.82899e-06, test acc 0.7031, test avg loss 1.54577, throughput 6.0116K wps
[Epoch 40 Batch 30/173] avg loss 7.59058e-06, throughput 6.11512K wps
[Epoch 40 Batch 60/173] avg loss 1.32104e-05, throughput 5.97918K wps
[Epoch 40 Batch 90/173] avg loss 4.92395e-06, throughput 5.98174K wps
[Epoch 40 Batch 120/173] avg loss 1.63782e-05, throughput 5.98176K wps
[Epoch 40 Batch 150/173] avg loss 7.47851e-06, throughput 5.97785K wps
Begin Testing...
[Epoch 40] train avg loss 9.87809e-06, test acc 0.7063, test avg loss 1.56127, throughput 6.00382K wps
[Epoch 41 Batch 30/173] avg loss 5.67379e-06, throughput 6.12314K wps
[Epoch 41 Batch 60/173] avg loss 6.45633e-06, throughput 5.98744K wps
[Epoch 41 Batch 90/173] avg loss 9.75208e-06, throughput 5.98562K wps
[Epoch 41 Batch 120/173] avg loss 7.49946e-06, throughput 5.97594K wps
[Epoch 41 Batch 150/173] avg loss 1.45634e-05, throughput 5.98157K wps
Begin Testing...
[Epoch 41] train avg loss 8.79629e-06, test acc 0.7073, test avg loss 1.58261, throughput 6.00699K wps
[Epoch 42 Batch 30/173] avg loss 4.11768e-06, throughput 6.13114K wps
[Epoch 42 Batch 60/173] avg loss 1.47226e-05, throughput 5.98257K wps
[Epoch 42 Batch 90/173] avg loss 5.0781e-06, throughput 5.97979K wps
[Epoch 42 Batch 120/173] avg loss 5.29281e-06, throughput 5.9832K wps
[Epoch 42 Batch 150/173] avg loss 5.65373e-06, throughput 5.98926K wps
Begin Testing...
[Epoch 42] train avg loss 6.9179e-06, test acc 0.7083, test avg loss 1.60431, throughput 6.00907K wps
[Epoch 43 Batch 30/173] avg loss 4.31898e-06, throughput 6.13038K wps
[Epoch 43 Batch 60/173] avg loss 4.82419e-06, throughput 5.98281K wps
[Epoch 43 Batch 90/173] avg loss 4.24711e-06, throughput 5.994K wps
[Epoch 43 Batch 120/173] avg loss 8.57614e-06, throughput 5.98469K wps
[Epoch 43 Batch 150/173] avg loss 1.31318e-05, throughput 5.99892K wps
Begin Testing...
[Epoch 43] train avg loss 6.61241e-06, test acc 0.7104, test avg loss 1.62474, throughput 6.01545K wps
[Epoch 44 Batch 30/173] avg loss 4.66937e-06, throughput 6.12042K wps
[Epoch 44 Batch 60/173] avg loss 6.73033e-06, throughput 5.97973K wps
[Epoch 44 Batch 90/173] avg loss 4.9816e-06, throughput 5.9911K wps
[Epoch 44 Batch 120/173] avg loss 5.16173e-06, throughput 5.99648K wps
[Epoch 44 Batch 150/173] avg loss 5.57143e-06, throughput 6.00287K wps
Begin Testing...
[Epoch 44] train avg loss 6.61697e-06, test acc 0.7146, test avg loss 1.65254, throughput 6.01538K wps
[Epoch 45 Batch 30/173] avg loss 3.81326e-06, throughput 6.13699K wps
[Epoch 45 Batch 60/173] avg loss 3.46173e-06, throughput 5.98196K wps
[Epoch 45 Batch 90/173] avg loss 3.43324e-06, throughput 5.98345K wps
[Epoch 45 Batch 120/173] avg loss 4.75545e-06, throughput 5.98076K wps
[Epoch 45 Batch 150/173] avg loss 3.85545e-06, throughput 5.98428K wps
Begin Testing...
[Epoch 45] train avg loss 5.50219e-06, test acc 0.7125, test avg loss 1.68146, throughput 6.01017K wps
[Epoch 46 Batch 30/173] avg loss 1.18726e-05, throughput 6.12349K wps
[Epoch 46 Batch 60/173] avg loss 2.54375e-06, throughput 5.98327K wps
[Epoch 46 Batch 90/173] avg loss 4.78538e-06, throughput 5.98032K wps
[Epoch 46 Batch 120/173] avg loss 4.72328e-06, throughput 5.98597K wps
[Epoch 46 Batch 150/173] avg loss 3.05994e-06, throughput 5.9845K wps
Begin Testing...
[Epoch 46] train avg loss 5.20747e-06, test acc 0.7052, test avg loss 1.70648, throughput 6.00643K wps
[Epoch 47 Batch 30/173] avg loss 3.72583e-06, throughput 6.13176K wps
[Epoch 47 Batch 60/173] avg loss 3.90052e-06, throughput 5.98796K wps
[Epoch 47 Batch 90/173] avg loss 3.03202e-06, throughput 5.9806K wps
[Epoch 47 Batch 120/173] avg loss 1.01903e-05, throughput 5.96874K wps
[Epoch 47 Batch 150/173] avg loss 3.81049e-06, throughput 5.98131K wps
Begin Testing...
[Epoch 47] train avg loss 4.86127e-06, test acc 0.7083, test avg loss 1.7181, throughput 6.00856K wps
[Epoch 48 Batch 30/173] avg loss 3.29104e-06, throughput 6.13742K wps
[Epoch 48 Batch 60/173] avg loss 2.33901e-06, throughput 5.98588K wps
[Epoch 48 Batch 90/173] avg loss 2.78702e-06, throughput 5.98459K wps
[Epoch 48 Batch 120/173] avg loss 2.90768e-06, throughput 5.98627K wps
[Epoch 48 Batch 150/173] avg loss 2.2526e-06, throughput 5.96734K wps
Begin Testing...
[Epoch 48] train avg loss 3.94466e-06, test acc 0.7073, test avg loss 1.73308, throughput 6.00803K wps
[Epoch 49 Batch 30/173] avg loss 1.05101e-05, throughput 6.12512K wps
[Epoch 49 Batch 60/173] avg loss 2.58134e-06, throughput 5.96896K wps
[Epoch 49 Batch 90/173] avg loss 3.87021e-06, throughput 5.97054K wps
[Epoch 49 Batch 120/173] avg loss 2.64492e-06, throughput 5.99257K wps
[Epoch 49 Batch 150/173] avg loss 2.68827e-06, throughput 5.98802K wps
Begin Testing...
[Epoch 49] train avg loss 4.30279e-06, test acc 0.7115, test avg loss 1.74895, throughput 6.00472K wps
[Epoch 50 Batch 30/173] avg loss 2.04447e-06, throughput 6.12857K wps
[Epoch 50 Batch 60/173] avg loss 9.46491e-06, throughput 5.9759K wps
[Epoch 50 Batch 90/173] avg loss 2.10653e-06, throughput 5.98538K wps
[Epoch 50 Batch 120/173] avg loss 3.38632e-06, throughput 5.98009K wps
[Epoch 50 Batch 150/173] avg loss 1.85839e-06, throughput 5.98343K wps
Begin Testing...
[Epoch 50] train avg loss 3.66172e-06, test acc 0.7104, test avg loss 1.76614, throughput 6.00842K wps
[Epoch 51 Batch 30/173] avg loss 1.79829e-06, throughput 6.12245K wps
[Epoch 51 Batch 60/173] avg loss 2.06375e-06, throughput 5.98101K wps
[Epoch 51 Batch 90/173] avg loss 1.00639e-05, throughput 5.9926K wps
[Epoch 51 Batch 120/173] avg loss 3.21841e-06, throughput 5.97649K wps
[Epoch 51 Batch 150/173] avg loss 3.24499e-06, throughput 5.97557K wps
Begin Testing...
[Epoch 51] train avg loss 4.00558e-06, test acc 0.7115, test avg loss 1.79048, throughput 6.00457K wps
[Epoch 52 Batch 30/173] avg loss 9.4539e-06, throughput 6.13122K wps
[Epoch 52 Batch 60/173] avg loss 1.60193e-06, throughput 5.98227K wps
[Epoch 52 Batch 90/173] avg loss 1.97185e-06, throughput 5.97813K wps
[Epoch 52 Batch 120/173] avg loss 1.73577e-06, throughput 5.99015K wps
[Epoch 52 Batch 150/173] avg loss 2.91008e-06, throughput 5.98199K wps
Begin Testing...
[Epoch 52] train avg loss 3.28672e-06, test acc 0.7094, test avg loss 1.80681, throughput 6.00991K wps
[Epoch 53 Batch 30/173] avg loss 1.55346e-05, throughput 6.14036K wps
[Epoch 53 Batch 60/173] avg loss 9.1518e-06, throughput 5.975K wps
[Epoch 53 Batch 90/173] avg loss 3.82502e-06, throughput 5.98238K wps
[Epoch 53 Batch 120/173] avg loss 2.98409e-06, throughput 5.98298K wps
[Epoch 53 Batch 150/173] avg loss 1.84752e-06, throughput 5.98272K wps
Begin Testing...
[Epoch 53] train avg loss 6.19617e-06, test acc 0.7063, test avg loss 1.84314, throughput 6.00888K wps
[Epoch 54 Batch 30/173] avg loss 1.91639e-06, throughput 6.12587K wps
[Epoch 54 Batch 60/173] avg loss 3.50454e-06, throughput 5.96825K wps
[Epoch 54 Batch 90/173] avg loss 1.85783e-06, throughput 5.97776K wps
[Epoch 54 Batch 120/173] avg loss 9.5514e-06, throughput 6.00333K wps
[Epoch 54 Batch 150/173] avg loss 1.99105e-06, throughput 5.98814K wps
Begin Testing...
[Epoch 54] train avg loss 3.62798e-06, test acc 0.7104, test avg loss 1.86374, throughput 6.00925K wps
[Epoch 55 Batch 30/173] avg loss 1.67316e-06, throughput 6.13828K wps
[Epoch 55 Batch 60/173] avg loss 8.64564e-06, throughput 5.98873K wps
[Epoch 55 Batch 90/173] avg loss 1.5069e-06, throughput 5.98959K wps
[Epoch 55 Batch 120/173] avg loss 2.9843e-06, throughput 5.98833K wps
[Epoch 55 Batch 150/173] avg loss 1.68559e-06, throughput 5.9804K wps
Begin Testing...
[Epoch 55] train avg loss 3.0837e-06, test acc 0.7010, test avg loss 1.88534, throughput 6.01511K wps
[Epoch 56 Batch 30/173] avg loss 2.11902e-06, throughput 6.13399K wps
[Epoch 56 Batch 60/173] avg loss 1.29473e-06, throughput 5.97733K wps
[Epoch 56 Batch 90/173] avg loss 1.5155e-06, throughput 5.99063K wps
[Epoch 56 Batch 120/173] avg loss 7.98634e-06, throughput 6.0016K wps
[Epoch 56 Batch 150/173] avg loss 9.49721e-07, throughput 5.98919K wps
Begin Testing...
[Epoch 56] train avg loss 2.54252e-06, test acc 0.7031, test avg loss 1.89973, throughput 6.01677K wps
[Epoch 57 Batch 30/173] avg loss 1.70678e-06, throughput 6.13875K wps
[Epoch 57 Batch 60/173] avg loss 1.45058e-06, throughput 5.96448K wps
[Epoch 57 Batch 90/173] avg loss 1.13017e-06, throughput 5.95822K wps
[Epoch 57 Batch 120/173] avg loss 7.01296e-06, throughput 5.98216K wps
[Epoch 57 Batch 150/173] avg loss 1.9151e-06, throughput 5.98868K wps
Begin Testing...
[Epoch 57] train avg loss 2.48682e-06, test acc 0.7083, test avg loss 1.91963, throughput 6.00371K wps
[Epoch 58 Batch 30/173] avg loss 1.34848e-06, throughput 6.12522K wps
[Epoch 58 Batch 60/173] avg loss 1.16423e-06, throughput 5.98128K wps
[Epoch 58 Batch 90/173] avg loss 1.23272e-06, throughput 5.98078K wps
[Epoch 58 Batch 120/173] avg loss 7.4452e-06, throughput 5.98635K wps
[Epoch 58 Batch 150/173] avg loss 1.19156e-06, throughput 5.985K wps
Begin Testing...
[Epoch 58] train avg loss 2.29072e-06, test acc 0.7063, test avg loss 1.9269, throughput 6.00859K wps
[Epoch 59 Batch 30/173] avg loss 1.53394e-06, throughput 6.1443K wps
[Epoch 59 Batch 60/173] avg loss 9.47942e-07, throughput 5.98269K wps
[Epoch 59 Batch 90/173] avg loss 9.73121e-07, throughput 5.99023K wps
[Epoch 59 Batch 120/173] avg loss 1.12436e-06, throughput 5.98766K wps
[Epoch 59 Batch 150/173] avg loss 6.49624e-06, throughput 5.98638K wps
Begin Testing...
[Epoch 59] train avg loss 2.05168e-06, test acc 0.7083, test avg loss 1.94126, throughput 6.01409K wps
Test loss 0.477378, test acc 0.7814
Total time cost 357.92s
9596 1066
[Epoch 0 Batch 30/173] avg loss 0.0138521, throughput 5.76978K wps
[Epoch 0 Batch 60/173] avg loss 0.0138282, throughput 5.98846K wps
[Epoch 0 Batch 90/173] avg loss 0.0138227, throughput 5.984K wps
[Epoch 0 Batch 120/173] avg loss 0.0137911, throughput 5.97968K wps
[Epoch 0 Batch 150/173] avg loss 0.0137698, throughput 5.99109K wps
Begin Testing...
[Epoch 0] train avg loss 0.0138253, test acc 0.5906, test avg loss 0.685265, throughput 5.94718K wps
Observed Improvement.
Begin Testing...
[Epoch 1 Batch 30/173] avg loss 0.0135142, throughput 6.1352K wps
[Epoch 1 Batch 60/173] avg loss 0.0134393, throughput 5.96927K wps
[Epoch 1 Batch 90/173] avg loss 0.0134582, throughput 5.98266K wps
[Epoch 1 Batch 120/173] avg loss 0.0134025, throughput 5.98312K wps
[Epoch 1 Batch 150/173] avg loss 0.0133302, throughput 5.98825K wps
Begin Testing...
[Epoch 1] train avg loss 0.0134329, test acc 0.7052, test avg loss 0.667851, throughput 6.01177K wps
Observed Improvement.
Begin Testing...
[Epoch 2 Batch 30/173] avg loss 0.012893, throughput 6.14974K wps
[Epoch 2 Batch 60/173] avg loss 0.0128063, throughput 5.99851K wps
[Epoch 2 Batch 90/173] avg loss 0.0126917, throughput 5.98599K wps
[Epoch 2 Batch 120/173] avg loss 0.0125421, throughput 5.98627K wps
[Epoch 2 Batch 150/173] avg loss 0.0123109, throughput 5.9848K wps
Begin Testing...
[Epoch 2] train avg loss 0.0126136, test acc 0.7219, test avg loss 0.620452, throughput 6.01502K wps
Observed Improvement.
Begin Testing...
[Epoch 3 Batch 30/173] avg loss 0.0115148, throughput 6.14034K wps
[Epoch 3 Batch 60/173] avg loss 0.0112418, throughput 5.98799K wps
[Epoch 3 Batch 90/173] avg loss 0.0108783, throughput 5.97824K wps
[Epoch 3 Batch 120/173] avg loss 0.0106575, throughput 5.98935K wps
[Epoch 3 Batch 150/173] avg loss 0.0103834, throughput 5.98748K wps
Begin Testing...
[Epoch 3] train avg loss 0.0108543, test acc 0.7542, test avg loss 0.540875, throughput 6.01362K wps
Observed Improvement.
Begin Testing...
[Epoch 4 Batch 30/173] avg loss 0.00897909, throughput 6.14987K wps
[Epoch 4 Batch 60/173] avg loss 0.00878545, throughput 5.98352K wps
[Epoch 4 Batch 90/173] avg loss 0.00811725, throughput 5.97814K wps
[Epoch 4 Batch 120/173] avg loss 0.00797468, throughput 5.97805K wps
[Epoch 4 Batch 150/173] avg loss 0.00793587, throughput 5.97532K wps
Begin Testing...
[Epoch 4] train avg loss 0.00828901, test acc 0.7823, test avg loss 0.471902, throughput 6.00902K wps
Observed Improvement.
Begin Testing...
[Epoch 5 Batch 30/173] avg loss 0.0061483, throughput 6.13731K wps
[Epoch 5 Batch 60/173] avg loss 0.00591934, throughput 6.00145K wps
[Epoch 5 Batch 90/173] avg loss 0.00596318, throughput 5.9937K wps
[Epoch 5 Batch 120/173] avg loss 0.00587453, throughput 6.00409K wps
[Epoch 5 Batch 150/173] avg loss 0.00553713, throughput 5.98513K wps
Begin Testing...
[Epoch 5] train avg loss 0.00588212, test acc 0.7802, test avg loss 0.453871, throughput 6.02087K wps
[Epoch 6 Batch 30/173] avg loss 0.00425994, throughput 6.13494K wps
[Epoch 6 Batch 60/173] avg loss 0.00398068, throughput 5.99873K wps
[Epoch 6 Batch 90/173] avg loss 0.00433606, throughput 5.98146K wps
[Epoch 6 Batch 120/173] avg loss 0.00388574, throughput 5.97811K wps
[Epoch 6 Batch 150/173] avg loss 0.00431717, throughput 5.9929K wps
Begin Testing...
[Epoch 6] train avg loss 0.00411263, test acc 0.7875, test avg loss 0.460471, throughput 6.01535K wps
Observed Improvement.
Begin Testing...
[Epoch 7 Batch 30/173] avg loss 0.00289052, throughput 6.12899K wps
[Epoch 7 Batch 60/173] avg loss 0.00262437, throughput 5.99669K wps
[Epoch 7 Batch 90/173] avg loss 0.00283834, throughput 5.98913K wps
[Epoch 7 Batch 120/173] avg loss 0.00280882, throughput 5.9956K wps
[Epoch 7 Batch 150/173] avg loss 0.00308369, throughput 5.98804K wps
Begin Testing...
[Epoch 7] train avg loss 0.00284077, test acc 0.7844, test avg loss 0.489485, throughput 6.01566K wps
[Epoch 8 Batch 30/173] avg loss 0.00204631, throughput 6.12592K wps
[Epoch 8 Batch 60/173] avg loss 0.00193785, throughput 5.98195K wps
[Epoch 8 Batch 90/173] avg loss 0.00186126, throughput 5.98143K wps
[Epoch 8 Batch 120/173] avg loss 0.00195122, throughput 5.99472K wps
[Epoch 8 Batch 150/173] avg loss 0.0019672, throughput 5.97967K wps
Begin Testing...
[Epoch 8] train avg loss 0.00198618, test acc 0.7750, test avg loss 0.530266, throughput 6.00754K wps
[Epoch 9 Batch 30/173] avg loss 0.00132333, throughput 6.13459K wps
[Epoch 9 Batch 60/173] avg loss 0.00140631, throughput 5.98958K wps
[Epoch 9 Batch 90/173] avg loss 0.00138181, throughput 5.99528K wps
[Epoch 9 Batch 120/173] avg loss 0.00139743, throughput 5.97967K wps
[Epoch 9 Batch 150/173] avg loss 0.00139284, throughput 5.99517K wps
Begin Testing...
[Epoch 9] train avg loss 0.00139272, test acc 0.7646, test avg loss 0.560898, throughput 6.0137K wps
[Epoch 10 Batch 30/173] avg loss 0.000855525, throughput 6.13169K wps
[Epoch 10 Batch 60/173] avg loss 0.00107081, throughput 5.98157K wps
[Epoch 10 Batch 90/173] avg loss 0.00101731, throughput 5.9738K wps
[Epoch 10 Batch 120/173] avg loss 0.00109201, throughput 5.97578K wps
[Epoch 10 Batch 150/173] avg loss 0.00103084, throughput 5.97936K wps
Begin Testing...
[Epoch 10] train avg loss 0.00102437, test acc 0.7625, test avg loss 0.590957, throughput 6.00792K wps
[Epoch 11 Batch 30/173] avg loss 0.00064386, throughput 6.12848K wps
[Epoch 11 Batch 60/173] avg loss 0.000784208, throughput 5.97848K wps
[Epoch 11 Batch 90/173] avg loss 0.000655113, throughput 5.98974K wps
[Epoch 11 Batch 120/173] avg loss 0.000619797, throughput 5.9823K wps
[Epoch 11 Batch 150/173] avg loss 0.000793801, throughput 5.98091K wps
Begin Testing...
[Epoch 11] train avg loss 0.000724083, test acc 0.7562, test avg loss 0.625685, throughput 6.00928K wps
[Epoch 12 Batch 30/173] avg loss 0.000534345, throughput 6.12531K wps
[Epoch 12 Batch 60/173] avg loss 0.000453992, throughput 5.97134K wps
[Epoch 12 Batch 90/173] avg loss 0.000536669, throughput 5.97618K wps
[Epoch 12 Batch 120/173] avg loss 0.000588809, throughput 5.97548K wps
[Epoch 12 Batch 150/173] avg loss 0.000669143, throughput 5.98208K wps
Begin Testing...
[Epoch 12] train avg loss 0.000553785, test acc 0.7542, test avg loss 0.658062, throughput 6.00321K wps
[Epoch 13 Batch 30/173] avg loss 0.000459955, throughput 6.13195K wps
[Epoch 13 Batch 60/173] avg loss 0.000417662, throughput 5.97835K wps
[Epoch 13 Batch 90/173] avg loss 0.000416936, throughput 5.9893K wps
[Epoch 13 Batch 120/173] avg loss 0.000435883, throughput 5.98459K wps
[Epoch 13 Batch 150/173] avg loss 0.000366363, throughput 5.97667K wps
Begin Testing...
[Epoch 13] train avg loss 0.000421374, test acc 0.7573, test avg loss 0.674669, throughput 6.00983K wps
[Epoch 14 Batch 30/173] avg loss 0.00029728, throughput 6.12099K wps
[Epoch 14 Batch 60/173] avg loss 0.000271373, throughput 5.97219K wps
[Epoch 14 Batch 90/173] avg loss 0.000317311, throughput 5.98683K wps
[Epoch 14 Batch 120/173] avg loss 0.000324557, throughput 5.98021K wps
[Epoch 14 Batch 150/173] avg loss 0.000331179, throughput 5.99205K wps
Begin Testing...
[Epoch 14] train avg loss 0.000336762, test acc 0.7625, test avg loss 0.698443, throughput 6.00846K wps
[Epoch 15 Batch 30/173] avg loss 0.000226717, throughput 6.12702K wps
[Epoch 15 Batch 60/173] avg loss 0.000307562, throughput 5.99059K wps
[Epoch 15 Batch 90/173] avg loss 0.000274966, throughput 5.99268K wps
[Epoch 15 Batch 120/173] avg loss 0.000304011, throughput 5.98037K wps
[Epoch 15 Batch 150/173] avg loss 0.000255189, throughput 5.97987K wps
Begin Testing...
[Epoch 15] train avg loss 0.000273555, test acc 0.7646, test avg loss 0.731322, throughput 6.01095K wps
[Epoch 16 Batch 30/173] avg loss 0.000209024, throughput 6.13967K wps
[Epoch 16 Batch 60/173] avg loss 0.000180538, throughput 5.98034K wps
[Epoch 16 Batch 90/173] avg loss 0.000198417, throughput 5.99461K wps
[Epoch 16 Batch 120/173] avg loss 0.000221998, throughput 5.9888K wps
[Epoch 16 Batch 150/173] avg loss 0.000229006, throughput 5.9939K wps
Begin Testing...
[Epoch 16] train avg loss 0.000215757, test acc 0.7635, test avg loss 0.757937, throughput 6.01567K wps
[Epoch 17 Batch 30/173] avg loss 0.0001563, throughput 6.14107K wps
[Epoch 17 Batch 60/173] avg loss 0.00014189, throughput 5.97658K wps
[Epoch 17 Batch 90/173] avg loss 0.000175436, throughput 5.99473K wps
[Epoch 17 Batch 120/173] avg loss 0.000174912, throughput 5.98703K wps
[Epoch 17 Batch 150/173] avg loss 0.000149378, throughput 5.97869K wps
Begin Testing...
[Epoch 17] train avg loss 0.000168847, test acc 0.7646, test avg loss 0.780364, throughput 6.01484K wps
[Epoch 18 Batch 30/173] avg loss 0.000141247, throughput 6.11362K wps
[Epoch 18 Batch 60/173] avg loss 9.90943e-05, throughput 5.9962K wps
[Epoch 18 Batch 90/173] avg loss 0.000160415, throughput 5.97768K wps
[Epoch 18 Batch 120/173] avg loss 0.000142411, throughput 5.99571K wps
[Epoch 18 Batch 150/173] avg loss 0.000146733, throughput 5.99005K wps
Begin Testing...
[Epoch 18] train avg loss 0.000138987, test acc 0.7688, test avg loss 0.800469, throughput 6.01235K wps
[Epoch 19 Batch 30/173] avg loss 0.000100085, throughput 6.1247K wps
[Epoch 19 Batch 60/173] avg loss 0.000128095, throughput 5.98444K wps
[Epoch 19 Batch 90/173] avg loss 0.00013299, throughput 5.98509K wps
[Epoch 19 Batch 120/173] avg loss 0.000101277, throughput 5.98242K wps
[Epoch 19 Batch 150/173] avg loss 0.000107229, throughput 5.98751K wps
Begin Testing...
[Epoch 19] train avg loss 0.000114945, test acc 0.7688, test avg loss 0.825846, throughput 6.00946K wps
[Epoch 20 Batch 30/173] avg loss 9.5949e-05, throughput 6.13198K wps
[Epoch 20 Batch 60/173] avg loss 8.17322e-05, throughput 5.99516K wps
[Epoch 20 Batch 90/173] avg loss 8.25559e-05, throughput 6.00127K wps
[Epoch 20 Batch 120/173] avg loss 8.97777e-05, throughput 5.99304K wps
[Epoch 20 Batch 150/173] avg loss 9.2097e-05, throughput 5.99587K wps
Begin Testing...
[Epoch 20] train avg loss 9.35539e-05, test acc 0.7677, test avg loss 0.836946, throughput 6.01958K wps
[Epoch 21 Batch 30/173] avg loss 7.46557e-05, throughput 6.13726K wps
[Epoch 21 Batch 60/173] avg loss 7.10517e-05, throughput 5.98231K wps
[Epoch 21 Batch 90/173] avg loss 7.95472e-05, throughput 5.98167K wps
[Epoch 21 Batch 120/173] avg loss 7.40267e-05, throughput 5.98785K wps
[Epoch 21 Batch 150/173] avg loss 9.82091e-05, throughput 5.99287K wps
Begin Testing...
[Epoch 21] train avg loss 7.95488e-05, test acc 0.7698, test avg loss 0.867808, throughput 6.01068K wps
[Epoch 22 Batch 30/173] avg loss 8.51819e-05, throughput 6.14092K wps
[Epoch 22 Batch 60/173] avg loss 5.59376e-05, throughput 5.9905K wps
[Epoch 22 Batch 90/173] avg loss 6.56493e-05, throughput 5.98715K wps
[Epoch 22 Batch 120/173] avg loss 6.71657e-05, throughput 5.98336K wps
[Epoch 22 Batch 150/173] avg loss 6.42572e-05, throughput 5.9901K wps
Begin Testing...
[Epoch 22] train avg loss 6.80841e-05, test acc 0.7677, test avg loss 0.885305, throughput 6.01484K wps
[Epoch 23 Batch 30/173] avg loss 4.66962e-05, throughput 6.13266K wps
[Epoch 23 Batch 60/173] avg loss 6.73149e-05, throughput 5.98737K wps
[Epoch 23 Batch 90/173] avg loss 6.82711e-05, throughput 5.9817K wps
[Epoch 23 Batch 120/173] avg loss 6.0561e-05, throughput 5.98737K wps
[Epoch 23 Batch 150/173] avg loss 6.03804e-05, throughput 5.97377K wps
Begin Testing...
[Epoch 23] train avg loss 6.12113e-05, test acc 0.7688, test avg loss 0.905006, throughput 6.00931K wps
[Epoch 24 Batch 30/173] avg loss 4.95059e-05, throughput 6.13607K wps
[Epoch 24 Batch 60/173] avg loss 4.47697e-05, throughput 5.9894K wps
[Epoch 24 Batch 90/173] avg loss 4.14702e-05, throughput 5.99009K wps
[Epoch 24 Batch 120/173] avg loss 3.8313e-05, throughput 5.98403K wps
[Epoch 24 Batch 150/173] avg loss 5.01296e-05, throughput 5.98964K wps
Begin Testing...
[Epoch 24] train avg loss 5.0465e-05, test acc 0.7677, test avg loss 0.92712, throughput 6.01256K wps
[Epoch 25 Batch 30/173] avg loss 3.91167e-05, throughput 6.13889K wps
[Epoch 25 Batch 60/173] avg loss 3.63573e-05, throughput 5.98712K wps
[Epoch 25 Batch 90/173] avg loss 4.11156e-05, throughput 5.98767K wps
[Epoch 25 Batch 120/173] avg loss 3.94698e-05, throughput 5.98516K wps
[Epoch 25 Batch 150/173] avg loss 5.37755e-05, throughput 5.98997K wps
Begin Testing...
[Epoch 25] train avg loss 4.18347e-05, test acc 0.7677, test avg loss 0.932815, throughput 6.01303K wps
[Epoch 26 Batch 30/173] avg loss 4.6224e-05, throughput 6.11812K wps
[Epoch 26 Batch 60/173] avg loss 3.32452e-05, throughput 5.97891K wps
[Epoch 26 Batch 90/173] avg loss 3.9961e-05, throughput 5.99297K wps
[Epoch 26 Batch 120/173] avg loss 3.15539e-05, throughput 5.97828K wps
[Epoch 26 Batch 150/173] avg loss 3.89384e-05, throughput 5.97739K wps
Begin Testing...
[Epoch 26] train avg loss 3.85776e-05, test acc 0.7625, test avg loss 0.953449, throughput 6.00801K wps
[Epoch 27 Batch 30/173] avg loss 2.83434e-05, throughput 6.1221K wps
[Epoch 27 Batch 60/173] avg loss 3.71571e-05, throughput 5.95368K wps
[Epoch 27 Batch 90/173] avg loss 2.85396e-05, throughput 5.95682K wps
[Epoch 27 Batch 120/173] avg loss 2.90676e-05, throughput 5.99232K wps
[Epoch 27 Batch 150/173] avg loss 4.61974e-05, throughput 5.97588K wps
Begin Testing...
[Epoch 27] train avg loss 3.35131e-05, test acc 0.7646, test avg loss 0.974889, throughput 5.99837K wps
[Epoch 28 Batch 30/173] avg loss 3.49548e-05, throughput 6.13715K wps
[Epoch 28 Batch 60/173] avg loss 2.62511e-05, throughput 5.98444K wps
[Epoch 28 Batch 90/173] avg loss 3.27766e-05, throughput 5.999K wps
[Epoch 28 Batch 120/173] avg loss 2.69438e-05, throughput 5.99933K wps
[Epoch 28 Batch 150/173] avg loss 2.81134e-05, throughput 5.9953K wps
Begin Testing...
[Epoch 28] train avg loss 2.88026e-05, test acc 0.7667, test avg loss 0.990822, throughput 6.0196K wps
[Epoch 29 Batch 30/173] avg loss 2.09698e-05, throughput 6.12086K wps
[Epoch 29 Batch 60/173] avg loss 2.59013e-05, throughput 5.98635K wps
[Epoch 29 Batch 90/173] avg loss 3.02052e-05, throughput 5.98103K wps
[Epoch 29 Batch 120/173] avg loss 2.64632e-05, throughput 6.00662K wps
[Epoch 29 Batch 150/173] avg loss 2.04906e-05, throughput 6.00356K wps
Begin Testing...
[Epoch 29] train avg loss 2.54185e-05, test acc 0.7677, test avg loss 1.01021, throughput 6.01657K wps
[Epoch 30 Batch 30/173] avg loss 3.14214e-05, throughput 6.13499K wps
[Epoch 30 Batch 60/173] avg loss 1.9255e-05, throughput 5.97138K wps
[Epoch 30 Batch 90/173] avg loss 1.81212e-05, throughput 5.97182K wps
[Epoch 30 Batch 120/173] avg loss 2.17179e-05, throughput 5.99424K wps
[Epoch 30 Batch 150/173] avg loss 2.50275e-05, throughput 5.98796K wps
Begin Testing...
[Epoch 30] train avg loss 2.28681e-05, test acc 0.7677, test avg loss 1.03121, throughput 6.01073K wps
[Epoch 31 Batch 30/173] avg loss 1.95483e-05, throughput 6.14599K wps
[Epoch 31 Batch 60/173] avg loss 2.00494e-05, throughput 6.00218K wps
[Epoch 31 Batch 90/173] avg loss 1.9977e-05, throughput 5.97959K wps
[Epoch 31 Batch 120/173] avg loss 1.96767e-05, throughput 5.97121K wps
[Epoch 31 Batch 150/173] avg loss 2.86104e-05, throughput 5.9763K wps
Begin Testing...
[Epoch 31] train avg loss 2.14763e-05, test acc 0.7656, test avg loss 1.04356, throughput 6.01276K wps
[Epoch 32 Batch 30/173] avg loss 1.51932e-05, throughput 6.12827K wps
[Epoch 32 Batch 60/173] avg loss 2.39136e-05, throughput 6.00245K wps
[Epoch 32 Batch 90/173] avg loss 1.86159e-05, throughput 5.98492K wps
[Epoch 32 Batch 120/173] avg loss 2.41304e-05, throughput 5.97935K wps
[Epoch 32 Batch 150/173] avg loss 1.46148e-05, throughput 5.9917K wps
Begin Testing...
[Epoch 32] train avg loss 1.90198e-05, test acc 0.7667, test avg loss 1.05871, throughput 6.01322K wps
[Epoch 33 Batch 30/173] avg loss 2.23353e-05, throughput 6.14706K wps
[Epoch 33 Batch 60/173] avg loss 1.17137e-05, throughput 5.98751K wps
[Epoch 33 Batch 90/173] avg loss 1.55884e-05, throughput 5.99728K wps
[Epoch 33 Batch 120/173] avg loss 1.23003e-05, throughput 5.98076K wps
[Epoch 33 Batch 150/173] avg loss 1.46369e-05, throughput 5.97466K wps
Begin Testing...
[Epoch 33] train avg loss 1.52955e-05, test acc 0.7667, test avg loss 1.07928, throughput 6.01382K wps
[Epoch 34 Batch 30/173] avg loss 1.51798e-05, throughput 6.12262K wps
[Epoch 34 Batch 60/173] avg loss 1.44105e-05, throughput 5.99976K wps
[Epoch 34 Batch 90/173] avg loss 1.98529e-05, throughput 5.99757K wps
[Epoch 34 Batch 120/173] avg loss 1.57588e-05, throughput 6.0026K wps
[Epoch 34 Batch 150/173] avg loss 1.55391e-05, throughput 5.98174K wps
Begin Testing...
[Epoch 34] train avg loss 1.54704e-05, test acc 0.7656, test avg loss 1.0847, throughput 6.01546K wps
[Epoch 35 Batch 30/173] avg loss 1.11697e-05, throughput 6.12056K wps
[Epoch 35 Batch 60/173] avg loss 8.61053e-06, throughput 5.97934K wps
[Epoch 35 Batch 90/173] avg loss 1.02798e-05, throughput 5.99576K wps
[Epoch 35 Batch 120/173] avg loss 1.32495e-05, throughput 5.98977K wps
[Epoch 35 Batch 150/173] avg loss 1.27374e-05, throughput 5.99492K wps
Begin Testing...
[Epoch 35] train avg loss 1.27538e-05, test acc 0.7646, test avg loss 1.09162, throughput 6.01255K wps
[Epoch 36 Batch 30/173] avg loss 1.00407e-05, throughput 6.1427K wps
[Epoch 36 Batch 60/173] avg loss 1.0478e-05, throughput 5.97898K wps
[Epoch 36 Batch 90/173] avg loss 1.0394e-05, throughput 5.98034K wps
[Epoch 36 Batch 120/173] avg loss 1.50965e-05, throughput 5.99098K wps
[Epoch 36 Batch 150/173] avg loss 1.07845e-05, throughput 5.99706K wps
Begin Testing...
[Epoch 36] train avg loss 1.26799e-05, test acc 0.7677, test avg loss 1.12319, throughput 6.01646K wps
[Epoch 37 Batch 30/173] avg loss 1.68041e-05, throughput 6.12454K wps
[Epoch 37 Batch 60/173] avg loss 1.14097e-05, throughput 5.98345K wps
[Epoch 37 Batch 90/173] avg loss 1.2708e-05, throughput 5.99782K wps
[Epoch 37 Batch 120/173] avg loss 7.54024e-06, throughput 5.99276K wps
[Epoch 37 Batch 150/173] avg loss 1.07086e-05, throughput 5.98584K wps
Begin Testing...
[Epoch 37] train avg loss 1.14028e-05, test acc 0.7594, test avg loss 1.141, throughput 6.0152K wps
[Epoch 38 Batch 30/173] avg loss 1.05919e-05, throughput 6.11781K wps
[Epoch 38 Batch 60/173] avg loss 8.1224e-06, throughput 5.99244K wps
[Epoch 38 Batch 90/173] avg loss 1.82024e-05, throughput 5.99261K wps
[Epoch 38 Batch 120/173] avg loss 8.29795e-06, throughput 6.00042K wps
[Epoch 38 Batch 150/173] avg loss 1.21966e-05, throughput 5.97919K wps
Begin Testing...
[Epoch 38] train avg loss 1.1143e-05, test acc 0.7667, test avg loss 1.15643, throughput 6.01236K wps
[Epoch 39 Batch 30/173] avg loss 1.57338e-05, throughput 6.12894K wps
[Epoch 39 Batch 60/173] avg loss 7.73477e-06, throughput 5.98482K wps
[Epoch 39 Batch 90/173] avg loss 7.46249e-06, throughput 5.9971K wps
[Epoch 39 Batch 120/173] avg loss 6.49e-06, throughput 5.98606K wps
[Epoch 39 Batch 150/173] avg loss 8.64611e-06, throughput 5.98312K wps
Begin Testing...
[Epoch 39] train avg loss 8.86526e-06, test acc 0.7677, test avg loss 1.17075, throughput 6.01129K wps
[Epoch 40 Batch 30/173] avg loss 1.48595e-05, throughput 6.12816K wps
[Epoch 40 Batch 60/173] avg loss 6.71336e-06, throughput 5.98903K wps
[Epoch 40 Batch 90/173] avg loss 6.17302e-06, throughput 5.99378K wps
[Epoch 40 Batch 120/173] avg loss 6.63345e-06, throughput 5.98541K wps
[Epoch 40 Batch 150/173] avg loss 5.20833e-06, throughput 5.98278K wps
Begin Testing...
[Epoch 40] train avg loss 7.95065e-06, test acc 0.7667, test avg loss 1.17944, throughput 6.01306K wps
[Epoch 41 Batch 30/173] avg loss 1.31859e-05, throughput 6.13164K wps
[Epoch 41 Batch 60/173] avg loss 8.05246e-06, throughput 5.9896K wps
[Epoch 41 Batch 90/173] avg loss 4.70599e-06, throughput 5.9863K wps
[Epoch 41 Batch 120/173] avg loss 8.86115e-06, throughput 6.00442K wps
[Epoch 41 Batch 150/173] avg loss 9.02312e-06, throughput 5.9842K wps
Begin Testing...
[Epoch 41] train avg loss 8.44088e-06, test acc 0.7667, test avg loss 1.20006, throughput 6.01571K wps
[Epoch 42 Batch 30/173] avg loss 5.44875e-06, throughput 6.14166K wps
[Epoch 42 Batch 60/173] avg loss 6.70148e-06, throughput 6.00021K wps
[Epoch 42 Batch 90/173] avg loss 4.97398e-06, throughput 5.98937K wps
[Epoch 42 Batch 120/173] avg loss 1.30722e-05, throughput 5.98588K wps
[Epoch 42 Batch 150/173] avg loss 3.98942e-06, throughput 5.98486K wps
Begin Testing...
[Epoch 42] train avg loss 7.30627e-06, test acc 0.7688, test avg loss 1.21321, throughput 6.01799K wps
[Epoch 43 Batch 30/173] avg loss 6.35382e-06, throughput 6.1293K wps
[Epoch 43 Batch 60/173] avg loss 3.77786e-06, throughput 5.99691K wps
[Epoch 43 Batch 90/173] avg loss 2.94743e-06, throughput 5.98366K wps
[Epoch 43 Batch 120/173] avg loss 4.7607e-06, throughput 5.98479K wps
[Epoch 43 Batch 150/173] avg loss 1.56429e-05, throughput 5.98462K wps
Begin Testing...
[Epoch 43] train avg loss 6.39626e-06, test acc 0.7688, test avg loss 1.219, throughput 6.01345K wps
[Epoch 44 Batch 30/173] avg loss 4.51651e-06, throughput 6.13106K wps
[Epoch 44 Batch 60/173] avg loss 5.34231e-06, throughput 5.9934K wps
[Epoch 44 Batch 90/173] avg loss 5.15818e-06, throughput 5.98989K wps
[Epoch 44 Batch 120/173] avg loss 3.83725e-06, throughput 6.0032K wps
[Epoch 44 Batch 150/173] avg loss 1.22539e-05, throughput 5.99681K wps
Begin Testing...
[Epoch 44] train avg loss 6.17835e-06, test acc 0.7677, test avg loss 1.23999, throughput 6.0191K wps
[Epoch 45 Batch 30/173] avg loss 3.49378e-06, throughput 6.13327K wps
[Epoch 45 Batch 60/173] avg loss 1.23377e-05, throughput 5.97928K wps
[Epoch 45 Batch 90/173] avg loss 5.06427e-06, throughput 6.00268K wps
[Epoch 45 Batch 120/173] avg loss 3.65817e-06, throughput 5.99853K wps
[Epoch 45 Batch 150/173] avg loss 3.4262e-06, throughput 6.00634K wps
Begin Testing...
[Epoch 45] train avg loss 5.38897e-06, test acc 0.7688, test avg loss 1.25145, throughput 6.02014K wps
[Epoch 46 Batch 30/173] avg loss 3.51829e-06, throughput 6.13066K wps
[Epoch 46 Batch 60/173] avg loss 3.71078e-06, throughput 5.98468K wps
[Epoch 46 Batch 90/173] avg loss 2.73311e-06, throughput 5.98029K wps
[Epoch 46 Batch 120/173] avg loss 2.98159e-06, throughput 5.99011K wps
[Epoch 46 Batch 150/173] avg loss 1.10342e-05, throughput 5.98771K wps
Begin Testing...
[Epoch 46] train avg loss 4.52115e-06, test acc 0.7688, test avg loss 1.27352, throughput 6.00934K wps
[Epoch 47 Batch 30/173] avg loss 2.48985e-06, throughput 6.1203K wps
[Epoch 47 Batch 60/173] avg loss 2.84922e-06, throughput 5.97352K wps
[Epoch 47 Batch 90/173] avg loss 2.76887e-06, throughput 5.98091K wps
[Epoch 47 Batch 120/173] avg loss 2.69741e-06, throughput 5.96999K wps
[Epoch 47 Batch 150/173] avg loss 1.28123e-05, throughput 5.97055K wps
Begin Testing...
[Epoch 47] train avg loss 4.46084e-06, test acc 0.7688, test avg loss 1.28445, throughput 5.99777K wps
[Epoch 48 Batch 30/173] avg loss 1.12782e-05, throughput 6.12656K wps
[Epoch 48 Batch 60/173] avg loss 3.85479e-06, throughput 5.98623K wps
[Epoch 48 Batch 90/173] avg loss 3.53027e-06, throughput 5.97434K wps
[Epoch 48 Batch 120/173] avg loss 3.1913e-06, throughput 5.99331K wps
[Epoch 48 Batch 150/173] avg loss 2.9924e-06, throughput 5.98459K wps
Begin Testing...
[Epoch 48] train avg loss 4.82835e-06, test acc 0.7760, test avg loss 1.29932, throughput 6.00992K wps
[Epoch 49 Batch 30/173] avg loss 2.24677e-06, throughput 6.14241K wps
[Epoch 49 Batch 60/173] avg loss 1.05824e-05, throughput 5.99733K wps
[Epoch 49 Batch 90/173] avg loss 2.38196e-06, throughput 5.9899K wps
[Epoch 49 Batch 120/173] avg loss 2.03423e-06, throughput 5.98442K wps
[Epoch 49 Batch 150/173] avg loss 2.94047e-06, throughput 5.98721K wps
Begin Testing...
[Epoch 49] train avg loss 3.93281e-06, test acc 0.7719, test avg loss 1.31346, throughput 6.01521K wps
[Epoch 50 Batch 30/173] avg loss 2.13649e-06, throughput 6.13147K wps
[Epoch 50 Batch 60/173] avg loss 1.0161e-05, throughput 5.98763K wps
[Epoch 50 Batch 90/173] avg loss 2.3406e-06, throughput 5.9909K wps
[Epoch 50 Batch 120/173] avg loss 2.42033e-06, throughput 5.9753K wps
[Epoch 50 Batch 150/173] avg loss 2.68355e-06, throughput 5.99204K wps
Begin Testing...
[Epoch 50] train avg loss 3.6876e-06, test acc 0.7708, test avg loss 1.33337, throughput 6.01143K wps
[Epoch 51 Batch 30/173] avg loss 1.39134e-06, throughput 6.12667K wps
[Epoch 51 Batch 60/173] avg loss 1.67248e-06, throughput 5.98994K wps
[Epoch 51 Batch 90/173] avg loss 2.21347e-06, throughput 5.99166K wps
[Epoch 51 Batch 120/173] avg loss 2.20941e-06, throughput 5.96644K wps
[Epoch 51 Batch 150/173] avg loss 1.01272e-05, throughput 5.97881K wps
Begin Testing...
[Epoch 51] train avg loss 3.33464e-06, test acc 0.7750, test avg loss 1.33719, throughput 6.00672K wps
[Epoch 52 Batch 30/173] avg loss 2.22999e-06, throughput 6.12576K wps
[Epoch 52 Batch 60/173] avg loss 2.12699e-06, throughput 5.98591K wps
[Epoch 52 Batch 90/173] avg loss 1.73823e-06, throughput 5.97978K wps
[Epoch 52 Batch 120/173] avg loss 1.06773e-05, throughput 5.98367K wps
[Epoch 52 Batch 150/173] avg loss 1.91587e-06, throughput 5.98104K wps
Begin Testing...
[Epoch 52] train avg loss 3.50815e-06, test acc 0.7708, test avg loss 1.35528, throughput 6.00818K wps
[Epoch 53 Batch 30/173] avg loss 9.41958e-06, throughput 6.14257K wps
[Epoch 53 Batch 60/173] avg loss 2.71266e-06, throughput 5.9928K wps
[Epoch 53 Batch 90/173] avg loss 2.04689e-06, throughput 5.99324K wps
[Epoch 53 Batch 120/173] avg loss 2.20792e-06, throughput 5.96988K wps
[Epoch 53 Batch 150/173] avg loss 1.30334e-06, throughput 5.97898K wps
Begin Testing...
[Epoch 53] train avg loss 3.28338e-06, test acc 0.7719, test avg loss 1.36496, throughput 6.01427K wps
[Epoch 54 Batch 30/173] avg loss 1.74688e-06, throughput 6.13788K wps
[Epoch 54 Batch 60/173] avg loss 2.46997e-06, throughput 5.99161K wps
[Epoch 54 Batch 90/173] avg loss 1.28232e-06, throughput 5.98779K wps
[Epoch 54 Batch 120/173] avg loss 2.04004e-06, throughput 5.98554K wps
[Epoch 54 Batch 150/173] avg loss 1.00021e-05, throughput 5.97684K wps
Begin Testing...
[Epoch 54] train avg loss 3.29085e-06, test acc 0.7677, test avg loss 1.39098, throughput 6.01316K wps
[Epoch 55 Batch 30/173] avg loss 2.13135e-06, throughput 6.12458K wps
[Epoch 55 Batch 60/173] avg loss 2.09941e-06, throughput 5.98435K wps
[Epoch 55 Batch 90/173] avg loss 1.63273e-06, throughput 5.99792K wps
[Epoch 55 Batch 120/173] avg loss 8.0191e-06, throughput 5.99431K wps
[Epoch 55 Batch 150/173] avg loss 1.37764e-06, throughput 5.99602K wps
Begin Testing...
[Epoch 55] train avg loss 2.8147e-06, test acc 0.7729, test avg loss 1.38897, throughput 6.01566K wps
[Epoch 56 Batch 30/173] avg loss 1.11727e-06, throughput 6.14451K wps
[Epoch 56 Batch 60/173] avg loss 9.66187e-07, throughput 5.99336K wps
[Epoch 56 Batch 90/173] avg loss 8.02431e-06, throughput 5.97936K wps
[Epoch 56 Batch 120/173] avg loss 1.42975e-06, throughput 5.97463K wps
[Epoch 56 Batch 150/173] avg loss 1.55507e-06, throughput 5.98175K wps
Begin Testing...
[Epoch 56] train avg loss 2.39751e-06, test acc 0.7708, test avg loss 1.42001, throughput 6.00969K wps
[Epoch 57 Batch 30/173] avg loss 1.26075e-06, throughput 6.1395K wps
[Epoch 57 Batch 60/173] avg loss 1.07689e-06, throughput 5.98456K wps
[Epoch 57 Batch 90/173] avg loss 1.65711e-06, throughput 5.97822K wps
[Epoch 57 Batch 120/173] avg loss 9.64835e-06, throughput 5.93814K wps
[Epoch 57 Batch 150/173] avg loss 2.22698e-06, throughput 5.94624K wps
Begin Testing...
[Epoch 57] train avg loss 3.0629e-06, test acc 0.7719, test avg loss 1.42076, throughput 5.99789K wps
[Epoch 58 Batch 30/173] avg loss 7.52799e-06, throughput 6.12815K wps
[Epoch 58 Batch 60/173] avg loss 9.90286e-07, throughput 5.9877K wps
[Epoch 58 Batch 90/173] avg loss 1.02309e-06, throughput 5.99616K wps
[Epoch 58 Batch 120/173] avg loss 1.55719e-06, throughput 5.99496K wps
[Epoch 58 Batch 150/173] avg loss 1.25346e-06, throughput 5.98494K wps
Begin Testing...
[Epoch 58] train avg loss 2.26809e-06, test acc 0.7708, test avg loss 1.44068, throughput 6.01536K wps
[Epoch 59 Batch 30/173] avg loss 8.53806e-07, throughput 6.13523K wps
[Epoch 59 Batch 60/173] avg loss 1.20555e-06, throughput 5.97752K wps
[Epoch 59 Batch 90/173] avg loss 1.09193e-06, throughput 5.98346K wps
[Epoch 59 Batch 120/173] avg loss 7.59217e-06, throughput 5.97812K wps
[Epoch 59 Batch 150/173] avg loss 9.31677e-07, throughput 5.98867K wps
Begin Testing...
[Epoch 59] train avg loss 2.15846e-06, test acc 0.7708, test avg loss 1.45065, throughput 6.01008K wps
Test loss 0.486356, test acc 0.7758
Total time cost 358.31s
9596 1066
[Epoch 0 Batch 30/173] avg loss 0.0138642, throughput 5.79488K wps
[Epoch 0 Batch 60/173] avg loss 0.0138202, throughput 5.98247K wps
[Epoch 0 Batch 90/173] avg loss 0.0138343, throughput 5.98691K wps
[Epoch 0 Batch 120/173] avg loss 0.0138026, throughput 5.9848K wps
[Epoch 0 Batch 150/173] avg loss 0.013747, throughput 5.97923K wps
Begin Testing...
[Epoch 0] train avg loss 0.0138261, test acc 0.6333, test avg loss 0.684795, throughput 5.95019K wps
Observed Improvement.
Begin Testing...
[Epoch 1 Batch 30/173] avg loss 0.0134773, throughput 6.13553K wps
[Epoch 1 Batch 60/173] avg loss 0.0134463, throughput 5.98371K wps
[Epoch 1 Batch 90/173] avg loss 0.0133635, throughput 5.99436K wps
[Epoch 1 Batch 120/173] avg loss 0.0133226, throughput 5.98611K wps
[Epoch 1 Batch 150/173] avg loss 0.0132735, throughput 5.98894K wps
Begin Testing...
[Epoch 1] train avg loss 0.0133762, test acc 0.6719, test avg loss 0.667333, throughput 6.01462K wps
Observed Improvement.
Begin Testing...
[Epoch 2 Batch 30/173] avg loss 0.0127935, throughput 6.14773K wps
[Epoch 2 Batch 60/173] avg loss 0.012713, throughput 5.98485K wps
[Epoch 2 Batch 90/173] avg loss 0.0126102, throughput 5.98551K wps
[Epoch 2 Batch 120/173] avg loss 0.0124084, throughput 5.99862K wps
[Epoch 2 Batch 150/173] avg loss 0.0122155, throughput 5.98429K wps
Begin Testing...
[Epoch 2] train avg loss 0.0125024, test acc 0.7146, test avg loss 0.622456, throughput 6.0147K wps
Observed Improvement.
Begin Testing...
[Epoch 3 Batch 30/173] avg loss 0.0113551, throughput 6.14279K wps
[Epoch 3 Batch 60/173] avg loss 0.0110548, throughput 5.98049K wps
[Epoch 3 Batch 90/173] avg loss 0.0110587, throughput 5.9875K wps
[Epoch 3 Batch 120/173] avg loss 0.010747, throughput 5.98274K wps
[Epoch 3 Batch 150/173] avg loss 0.0105032, throughput 5.98327K wps
Begin Testing...
[Epoch 3] train avg loss 0.0108473, test acc 0.7771, test avg loss 0.543502, throughput 6.01265K wps
Observed Improvement.
Begin Testing...
[Epoch 4 Batch 30/173] avg loss 0.00902327, throughput 6.1284K wps
[Epoch 4 Batch 60/173] avg loss 0.00859206, throughput 5.97601K wps
[Epoch 4 Batch 90/173] avg loss 0.00860982, throughput 5.99055K wps
[Epoch 4 Batch 120/173] avg loss 0.00818004, throughput 5.99282K wps
[Epoch 4 Batch 150/173] avg loss 0.00803162, throughput 5.9824K wps
Begin Testing...
[Epoch 4] train avg loss 0.00840318, test acc 0.7906, test avg loss 0.474606, throughput 6.01085K wps
Observed Improvement.
Begin Testing...
[Epoch 5 Batch 30/173] avg loss 0.00656532, throughput 6.1274K wps
[Epoch 5 Batch 60/173] avg loss 0.0062881, throughput 5.99397K wps
[Epoch 5 Batch 90/173] avg loss 0.00618631, throughput 5.98513K wps
[Epoch 5 Batch 120/173] avg loss 0.00571909, throughput 5.99544K wps
[Epoch 5 Batch 150/173] avg loss 0.00562037, throughput 5.99653K wps
Begin Testing...
[Epoch 5] train avg loss 0.00601378, test acc 0.7885, test avg loss 0.446974, throughput 6.01639K wps
[Epoch 6 Batch 30/173] avg loss 0.00418466, throughput 6.1338K wps
[Epoch 6 Batch 60/173] avg loss 0.00415374, throughput 5.98764K wps
[Epoch 6 Batch 90/173] avg loss 0.00421718, throughput 5.97544K wps
[Epoch 6 Batch 120/173] avg loss 0.00432242, throughput 5.99588K wps
[Epoch 6 Batch 150/173] avg loss 0.00455855, throughput 5.99698K wps
Begin Testing...
[Epoch 6] train avg loss 0.00424279, test acc 0.7833, test avg loss 0.459161, throughput 6.01408K wps
[Epoch 7 Batch 30/173] avg loss 0.00286645, throughput 6.13498K wps
[Epoch 7 Batch 60/173] avg loss 0.00308125, throughput 5.99476K wps
[Epoch 7 Batch 90/173] avg loss 0.00282598, throughput 5.98691K wps
[Epoch 7 Batch 120/173] avg loss 0.00288093, throughput 5.97723K wps
[Epoch 7 Batch 150/173] avg loss 0.00308242, throughput 5.99014K wps
Begin Testing...
[Epoch 7] train avg loss 0.00292961, test acc 0.7854, test avg loss 0.489417, throughput 6.01439K wps
[Epoch 8 Batch 30/173] avg loss 0.00198144, throughput 6.12687K wps
[Epoch 8 Batch 60/173] avg loss 0.0022266, throughput 5.99542K wps
[Epoch 8 Batch 90/173] avg loss 0.00199596, throughput 6.00178K wps
[Epoch 8 Batch 120/173] avg loss 0.00199158, throughput 5.99585K wps
[Epoch 8 Batch 150/173] avg loss 0.00195494, throughput 5.98845K wps
Begin Testing...
[Epoch 8] train avg loss 0.002044, test acc 0.7771, test avg loss 0.528848, throughput 6.01895K wps
[Epoch 9 Batch 30/173] avg loss 0.00137647, throughput 6.13053K wps
[Epoch 9 Batch 60/173] avg loss 0.00135537, throughput 5.99955K wps
[Epoch 9 Batch 90/173] avg loss 0.00135714, throughput 5.98943K wps
[Epoch 9 Batch 120/173] avg loss 0.00162361, throughput 5.9892K wps
[Epoch 9 Batch 150/173] avg loss 0.0015268, throughput 5.98223K wps
Begin Testing...
[Epoch 9] train avg loss 0.00144756, test acc 0.7719, test avg loss 0.569962, throughput 6.01437K wps
[Epoch 10 Batch 30/173] avg loss 0.000933366, throughput 6.13712K wps
[Epoch 10 Batch 60/173] avg loss 0.00109164, throughput 5.99028K wps
[Epoch 10 Batch 90/173] avg loss 0.00107098, throughput 5.98025K wps
[Epoch 10 Batch 120/173] avg loss 0.000947505, throughput 5.97908K wps
[Epoch 10 Batch 150/173] avg loss 0.000899051, throughput 5.9947K wps
Begin Testing...
[Epoch 10] train avg loss 0.00100599, test acc 0.7635, test avg loss 0.614553, throughput 6.01308K wps
[Epoch 11 Batch 30/173] avg loss 0.000735668, throughput 6.13257K wps
[Epoch 11 Batch 60/173] avg loss 0.000778535, throughput 5.98835K wps
[Epoch 11 Batch 90/173] avg loss 0.000607436, throughput 5.99701K wps
[Epoch 11 Batch 120/173] avg loss 0.000751207, throughput 5.98012K wps
[Epoch 11 Batch 150/173] avg loss 0.000801654, throughput 6.00931K wps
Begin Testing...
[Epoch 11] train avg loss 0.000751838, test acc 0.7594, test avg loss 0.658323, throughput 6.0187K wps
[Epoch 12 Batch 30/173] avg loss 0.000506868, throughput 6.13405K wps
[Epoch 12 Batch 60/173] avg loss 0.000589814, throughput 5.99927K wps
[Epoch 12 Batch 90/173] avg loss 0.000502215, throughput 5.99236K wps
[Epoch 12 Batch 120/173] avg loss 0.000531221, throughput 5.97902K wps
[Epoch 12 Batch 150/173] avg loss 0.00055875, throughput 5.98348K wps
Begin Testing...
[Epoch 12] train avg loss 0.000549598, test acc 0.7604, test avg loss 0.696421, throughput 6.0159K wps
[Epoch 13 Batch 30/173] avg loss 0.000431999, throughput 6.13904K wps
[Epoch 13 Batch 60/173] avg loss 0.000386904, throughput 6.00046K wps
[Epoch 13 Batch 90/173] avg loss 0.000367795, throughput 5.98986K wps
[Epoch 13 Batch 120/173] avg loss 0.000415489, throughput 5.97916K wps
[Epoch 13 Batch 150/173] avg loss 0.000427722, throughput 5.986K wps
Begin Testing...
[Epoch 13] train avg loss 0.00041125, test acc 0.7552, test avg loss 0.7284, throughput 6.01757K wps
[Epoch 14 Batch 30/173] avg loss 0.000289291, throughput 6.13015K wps
[Epoch 14 Batch 60/173] avg loss 0.000277476, throughput 5.99394K wps
[Epoch 14 Batch 90/173] avg loss 0.000349846, throughput 5.98313K wps
[Epoch 14 Batch 120/173] avg loss 0.000329109, throughput 5.97887K wps
[Epoch 14 Batch 150/173] avg loss 0.000306589, throughput 5.98359K wps
Begin Testing...
[Epoch 14] train avg loss 0.000320008, test acc 0.7552, test avg loss 0.76416, throughput 6.00926K wps
[Epoch 15 Batch 30/173] avg loss 0.000242697, throughput 6.1409K wps
[Epoch 15 Batch 60/173] avg loss 0.000260652, throughput 5.98638K wps
[Epoch 15 Batch 90/173] avg loss 0.000258914, throughput 5.9765K wps
[Epoch 15 Batch 120/173] avg loss 0.000233059, throughput 5.98865K wps
[Epoch 15 Batch 150/173] avg loss 0.00024452, throughput 5.99538K wps
Begin Testing...
[Epoch 15] train avg loss 0.000250901, test acc 0.7500, test avg loss 0.801536, throughput 6.0144K wps
[Epoch 16 Batch 30/173] avg loss 0.00020454, throughput 6.13597K wps
[Epoch 16 Batch 60/173] avg loss 0.000179461, throughput 5.99725K wps
[Epoch 16 Batch 90/173] avg loss 0.000194386, throughput 5.99862K wps
[Epoch 16 Batch 120/173] avg loss 0.00020249, throughput 5.99633K wps
[Epoch 16 Batch 150/173] avg loss 0.000216384, throughput 5.99017K wps
Begin Testing...
[Epoch 16] train avg loss 0.000200047, test acc 0.7490, test avg loss 0.832099, throughput 6.01842K wps
[Epoch 17 Batch 30/173] avg loss 0.000132867, throughput 6.13486K wps
[Epoch 17 Batch 60/173] avg loss 0.000167409, throughput 5.99086K wps
[Epoch 17 Batch 90/173] avg loss 0.000148235, throughput 5.98445K wps
[Epoch 17 Batch 120/173] avg loss 0.000167064, throughput 5.97497K wps
[Epoch 17 Batch 150/173] avg loss 0.000154211, throughput 5.97771K wps
Begin Testing...
[Epoch 17] train avg loss 0.000160021, test acc 0.7521, test avg loss 0.856437, throughput 6.00819K wps
[Epoch 18 Batch 30/173] avg loss 0.000126049, throughput 6.13921K wps
[Epoch 18 Batch 60/173] avg loss 0.000135767, throughput 5.99068K wps
[Epoch 18 Batch 90/173] avg loss 0.000137529, throughput 5.98484K wps
[Epoch 18 Batch 120/173] avg loss 0.000132137, throughput 5.99003K wps
[Epoch 18 Batch 150/173] avg loss 0.000134078, throughput 5.99792K wps
Begin Testing...
[Epoch 18] train avg loss 0.0001336, test acc 0.7510, test avg loss 0.886138, throughput 6.01738K wps
[Epoch 19 Batch 30/173] avg loss 9.57306e-05, throughput 6.13751K wps
[Epoch 19 Batch 60/173] avg loss 0.000110654, throughput 5.99122K wps
[Epoch 19 Batch 90/173] avg loss 9.49765e-05, throughput 5.98755K wps
[Epoch 19 Batch 120/173] avg loss 0.000133226, throughput 5.98321K wps
[Epoch 19 Batch 150/173] avg loss 0.000115283, throughput 5.98522K wps
Begin Testing...
[Epoch 19] train avg loss 0.000110996, test acc 0.7521, test avg loss 0.917484, throughput 6.01395K wps
[Epoch 20 Batch 30/173] avg loss 8.55233e-05, throughput 6.12972K wps
[Epoch 20 Batch 60/173] avg loss 9.8235e-05, throughput 5.97537K wps
[Epoch 20 Batch 90/173] avg loss 8.99932e-05, throughput 5.98223K wps
[Epoch 20 Batch 120/173] avg loss 0.000117732, throughput 5.97786K wps
[Epoch 20 Batch 150/173] avg loss 9.08978e-05, throughput 5.97579K wps
Begin Testing...
[Epoch 20] train avg loss 9.76031e-05, test acc 0.7479, test avg loss 0.950037, throughput 6.00584K wps
[Epoch 21 Batch 30/173] avg loss 7.24207e-05, throughput 6.13712K wps
[Epoch 21 Batch 60/173] avg loss 7.15401e-05, throughput 5.98834K wps
[Epoch 21 Batch 90/173] avg loss 8.09732e-05, throughput 5.9805K wps
[Epoch 21 Batch 120/173] avg loss 8.18917e-05, throughput 5.9859K wps
[Epoch 21 Batch 150/173] avg loss 7.88478e-05, throughput 5.97882K wps
Begin Testing...
[Epoch 21] train avg loss 7.87207e-05, test acc 0.7448, test avg loss 0.979831, throughput 6.01049K wps
[Epoch 22 Batch 30/173] avg loss 5.82003e-05, throughput 6.12768K wps
[Epoch 22 Batch 60/173] avg loss 5.71638e-05, throughput 5.98038K wps
[Epoch 22 Batch 90/173] avg loss 6.97338e-05, throughput 5.9883K wps
[Epoch 22 Batch 120/173] avg loss 6.86394e-05, throughput 5.99085K wps
[Epoch 22 Batch 150/173] avg loss 6.56037e-05, throughput 5.98789K wps
Begin Testing...
[Epoch 22] train avg loss 6.46866e-05, test acc 0.7469, test avg loss 0.999689, throughput 6.01114K wps
[Epoch 23 Batch 30/173] avg loss 4.54341e-05, throughput 6.12785K wps
[Epoch 23 Batch 60/173] avg loss 6.2586e-05, throughput 5.98704K wps
[Epoch 23 Batch 90/173] avg loss 5.10096e-05, throughput 5.98129K wps
[Epoch 23 Batch 120/173] avg loss 5.03213e-05, throughput 5.97943K wps
[Epoch 23 Batch 150/173] avg loss 5.62567e-05, throughput 5.97497K wps
Begin Testing...
[Epoch 23] train avg loss 5.29916e-05, test acc 0.7510, test avg loss 1.02037, throughput 6.00784K wps
[Epoch 24 Batch 30/173] avg loss 4.80646e-05, throughput 6.12697K wps
[Epoch 24 Batch 60/173] avg loss 5.57075e-05, throughput 5.9952K wps
[Epoch 24 Batch 90/173] avg loss 4.41824e-05, throughput 5.98421K wps
[Epoch 24 Batch 120/173] avg loss 3.57295e-05, throughput 5.98973K wps
[Epoch 24 Batch 150/173] avg loss 4.3924e-05, throughput 5.99429K wps
Begin Testing...
[Epoch 24] train avg loss 4.70338e-05, test acc 0.7479, test avg loss 1.04723, throughput 6.01491K wps
[Epoch 25 Batch 30/173] avg loss 5.2266e-05, throughput 6.13739K wps
[Epoch 25 Batch 60/173] avg loss 3.7513e-05, throughput 5.98112K wps
[Epoch 25 Batch 90/173] avg loss 5.295e-05, throughput 5.98838K wps
[Epoch 25 Batch 120/173] avg loss 4.64562e-05, throughput 5.97548K wps
[Epoch 25 Batch 150/173] avg loss 4.57396e-05, throughput 5.98073K wps
Begin Testing...
[Epoch 25] train avg loss 4.5399e-05, test acc 0.7479, test avg loss 1.06641, throughput 6.0081K wps
[Epoch 26 Batch 30/173] avg loss 3.384e-05, throughput 6.13025K wps
[Epoch 26 Batch 60/173] avg loss 3.86538e-05, throughput 5.98843K wps
[Epoch 26 Batch 90/173] avg loss 3.45258e-05, throughput 5.98179K wps
[Epoch 26 Batch 120/173] avg loss 3.79312e-05, throughput 5.98243K wps
[Epoch 26 Batch 150/173] avg loss 4.44872e-05, throughput 5.97626K wps
Begin Testing...
[Epoch 26] train avg loss 3.81387e-05, test acc 0.7510, test avg loss 1.08724, throughput 6.00853K wps
[Epoch 27 Batch 30/173] avg loss 3.06525e-05, throughput 6.13237K wps
[Epoch 27 Batch 60/173] avg loss 3.7619e-05, throughput 5.98286K wps
[Epoch 27 Batch 90/173] avg loss 2.44141e-05, throughput 5.98112K wps
[Epoch 27 Batch 120/173] avg loss 2.60421e-05, throughput 5.97055K wps
[Epoch 27 Batch 150/173] avg loss 3.9855e-05, throughput 5.96313K wps
Begin Testing...
[Epoch 27] train avg loss 3.17773e-05, test acc 0.7521, test avg loss 1.11354, throughput 6.00033K wps
[Epoch 28 Batch 30/173] avg loss 2.63647e-05, throughput 6.12587K wps
[Epoch 28 Batch 60/173] avg loss 2.36105e-05, throughput 5.97486K wps
[Epoch 28 Batch 90/173] avg loss 2.48442e-05, throughput 5.97069K wps
[Epoch 28 Batch 120/173] avg loss 2.45937e-05, throughput 5.9867K wps
[Epoch 28 Batch 150/173] avg loss 4.15401e-05, throughput 5.98095K wps
Begin Testing...
[Epoch 28] train avg loss 2.88549e-05, test acc 0.7552, test avg loss 1.13003, throughput 6.00549K wps
[Epoch 29 Batch 30/173] avg loss 2.06105e-05, throughput 6.12836K wps
[Epoch 29 Batch 60/173] avg loss 2.54567e-05, throughput 5.98723K wps
[Epoch 29 Batch 90/173] avg loss 3.44591e-05, throughput 5.99404K wps
[Epoch 29 Batch 120/173] avg loss 2.69178e-05, throughput 5.99727K wps
[Epoch 29 Batch 150/173] avg loss 2.18006e-05, throughput 5.98167K wps
Begin Testing...
[Epoch 29] train avg loss 2.65361e-05, test acc 0.7510, test avg loss 1.14788, throughput 6.01347K wps
[Epoch 30 Batch 30/173] avg loss 2.95857e-05, throughput 6.13332K wps
[Epoch 30 Batch 60/173] avg loss 2.27767e-05, throughput 5.97933K wps
[Epoch 30 Batch 90/173] avg loss 2.12589e-05, throughput 5.97872K wps
[Epoch 30 Batch 120/173] avg loss 2.18786e-05, throughput 5.97786K wps
[Epoch 30 Batch 150/173] avg loss 2.24123e-05, throughput 5.98067K wps
Begin Testing...
[Epoch 30] train avg loss 2.29175e-05, test acc 0.7510, test avg loss 1.17502, throughput 6.00716K wps
[Epoch 31 Batch 30/173] avg loss 1.64555e-05, throughput 6.13018K wps
[Epoch 31 Batch 60/173] avg loss 1.61825e-05, throughput 5.99071K wps
[Epoch 31 Batch 90/173] avg loss 1.91357e-05, throughput 5.98384K wps
[Epoch 31 Batch 120/173] avg loss 4.31037e-05, throughput 5.98996K wps
[Epoch 31 Batch 150/173] avg loss 2.88988e-05, throughput 5.97632K wps
Begin Testing...
[Epoch 31] train avg loss 2.45693e-05, test acc 0.7490, test avg loss 1.20184, throughput 6.00849K wps
[Epoch 32 Batch 30/173] avg loss 1.75055e-05, throughput 6.11347K wps
[Epoch 32 Batch 60/173] avg loss 1.9984e-05, throughput 5.97897K wps
[Epoch 32 Batch 90/173] avg loss 1.74297e-05, throughput 5.96479K wps
[Epoch 32 Batch 120/173] avg loss 1.74739e-05, throughput 5.98414K wps
[Epoch 32 Batch 150/173] avg loss 1.87274e-05, throughput 5.97829K wps
Begin Testing...
[Epoch 32] train avg loss 1.93585e-05, test acc 0.7510, test avg loss 1.21723, throughput 6.00133K wps
[Epoch 33 Batch 30/173] avg loss 2.19207e-05, throughput 6.13283K wps
[Epoch 33 Batch 60/173] avg loss 1.10084e-05, throughput 5.98091K wps
[Epoch 33 Batch 90/173] avg loss 1.42996e-05, throughput 5.97156K wps
[Epoch 33 Batch 120/173] avg loss 1.2425e-05, throughput 5.97662K wps
[Epoch 33 Batch 150/173] avg loss 1.73956e-05, throughput 5.97701K wps
Begin Testing...
[Epoch 33] train avg loss 1.51287e-05, test acc 0.7500, test avg loss 1.23666, throughput 6.00539K wps
[Epoch 34 Batch 30/173] avg loss 1.46327e-05, throughput 6.13536K wps
[Epoch 34 Batch 60/173] avg loss 1.18102e-05, throughput 5.98665K wps
[Epoch 34 Batch 90/173] avg loss 1.23708e-05, throughput 5.98754K wps
[Epoch 34 Batch 120/173] avg loss 1.1957e-05, throughput 5.98406K wps
[Epoch 34 Batch 150/173] avg loss 1.4728e-05, throughput 5.99681K wps
Begin Testing...
[Epoch 34] train avg loss 1.4731e-05, test acc 0.7469, test avg loss 1.25638, throughput 6.01502K wps
[Epoch 35 Batch 30/173] avg loss 1.18391e-05, throughput 6.13685K wps
[Epoch 35 Batch 60/173] avg loss 1.97885e-05, throughput 5.98784K wps
[Epoch 35 Batch 90/173] avg loss 1.00059e-05, throughput 5.97764K wps
[Epoch 35 Batch 120/173] avg loss 1.19416e-05, throughput 5.98759K wps
[Epoch 35 Batch 150/173] avg loss 1.09346e-05, throughput 5.97882K wps
Begin Testing...
[Epoch 35] train avg loss 1.23661e-05, test acc 0.7438, test avg loss 1.27827, throughput 6.01043K wps
[Epoch 36 Batch 30/173] avg loss 9.62497e-06, throughput 6.13731K wps
[Epoch 36 Batch 60/173] avg loss 1.11668e-05, throughput 5.99325K wps
[Epoch 36 Batch 90/173] avg loss 9.53369e-06, throughput 5.97663K wps
[Epoch 36 Batch 120/173] avg loss 2.30673e-05, throughput 5.9779K wps
[Epoch 36 Batch 150/173] avg loss 1.24135e-05, throughput 5.98427K wps
Begin Testing...
[Epoch 36] train avg loss 1.25856e-05, test acc 0.7448, test avg loss 1.29557, throughput 6.01088K wps
[Epoch 37 Batch 30/173] avg loss 8.57462e-06, throughput 6.13525K wps
[Epoch 37 Batch 60/173] avg loss 8.27841e-06, throughput 5.98843K wps
[Epoch 37 Batch 90/173] avg loss 1.82466e-05, throughput 5.96837K wps
[Epoch 37 Batch 120/173] avg loss 1.45294e-05, throughput 5.97305K wps
[Epoch 37 Batch 150/173] avg loss 9.77792e-06, throughput 5.98374K wps
Begin Testing...
[Epoch 37] train avg loss 1.14416e-05, test acc 0.7500, test avg loss 1.31855, throughput 6.00568K wps
[Epoch 38 Batch 30/173] avg loss 1.08129e-05, throughput 6.13235K wps
[Epoch 38 Batch 60/173] avg loss 8.40445e-06, throughput 5.99123K wps
[Epoch 38 Batch 90/173] avg loss 8.78759e-06, throughput 5.98658K wps
[Epoch 38 Batch 120/173] avg loss 8.23223e-06, throughput 5.97887K wps
[Epoch 38 Batch 150/173] avg loss 1.72289e-05, throughput 5.96586K wps
Begin Testing...
[Epoch 38] train avg loss 1.07872e-05, test acc 0.7438, test avg loss 1.33714, throughput 6.00663K wps
[Epoch 39 Batch 30/173] avg loss 6.18974e-06, throughput 6.12442K wps
[Epoch 39 Batch 60/173] avg loss 1.52328e-05, throughput 5.9745K wps
[Epoch 39 Batch 90/173] avg loss 6.79298e-06, throughput 5.97918K wps
[Epoch 39 Batch 120/173] avg loss 1.09306e-05, throughput 5.98387K wps
[Epoch 39 Batch 150/173] avg loss 7.27361e-06, throughput 5.98732K wps
Begin Testing...
[Epoch 39] train avg loss 9.10926e-06, test acc 0.7448, test avg loss 1.35654, throughput 6.00815K wps
[Epoch 40 Batch 30/173] avg loss 6.29796e-06, throughput 6.12395K wps
[Epoch 40 Batch 60/173] avg loss 5.35228e-06, throughput 5.98537K wps
[Epoch 40 Batch 90/173] avg loss 1.67508e-05, throughput 5.98736K wps
[Epoch 40 Batch 120/173] avg loss 5.0875e-06, throughput 5.98043K wps
[Epoch 40 Batch 150/173] avg loss 8.14877e-06, throughput 5.99292K wps
Begin Testing...
[Epoch 40] train avg loss 8.11208e-06, test acc 0.7458, test avg loss 1.37802, throughput 6.01007K wps
[Epoch 41 Batch 30/173] avg loss 1.44154e-05, throughput 6.12632K wps
[Epoch 41 Batch 60/173] avg loss 5.75118e-06, throughput 5.98277K wps
[Epoch 41 Batch 90/173] avg loss 1.59704e-05, throughput 5.98127K wps
[Epoch 41 Batch 120/173] avg loss 1.51706e-05, throughput 5.99604K wps
[Epoch 41 Batch 150/173] avg loss 9.28761e-06, throughput 5.98009K wps
Begin Testing...
[Epoch 41] train avg loss 1.16071e-05, test acc 0.7448, test avg loss 1.39576, throughput 6.01073K wps
[Epoch 42 Batch 30/173] avg loss 1.40327e-05, throughput 6.12928K wps
[Epoch 42 Batch 60/173] avg loss 5.94e-06, throughput 5.98771K wps
[Epoch 42 Batch 90/173] avg loss 5.29047e-06, throughput 5.99536K wps
[Epoch 42 Batch 120/173] avg loss 4.88393e-06, throughput 5.98737K wps
[Epoch 42 Batch 150/173] avg loss 5.14627e-06, throughput 5.9852K wps
Begin Testing...
[Epoch 42] train avg loss 6.8538e-06, test acc 0.7448, test avg loss 1.40756, throughput 6.01091K wps
[Epoch 43 Batch 30/173] avg loss 3.86584e-06, throughput 6.13691K wps
[Epoch 43 Batch 60/173] avg loss 1.21623e-05, throughput 5.97222K wps
[Epoch 43 Batch 90/173] avg loss 5.22943e-06, throughput 5.97847K wps
[Epoch 43 Batch 120/173] avg loss 4.34023e-06, throughput 5.98298K wps
[Epoch 43 Batch 150/173] avg loss 4.96306e-06, throughput 5.98133K wps
Begin Testing...
[Epoch 43] train avg loss 5.82241e-06, test acc 0.7469, test avg loss 1.41796, throughput 6.00622K wps
[Epoch 44 Batch 30/173] avg loss 3.72795e-06, throughput 6.12015K wps
[Epoch 44 Batch 60/173] avg loss 1.26716e-05, throughput 5.99162K wps
[Epoch 44 Batch 90/173] avg loss 5.1211e-06, throughput 5.98768K wps
[Epoch 44 Batch 120/173] avg loss 3.97122e-06, throughput 5.98457K wps
[Epoch 44 Batch 150/173] avg loss 4.98998e-06, throughput 5.97927K wps
Begin Testing...
[Epoch 44] train avg loss 5.88939e-06, test acc 0.7427, test avg loss 1.44496, throughput 6.00869K wps
[Epoch 45 Batch 30/173] avg loss 1.14037e-05, throughput 6.13498K wps
[Epoch 45 Batch 60/173] avg loss 4.37217e-06, throughput 5.97805K wps
[Epoch 45 Batch 90/173] avg loss 5.84661e-06, throughput 5.98102K wps
[Epoch 45 Batch 120/173] avg loss 4.62939e-06, throughput 5.98062K wps
[Epoch 45 Batch 150/173] avg loss 4.47421e-06, throughput 5.98624K wps
Begin Testing...
[Epoch 45] train avg loss 5.99242e-06, test acc 0.7417, test avg loss 1.46856, throughput 6.00737K wps
[Epoch 46 Batch 30/173] avg loss 1.07511e-05, throughput 6.12824K wps
[Epoch 46 Batch 60/173] avg loss 3.47802e-06, throughput 5.98048K wps
[Epoch 46 Batch 90/173] avg loss 3.6014e-06, throughput 5.98463K wps
[Epoch 46 Batch 120/173] avg loss 3.03345e-06, throughput 5.97889K wps
[Epoch 46 Batch 150/173] avg loss 3.58846e-06, throughput 5.97762K wps
Begin Testing...
[Epoch 46] train avg loss 4.89847e-06, test acc 0.7427, test avg loss 1.47936, throughput 6.00624K wps
[Epoch 47 Batch 30/173] avg loss 2.74473e-06, throughput 6.13685K wps
[Epoch 47 Batch 60/173] avg loss 2.88617e-06, throughput 5.98188K wps
[Epoch 47 Batch 90/173] avg loss 3.66255e-06, throughput 5.98731K wps
[Epoch 47 Batch 120/173] avg loss 2.47608e-06, throughput 5.97875K wps
[Epoch 47 Batch 150/173] avg loss 2.8938e-06, throughput 5.98482K wps
Begin Testing...
[Epoch 47] train avg loss 4.57725e-06, test acc 0.7458, test avg loss 1.50052, throughput 6.0101K wps
[Epoch 48 Batch 30/173] avg loss 1.02683e-05, throughput 6.1226K wps
[Epoch 48 Batch 60/173] avg loss 2.75057e-06, throughput 5.9658K wps
[Epoch 48 Batch 90/173] avg loss 2.05894e-06, throughput 5.98347K wps
[Epoch 48 Batch 120/173] avg loss 2.71456e-06, throughput 5.9807K wps
[Epoch 48 Batch 150/173] avg loss 5.40724e-06, throughput 5.97986K wps
Begin Testing...
[Epoch 48] train avg loss 4.48147e-06, test acc 0.7427, test avg loss 1.52379, throughput 6.00271K wps
[Epoch 49 Batch 30/173] avg loss 2.48939e-06, throughput 6.13609K wps
[Epoch 49 Batch 60/173] avg loss 2.39546e-06, throughput 5.97544K wps
[Epoch 49 Batch 90/173] avg loss 1.07668e-05, throughput 5.97278K wps
[Epoch 49 Batch 120/173] avg loss 4.83212e-06, throughput 5.97983K wps
[Epoch 49 Batch 150/173] avg loss 2.2941e-06, throughput 5.98684K wps
Begin Testing...
[Epoch 49] train avg loss 4.31424e-06, test acc 0.7438, test avg loss 1.5364, throughput 6.00547K wps
[Epoch 50 Batch 30/173] avg loss 9.09173e-06, throughput 6.13558K wps
[Epoch 50 Batch 60/173] avg loss 2.58264e-06, throughput 5.99667K wps
[Epoch 50 Batch 90/173] avg loss 3.42192e-06, throughput 5.97915K wps
[Epoch 50 Batch 120/173] avg loss 2.09704e-06, throughput 5.9718K wps
[Epoch 50 Batch 150/173] avg loss 2.1764e-06, throughput 5.98313K wps
Begin Testing...
[Epoch 50] train avg loss 3.60932e-06, test acc 0.7448, test avg loss 1.55198, throughput 6.00871K wps
[Epoch 51 Batch 30/173] avg loss 2.03249e-06, throughput 6.12567K wps
[Epoch 51 Batch 60/173] avg loss 8.36567e-06, throughput 5.9669K wps
[Epoch 51 Batch 90/173] avg loss 1.51821e-06, throughput 5.97332K wps
[Epoch 51 Batch 120/173] avg loss 2.31224e-06, throughput 5.96645K wps
[Epoch 51 Batch 150/173] avg loss 2.03707e-06, throughput 5.98957K wps
Begin Testing...
[Epoch 51] train avg loss 3.09404e-06, test acc 0.7448, test avg loss 1.57113, throughput 6.00278K wps
[Epoch 52 Batch 30/173] avg loss 1.98191e-06, throughput 6.12971K wps
[Epoch 52 Batch 60/173] avg loss 1.64565e-06, throughput 5.99026K wps
[Epoch 52 Batch 90/173] avg loss 1.72294e-06, throughput 5.97842K wps
[Epoch 52 Batch 120/173] avg loss 8.93141e-06, throughput 5.97831K wps
[Epoch 52 Batch 150/173] avg loss 1.78452e-06, throughput 5.9816K wps
Begin Testing...
[Epoch 52] train avg loss 3.11265e-06, test acc 0.7458, test avg loss 1.59049, throughput 6.00754K wps
[Epoch 53 Batch 30/173] avg loss 1.86162e-06, throughput 6.13534K wps
[Epoch 53 Batch 60/173] avg loss 1.49564e-06, throughput 5.99656K wps
[Epoch 53 Batch 90/173] avg loss 1.31588e-06, throughput 5.97827K wps
[Epoch 53 Batch 120/173] avg loss 2.48401e-06, throughput 5.98976K wps
[Epoch 53 Batch 150/173] avg loss 6.75293e-06, throughput 5.98785K wps
Begin Testing...
[Epoch 53] train avg loss 2.68622e-06, test acc 0.7427, test avg loss 1.61367, throughput 6.01467K wps
[Epoch 54 Batch 30/173] avg loss 1.52778e-06, throughput 6.13837K wps
[Epoch 54 Batch 60/173] avg loss 1.76859e-06, throughput 5.97697K wps
[Epoch 54 Batch 90/173] avg loss 1.78818e-06, throughput 5.98668K wps
[Epoch 54 Batch 120/173] avg loss 1.24202e-06, throughput 5.988K wps
[Epoch 54 Batch 150/173] avg loss 8.60824e-06, throughput 5.97004K wps
Begin Testing...
[Epoch 54] train avg loss 3.06052e-06, test acc 0.7417, test avg loss 1.64088, throughput 6.00905K wps
[Epoch 55 Batch 30/173] avg loss 2.22647e-06, throughput 6.1296K wps
[Epoch 55 Batch 60/173] avg loss 1.36579e-05, throughput 5.98699K wps
[Epoch 55 Batch 90/173] avg loss 3.19969e-06, throughput 5.96841K wps
[Epoch 55 Batch 120/173] avg loss 4.24505e-06, throughput 5.98159K wps
[Epoch 55 Batch 150/173] avg loss 3.7149e-06, throughput 5.98008K wps
Begin Testing...
[Epoch 55] train avg loss 5.04798e-06, test acc 0.7427, test avg loss 1.65383, throughput 6.00674K wps
[Epoch 56 Batch 30/173] avg loss 1.97097e-06, throughput 6.13993K wps
[Epoch 56 Batch 60/173] avg loss 1.71246e-06, throughput 5.97874K wps
[Epoch 56 Batch 90/173] avg loss 1.48729e-06, throughput 5.98429K wps
[Epoch 56 Batch 120/173] avg loss 3.67018e-06, throughput 5.96894K wps
[Epoch 56 Batch 150/173] avg loss 1.7156e-06, throughput 5.98637K wps
Begin Testing...
[Epoch 56] train avg loss 2.15941e-06, test acc 0.7427, test avg loss 1.658, throughput 6.00896K wps
[Epoch 57 Batch 30/173] avg loss 1.72715e-06, throughput 6.13063K wps
[Epoch 57 Batch 60/173] avg loss 1.56313e-06, throughput 5.99081K wps
[Epoch 57 Batch 90/173] avg loss 1.24814e-06, throughput 5.98388K wps
[Epoch 57 Batch 120/173] avg loss 1.72987e-06, throughput 5.98324K wps
[Epoch 57 Batch 150/173] avg loss 2.10875e-06, throughput 5.9786K wps
Begin Testing...
[Epoch 57] train avg loss 1.9084e-06, test acc 0.7427, test avg loss 1.68249, throughput 6.00838K wps
[Epoch 58 Batch 30/173] avg loss 1.06136e-06, throughput 6.07838K wps
[Epoch 58 Batch 60/173] avg loss 2.44536e-06, throughput 5.99021K wps
[Epoch 58 Batch 90/173] avg loss 1.86275e-06, throughput 5.99688K wps
[Epoch 58 Batch 120/173] avg loss 1.41498e-06, throughput 5.99125K wps
[Epoch 58 Batch 150/173] avg loss 1.15219e-06, throughput 5.98519K wps
Begin Testing...
[Epoch 58] train avg loss 1.55186e-06, test acc 0.7417, test avg loss 1.70434, throughput 6.00635K wps
[Epoch 59 Batch 30/173] avg loss 1.10201e-06, throughput 6.13697K wps
[Epoch 59 Batch 60/173] avg loss 1.3027e-06, throughput 5.98766K wps
[Epoch 59 Batch 90/173] avg loss 1.04982e-06, throughput 5.99394K wps
[Epoch 59 Batch 120/173] avg loss 1.43354e-06, throughput 5.99705K wps
[Epoch 59 Batch 150/173] avg loss 1.42297e-06, throughput 5.99276K wps
Begin Testing...
[Epoch 59] train avg loss 1.24404e-06, test acc 0.7406, test avg loss 1.71494, throughput 6.01712K wps
Test loss 0.491276, test acc 0.7617
Total time cost 358.21s
9596 1066
[Epoch 0 Batch 30/173] avg loss 0.013866, throughput 5.79108K wps
[Epoch 0 Batch 60/173] avg loss 0.0138419, throughput 5.9924K wps
[Epoch 0 Batch 90/173] avg loss 0.013809, throughput 5.98266K wps
[Epoch 0 Batch 120/173] avg loss 0.0137772, throughput 5.97531K wps
[Epoch 0 Batch 150/173] avg loss 0.0137503, throughput 5.97549K wps
Begin Testing...
[Epoch 0] train avg loss 0.0138167, test acc 0.6271, test avg loss 0.686751, throughput 5.94784K wps
Observed Improvement.
Begin Testing...
[Epoch 1 Batch 30/173] avg loss 0.0134659, throughput 6.15192K wps
[Epoch 1 Batch 60/173] avg loss 0.0134662, throughput 5.9926K wps
[Epoch 1 Batch 90/173] avg loss 0.0134183, throughput 5.98578K wps
[Epoch 1 Batch 120/173] avg loss 0.0133427, throughput 5.98958K wps
[Epoch 1 Batch 150/173] avg loss 0.0133146, throughput 5.97682K wps
Begin Testing...
[Epoch 1] train avg loss 0.0134032, test acc 0.6625, test avg loss 0.673258, throughput 6.01366K wps
Observed Improvement.
Begin Testing...
[Epoch 2 Batch 30/173] avg loss 0.0128958, throughput 6.12729K wps
[Epoch 2 Batch 60/173] avg loss 0.0127597, throughput 5.97963K wps
[Epoch 2 Batch 90/173] avg loss 0.0126978, throughput 5.97569K wps
[Epoch 2 Batch 120/173] avg loss 0.0125782, throughput 5.98475K wps
[Epoch 2 Batch 150/173] avg loss 0.0123819, throughput 5.98974K wps
Begin Testing...
[Epoch 2] train avg loss 0.0126192, test acc 0.7177, test avg loss 0.6349, throughput 6.00907K wps
Observed Improvement.
Begin Testing...
[Epoch 3 Batch 30/173] avg loss 0.011457, throughput 6.12667K wps
[Epoch 3 Batch 60/173] avg loss 0.0113745, throughput 5.97278K wps
[Epoch 3 Batch 90/173] avg loss 0.0111195, throughput 5.97989K wps
[Epoch 3 Batch 120/173] avg loss 0.0107874, throughput 5.97578K wps
[Epoch 3 Batch 150/173] avg loss 0.0104942, throughput 5.97918K wps
Begin Testing...
[Epoch 3] train avg loss 0.0109449, test acc 0.7490, test avg loss 0.563821, throughput 6.00288K wps
Observed Improvement.
Begin Testing...
[Epoch 4 Batch 30/173] avg loss 0.00900936, throughput 6.1241K wps
[Epoch 4 Batch 60/173] avg loss 0.00860808, throughput 6.00121K wps
[Epoch 4 Batch 90/173] avg loss 0.0085382, throughput 5.97513K wps
[Epoch 4 Batch 120/173] avg loss 0.00832825, throughput 5.986K wps
[Epoch 4 Batch 150/173] avg loss 0.00805825, throughput 5.99681K wps
Begin Testing...
[Epoch 4] train avg loss 0.00847002, test acc 0.7771, test avg loss 0.495598, throughput 6.01409K wps
Observed Improvement.
Begin Testing...
[Epoch 5 Batch 30/173] avg loss 0.00614454, throughput 6.13975K wps
[Epoch 5 Batch 60/173] avg loss 0.00619716, throughput 5.98902K wps
[Epoch 5 Batch 90/173] avg loss 0.00624855, throughput 5.98393K wps
[Epoch 5 Batch 120/173] avg loss 0.00623449, throughput 5.99341K wps
[Epoch 5 Batch 150/173] avg loss 0.00598117, throughput 5.9815K wps
Begin Testing...
[Epoch 5] train avg loss 0.00612456, test acc 0.7750, test avg loss 0.471689, throughput 6.01271K wps
[Epoch 6 Batch 30/173] avg loss 0.00444746, throughput 6.12954K wps
[Epoch 6 Batch 60/173] avg loss 0.00438795, throughput 5.98331K wps
[Epoch 6 Batch 90/173] avg loss 0.00424073, throughput 5.97317K wps
[Epoch 6 Batch 120/173] avg loss 0.00437794, throughput 5.97864K wps
[Epoch 6 Batch 150/173] avg loss 0.00411979, throughput 5.98063K wps
Begin Testing...
[Epoch 6] train avg loss 0.00430728, test acc 0.7635, test avg loss 0.483323, throughput 6.00511K wps
[Epoch 7 Batch 30/173] avg loss 0.00317482, throughput 6.12967K wps
[Epoch 7 Batch 60/173] avg loss 0.00288571, throughput 5.98853K wps
[Epoch 7 Batch 90/173] avg loss 0.00296597, throughput 5.99607K wps
[Epoch 7 Batch 120/173] avg loss 0.00282836, throughput 5.98706K wps
[Epoch 7 Batch 150/173] avg loss 0.00321006, throughput 5.97625K wps
Begin Testing...
[Epoch 7] train avg loss 0.00301089, test acc 0.7615, test avg loss 0.516595, throughput 6.01005K wps
[Epoch 8 Batch 30/173] avg loss 0.0021655, throughput 6.14781K wps
[Epoch 8 Batch 60/173] avg loss 0.00220101, throughput 5.9853K wps
[Epoch 8 Batch 90/173] avg loss 0.00217394, throughput 5.99573K wps
[Epoch 8 Batch 120/173] avg loss 0.00212772, throughput 5.99456K wps
[Epoch 8 Batch 150/173] avg loss 0.0020491, throughput 5.98919K wps
Begin Testing...
[Epoch 8] train avg loss 0.00213599, test acc 0.7667, test avg loss 0.551343, throughput 6.01637K wps
[Epoch 9 Batch 30/173] avg loss 0.00142699, throughput 6.13662K wps
[Epoch 9 Batch 60/173] avg loss 0.00147434, throughput 5.97697K wps
[Epoch 9 Batch 90/173] avg loss 0.00141455, throughput 5.99079K wps
[Epoch 9 Batch 120/173] avg loss 0.00150817, throughput 5.99397K wps
[Epoch 9 Batch 150/173] avg loss 0.00155068, throughput 5.99419K wps
Begin Testing...
[Epoch 9] train avg loss 0.00149231, test acc 0.7562, test avg loss 0.594986, throughput 6.01624K wps
[Epoch 10 Batch 30/173] avg loss 0.00109088, throughput 6.14066K wps
[Epoch 10 Batch 60/173] avg loss 0.00108974, throughput 5.98409K wps
[Epoch 10 Batch 90/173] avg loss 0.00109361, throughput 5.97321K wps
[Epoch 10 Batch 120/173] avg loss 0.00101491, throughput 5.98613K wps
[Epoch 10 Batch 150/173] avg loss 0.0011565, throughput 5.97897K wps
Begin Testing...
[Epoch 10] train avg loss 0.00107526, test acc 0.7562, test avg loss 0.632034, throughput 6.00908K wps
[Epoch 11 Batch 30/173] avg loss 0.000859169, throughput 6.13717K wps
[Epoch 11 Batch 60/173] avg loss 0.000769313, throughput 5.99627K wps
[Epoch 11 Batch 90/173] avg loss 0.000756585, throughput 6.0028K wps
[Epoch 11 Batch 120/173] avg loss 0.00076884, throughput 5.99728K wps
[Epoch 11 Batch 150/173] avg loss 0.000811032, throughput 5.98765K wps
Begin Testing...
[Epoch 11] train avg loss 0.000795718, test acc 0.7500, test avg loss 0.672038, throughput 6.02102K wps
[Epoch 12 Batch 30/173] avg loss 0.000571135, throughput 6.1359K wps
[Epoch 12 Batch 60/173] avg loss 0.000557391, throughput 5.97453K wps
[Epoch 12 Batch 90/173] avg loss 0.000606707, throughput 5.99308K wps
[Epoch 12 Batch 120/173] avg loss 0.000598845, throughput 5.98218K wps
[Epoch 12 Batch 150/173] avg loss 0.000588169, throughput 5.97923K wps
Begin Testing...
[Epoch 12] train avg loss 0.000586398, test acc 0.7490, test avg loss 0.717663, throughput 6.00797K wps
[Epoch 13 Batch 30/173] avg loss 0.000426139, throughput 6.13249K wps
[Epoch 13 Batch 60/173] avg loss 0.000428648, throughput 5.98935K wps
[Epoch 13 Batch 90/173] avg loss 0.000503411, throughput 5.99404K wps
[Epoch 13 Batch 120/173] avg loss 0.000382585, throughput 6.00158K wps
[Epoch 13 Batch 150/173] avg loss 0.000485365, throughput 5.97717K wps
Begin Testing...
[Epoch 13] train avg loss 0.000454356, test acc 0.7417, test avg loss 0.751343, throughput 6.0152K wps
[Epoch 14 Batch 30/173] avg loss 0.000349986, throughput 6.13399K wps
[Epoch 14 Batch 60/173] avg loss 0.000333229, throughput 5.99093K wps
[Epoch 14 Batch 90/173] avg loss 0.000317865, throughput 5.98905K wps
[Epoch 14 Batch 120/173] avg loss 0.000324036, throughput 5.99472K wps
[Epoch 14 Batch 150/173] avg loss 0.000343027, throughput 5.98862K wps
Begin Testing...
[Epoch 14] train avg loss 0.000342141, test acc 0.7448, test avg loss 0.78637, throughput 6.01746K wps
[Epoch 15 Batch 30/173] avg loss 0.000247519, throughput 6.12728K wps
[Epoch 15 Batch 60/173] avg loss 0.000283147, throughput 5.99331K wps
[Epoch 15 Batch 90/173] avg loss 0.000266069, throughput 5.98606K wps
[Epoch 15 Batch 120/173] avg loss 0.000279376, throughput 5.97998K wps
[Epoch 15 Batch 150/173] avg loss 0.000266317, throughput 5.98952K wps
Begin Testing...
[Epoch 15] train avg loss 0.00027928, test acc 0.7469, test avg loss 0.823986, throughput 6.01298K wps
[Epoch 16 Batch 30/173] avg loss 0.000202613, throughput 6.12621K wps
[Epoch 16 Batch 60/173] avg loss 0.000205684, throughput 5.98459K wps
[Epoch 16 Batch 90/173] avg loss 0.000243968, throughput 6.00198K wps
[Epoch 16 Batch 120/173] avg loss 0.00021957, throughput 5.99556K wps
[Epoch 16 Batch 150/173] avg loss 0.000191744, throughput 5.98556K wps
Begin Testing...
[Epoch 16] train avg loss 0.000214337, test acc 0.7438, test avg loss 0.85372, throughput 6.01415K wps
[Epoch 17 Batch 30/173] avg loss 0.000178018, throughput 6.12869K wps
[Epoch 17 Batch 60/173] avg loss 0.000165186, throughput 5.97974K wps
[Epoch 17 Batch 90/173] avg loss 0.00018133, throughput 5.99307K wps
[Epoch 17 Batch 120/173] avg loss 0.000181141, throughput 5.99043K wps
[Epoch 17 Batch 150/173] avg loss 0.00017923, throughput 5.98466K wps
Begin Testing...
[Epoch 17] train avg loss 0.00017829, test acc 0.7427, test avg loss 0.88777, throughput 6.01413K wps
[Epoch 18 Batch 30/173] avg loss 0.000135129, throughput 6.15037K wps
[Epoch 18 Batch 60/173] avg loss 0.000143817, throughput 5.9924K wps
[Epoch 18 Batch 90/173] avg loss 0.000146821, throughput 5.99405K wps
[Epoch 18 Batch 120/173] avg loss 0.000134405, throughput 5.99721K wps
[Epoch 18 Batch 150/173] avg loss 0.000147072, throughput 5.97228K wps
Begin Testing...
[Epoch 18] train avg loss 0.000140366, test acc 0.7469, test avg loss 0.909757, throughput 6.01553K wps
[Epoch 19 Batch 30/173] avg loss 0.000125408, throughput 6.14552K wps
[Epoch 19 Batch 60/173] avg loss 0.000113515, throughput 6.00002K wps
[Epoch 19 Batch 90/173] avg loss 0.000113289, throughput 5.99058K wps
[Epoch 19 Batch 120/173] avg loss 0.000116137, throughput 5.98774K wps
[Epoch 19 Batch 150/173] avg loss 0.000100628, throughput 5.98864K wps
Begin Testing...
[Epoch 19] train avg loss 0.000113308, test acc 0.7448, test avg loss 0.941091, throughput 6.01978K wps
[Epoch 20 Batch 30/173] avg loss 8.25543e-05, throughput 6.13914K wps
[Epoch 20 Batch 60/173] avg loss 8.8011e-05, throughput 5.98949K wps
[Epoch 20 Batch 90/173] avg loss 9.91262e-05, throughput 6.00172K wps
[Epoch 20 Batch 120/173] avg loss 0.000104192, throughput 5.99892K wps
[Epoch 20 Batch 150/173] avg loss 0.000108543, throughput 5.98868K wps
Begin Testing...
[Epoch 20] train avg loss 9.73045e-05, test acc 0.7396, test avg loss 0.97229, throughput 6.01692K wps
[Epoch 21 Batch 30/173] avg loss 7.33208e-05, throughput 6.15096K wps
[Epoch 21 Batch 60/173] avg loss 0.00010104, throughput 5.99504K wps
[Epoch 21 Batch 90/173] avg loss 9.10097e-05, throughput 6.00423K wps
[Epoch 21 Batch 120/173] avg loss 7.45659e-05, throughput 6.00457K wps
[Epoch 21 Batch 150/173] avg loss 7.92766e-05, throughput 5.99896K wps
Begin Testing...
[Epoch 21] train avg loss 8.27844e-05, test acc 0.7406, test avg loss 1.00134, throughput 6.02669K wps
[Epoch 22 Batch 30/173] avg loss 7.53861e-05, throughput 6.13897K wps
[Epoch 22 Batch 60/173] avg loss 7.3093e-05, throughput 6.00331K wps
[Epoch 22 Batch 90/173] avg loss 8.50138e-05, throughput 5.98578K wps
[Epoch 22 Batch 120/173] avg loss 6.97952e-05, throughput 5.97678K wps
[Epoch 22 Batch 150/173] avg loss 7.11499e-05, throughput 5.98775K wps
Begin Testing...
[Epoch 22] train avg loss 7.72798e-05, test acc 0.7417, test avg loss 1.02588, throughput 6.01616K wps
[Epoch 23 Batch 30/173] avg loss 6.81157e-05, throughput 6.13503K wps
[Epoch 23 Batch 60/173] avg loss 7.2502e-05, throughput 5.97672K wps
[Epoch 23 Batch 90/173] avg loss 5.36027e-05, throughput 5.98675K wps
[Epoch 23 Batch 120/173] avg loss 6.22903e-05, throughput 5.98188K wps
[Epoch 23 Batch 150/173] avg loss 6.11611e-05, throughput 5.98565K wps
Begin Testing...
[Epoch 23] train avg loss 6.40787e-05, test acc 0.7396, test avg loss 1.04489, throughput 6.0101K wps
[Epoch 24 Batch 30/173] avg loss 4.59431e-05, throughput 6.12587K wps
[Epoch 24 Batch 60/173] avg loss 6.004e-05, throughput 5.99853K wps
[Epoch 24 Batch 90/173] avg loss 6.27041e-05, throughput 5.97766K wps
[Epoch 24 Batch 120/173] avg loss 4.58117e-05, throughput 5.99368K wps
[Epoch 24 Batch 150/173] avg loss 5.19312e-05, throughput 6.00062K wps
Begin Testing...
[Epoch 24] train avg loss 5.32462e-05, test acc 0.7417, test avg loss 1.06232, throughput 6.01681K wps
[Epoch 25 Batch 30/173] avg loss 4.63102e-05, throughput 6.13779K wps
[Epoch 25 Batch 60/173] avg loss 4.01835e-05, throughput 5.98684K wps
[Epoch 25 Batch 90/173] avg loss 3.83209e-05, throughput 5.97746K wps
[Epoch 25 Batch 120/173] avg loss 3.90822e-05, throughput 5.99571K wps
[Epoch 25 Batch 150/173] avg loss 4.84239e-05, throughput 6.00349K wps
Begin Testing...
[Epoch 25] train avg loss 4.36608e-05, test acc 0.7417, test avg loss 1.09248, throughput 6.01633K wps
[Epoch 26 Batch 30/173] avg loss 3.66046e-05, throughput 6.12622K wps
[Epoch 26 Batch 60/173] avg loss 3.59365e-05, throughput 5.98931K wps
[Epoch 26 Batch 90/173] avg loss 3.48311e-05, throughput 6.00157K wps
[Epoch 26 Batch 120/173] avg loss 4.76445e-05, throughput 5.98072K wps
[Epoch 26 Batch 150/173] avg loss 4.41632e-05, throughput 5.98102K wps
Begin Testing...
[Epoch 26] train avg loss 3.91267e-05, test acc 0.7406, test avg loss 1.12163, throughput 6.01291K wps
[Epoch 27 Batch 30/173] avg loss 3.30606e-05, throughput 6.13614K wps
[Epoch 27 Batch 60/173] avg loss 3.779e-05, throughput 5.97988K wps
[Epoch 27 Batch 90/173] avg loss 3.74905e-05, throughput 5.9755K wps
[Epoch 27 Batch 120/173] avg loss 3.10485e-05, throughput 5.97855K wps
[Epoch 27 Batch 150/173] avg loss 4.25601e-05, throughput 5.98219K wps
Begin Testing...
[Epoch 27] train avg loss 3.54043e-05, test acc 0.7406, test avg loss 1.14397, throughput 6.005K wps
[Epoch 28 Batch 30/173] avg loss 3.15953e-05, throughput 6.08458K wps
[Epoch 28 Batch 60/173] avg loss 3.48197e-05, throughput 6.01714K wps
[Epoch 28 Batch 90/173] avg loss 3.01443e-05, throughput 5.99223K wps
[Epoch 28 Batch 120/173] avg loss 2.5119e-05, throughput 5.99242K wps
[Epoch 28 Batch 150/173] avg loss 3.34809e-05, throughput 5.98632K wps
Begin Testing...
[Epoch 28] train avg loss 3.02547e-05, test acc 0.7396, test avg loss 1.16631, throughput 6.01094K wps
[Epoch 29 Batch 30/173] avg loss 2.44594e-05, throughput 6.14153K wps
[Epoch 29 Batch 60/173] avg loss 1.95538e-05, throughput 5.98258K wps
[Epoch 29 Batch 90/173] avg loss 3.01323e-05, throughput 5.96668K wps
[Epoch 29 Batch 120/173] avg loss 2.84569e-05, throughput 5.98789K wps
[Epoch 29 Batch 150/173] avg loss 2.9739e-05, throughput 5.99742K wps
Begin Testing...
[Epoch 29] train avg loss 2.65978e-05, test acc 0.7385, test avg loss 1.1835, throughput 6.01137K wps
[Epoch 30 Batch 30/173] avg loss 3.12519e-05, throughput 6.12988K wps
[Epoch 30 Batch 60/173] avg loss 2.31474e-05, throughput 5.99902K wps
[Epoch 30 Batch 90/173] avg loss 3.86383e-05, throughput 5.99809K wps
[Epoch 30 Batch 120/173] avg loss 3.68533e-05, throughput 5.97981K wps
[Epoch 30 Batch 150/173] avg loss 2.09946e-05, throughput 5.9802K wps
Begin Testing...
[Epoch 30] train avg loss 2.94825e-05, test acc 0.7375, test avg loss 1.20245, throughput 6.01295K wps
[Epoch 31 Batch 30/173] avg loss 2.30276e-05, throughput 6.12136K wps
[Epoch 31 Batch 60/173] avg loss 2.78341e-05, throughput 5.97896K wps
[Epoch 31 Batch 90/173] avg loss 1.77675e-05, throughput 6.00368K wps
[Epoch 31 Batch 120/173] avg loss 3.94037e-05, throughput 5.98757K wps
[Epoch 31 Batch 150/173] avg loss 1.77469e-05, throughput 5.97608K wps
Begin Testing...
[Epoch 31] train avg loss 2.40484e-05, test acc 0.7385, test avg loss 1.22563, throughput 6.01116K wps
[Epoch 32 Batch 30/173] avg loss 1.52178e-05, throughput 6.12411K wps
[Epoch 32 Batch 60/173] avg loss 1.58934e-05, throughput 5.98691K wps
[Epoch 32 Batch 90/173] avg loss 1.44975e-05, throughput 5.98651K wps
[Epoch 32 Batch 120/173] avg loss 2.72551e-05, throughput 5.97986K wps
[Epoch 32 Batch 150/173] avg loss 1.8106e-05, throughput 5.97511K wps
Begin Testing...
[Epoch 32] train avg loss 2.00958e-05, test acc 0.7375, test avg loss 1.25628, throughput 6.00895K wps
[Epoch 33 Batch 30/173] avg loss 1.51392e-05, throughput 6.13703K wps
[Epoch 33 Batch 60/173] avg loss 1.70606e-05, throughput 5.97828K wps
[Epoch 33 Batch 90/173] avg loss 2.15204e-05, throughput 5.97453K wps
[Epoch 33 Batch 120/173] avg loss 1.60546e-05, throughput 5.99385K wps
[Epoch 33 Batch 150/173] avg loss 2.04311e-05, throughput 5.97753K wps
Begin Testing...
[Epoch 33] train avg loss 1.83942e-05, test acc 0.7385, test avg loss 1.27473, throughput 6.00927K wps
[Epoch 34 Batch 30/173] avg loss 1.20254e-05, throughput 6.12351K wps
[Epoch 34 Batch 60/173] avg loss 2.37896e-05, throughput 5.97481K wps
[Epoch 34 Batch 90/173] avg loss 1.54525e-05, throughput 5.97822K wps
[Epoch 34 Batch 120/173] avg loss 1.38208e-05, throughput 5.98355K wps
[Epoch 34 Batch 150/173] avg loss 1.45717e-05, throughput 5.99805K wps
Begin Testing...
[Epoch 34] train avg loss 1.5756e-05, test acc 0.7375, test avg loss 1.29703, throughput 6.00976K wps
[Epoch 35 Batch 30/173] avg loss 1.10586e-05, throughput 6.13463K wps
[Epoch 35 Batch 60/173] avg loss 1.42761e-05, throughput 5.98731K wps
[Epoch 35 Batch 90/173] avg loss 1.46414e-05, throughput 5.98179K wps
[Epoch 35 Batch 120/173] avg loss 1.29016e-05, throughput 5.99126K wps
[Epoch 35 Batch 150/173] avg loss 1.07773e-05, throughput 6.0011K wps
Begin Testing...
[Epoch 35] train avg loss 1.37288e-05, test acc 0.7385, test avg loss 1.32086, throughput 6.01632K wps
[Epoch 36 Batch 30/173] avg loss 1.18224e-05, throughput 6.15156K wps
[Epoch 36 Batch 60/173] avg loss 1.10023e-05, throughput 5.98803K wps
[Epoch 36 Batch 90/173] avg loss 1.08633e-05, throughput 5.98336K wps
[Epoch 36 Batch 120/173] avg loss 1.05314e-05, throughput 5.98984K wps
[Epoch 36 Batch 150/173] avg loss 1.37774e-05, throughput 5.98343K wps
Begin Testing...
[Epoch 36] train avg loss 1.18209e-05, test acc 0.7417, test avg loss 1.34399, throughput 6.01533K wps
[Epoch 37 Batch 30/173] avg loss 8.07511e-06, throughput 6.12994K wps
[Epoch 37 Batch 60/173] avg loss 8.31943e-06, throughput 5.98788K wps
[Epoch 37 Batch 90/173] avg loss 1.16247e-05, throughput 5.99152K wps
[Epoch 37 Batch 120/173] avg loss 1.12137e-05, throughput 5.98972K wps
[Epoch 37 Batch 150/173] avg loss 9.15796e-06, throughput 5.97254K wps
Begin Testing...
[Epoch 37] train avg loss 9.8797e-06, test acc 0.7396, test avg loss 1.36414, throughput 6.01121K wps
[Epoch 38 Batch 30/173] avg loss 7.67866e-06, throughput 6.14712K wps
[Epoch 38 Batch 60/173] avg loss 7.33512e-06, throughput 5.97299K wps
[Epoch 38 Batch 90/173] avg loss 1.00215e-05, throughput 5.98829K wps
[Epoch 38 Batch 120/173] avg loss 9.74335e-06, throughput 5.98976K wps
[Epoch 38 Batch 150/173] avg loss 7.94125e-06, throughput 5.99137K wps
Begin Testing...
[Epoch 38] train avg loss 9.89334e-06, test acc 0.7354, test avg loss 1.3839, throughput 6.01396K wps
[Epoch 39 Batch 30/173] avg loss 8.26463e-06, throughput 6.13205K wps
[Epoch 39 Batch 60/173] avg loss 8.04404e-06, throughput 5.98528K wps
[Epoch 39 Batch 90/173] avg loss 8.46315e-06, throughput 5.98004K wps
[Epoch 39 Batch 120/173] avg loss 8.20577e-06, throughput 6.0059K wps
[Epoch 39 Batch 150/173] avg loss 7.05331e-06, throughput 6.00286K wps
Begin Testing...
[Epoch 39] train avg loss 8.27501e-06, test acc 0.7354, test avg loss 1.40806, throughput 6.01959K wps
[Epoch 40 Batch 30/173] avg loss 6.85211e-06, throughput 6.1433K wps
[Epoch 40 Batch 60/173] avg loss 7.47753e-06, throughput 5.98321K wps
[Epoch 40 Batch 90/173] avg loss 7.10128e-06, throughput 5.98656K wps
[Epoch 40 Batch 120/173] avg loss 7.68818e-06, throughput 5.99293K wps
[Epoch 40 Batch 150/173] avg loss 6.43959e-06, throughput 5.98491K wps
Begin Testing...
[Epoch 40] train avg loss 7.31394e-06, test acc 0.7375, test avg loss 1.4308, throughput 6.0147K wps
[Epoch 41 Batch 30/173] avg loss 7.70442e-06, throughput 6.13148K wps
[Epoch 41 Batch 60/173] avg loss 7.55692e-06, throughput 5.98146K wps
[Epoch 41 Batch 90/173] avg loss 5.03413e-06, throughput 5.98891K wps
[Epoch 41 Batch 120/173] avg loss 2.0554e-05, throughput 6.00105K wps
[Epoch 41 Batch 150/173] avg loss 1.04813e-05, throughput 5.99232K wps
Begin Testing...
[Epoch 41] train avg loss 1.022e-05, test acc 0.7323, test avg loss 1.4521, throughput 6.01637K wps
[Epoch 42 Batch 30/173] avg loss 6.78142e-06, throughput 6.1309K wps
[Epoch 42 Batch 60/173] avg loss 7.49941e-06, throughput 5.97045K wps
[Epoch 42 Batch 90/173] avg loss 6.19729e-06, throughput 6.00282K wps
[Epoch 42 Batch 120/173] avg loss 6.50994e-06, throughput 6.00118K wps
[Epoch 42 Batch 150/173] avg loss 5.19848e-06, throughput 5.99235K wps
Begin Testing...
[Epoch 42] train avg loss 6.67117e-06, test acc 0.7333, test avg loss 1.46699, throughput 6.01804K wps
[Epoch 43 Batch 30/173] avg loss 4.03832e-06, throughput 6.13646K wps
[Epoch 43 Batch 60/173] avg loss 5.43114e-06, throughput 5.97592K wps
[Epoch 43 Batch 90/173] avg loss 6.73034e-06, throughput 5.9848K wps
[Epoch 43 Batch 120/173] avg loss 5.05032e-06, throughput 5.99527K wps
[Epoch 43 Batch 150/173] avg loss 4.33742e-06, throughput 5.99569K wps
Begin Testing...
[Epoch 43] train avg loss 6.20079e-06, test acc 0.7323, test avg loss 1.4876, throughput 6.01392K wps
[Epoch 44 Batch 30/173] avg loss 4.83215e-06, throughput 6.11618K wps
[Epoch 44 Batch 60/173] avg loss 6.64725e-06, throughput 5.97564K wps
[Epoch 44 Batch 90/173] avg loss 4.60396e-06, throughput 5.97752K wps
[Epoch 44 Batch 120/173] avg loss 4.3874e-06, throughput 5.97628K wps
[Epoch 44 Batch 150/173] avg loss 5.40747e-06, throughput 5.99798K wps
Begin Testing...
[Epoch 44] train avg loss 5.3712e-06, test acc 0.7365, test avg loss 1.5163, throughput 6.00693K wps
[Epoch 45 Batch 30/173] avg loss 4.56962e-06, throughput 6.12458K wps
[Epoch 45 Batch 60/173] avg loss 7.25676e-06, throughput 5.97835K wps
[Epoch 45 Batch 90/173] avg loss 4.51988e-06, throughput 5.98472K wps
[Epoch 45 Batch 120/173] avg loss 4.64161e-06, throughput 5.98304K wps
[Epoch 45 Batch 150/173] avg loss 4.83884e-06, throughput 5.99656K wps
Begin Testing...
[Epoch 45] train avg loss 5.10904e-06, test acc 0.7406, test avg loss 1.53299, throughput 6.01109K wps
[Epoch 46 Batch 30/173] avg loss 3.37518e-06, throughput 6.13963K wps
[Epoch 46 Batch 60/173] avg loss 3.71905e-06, throughput 5.99521K wps
[Epoch 46 Batch 90/173] avg loss 3.25583e-06, throughput 5.98835K wps
[Epoch 46 Batch 120/173] avg loss 4.21334e-06, throughput 5.9819K wps
[Epoch 46 Batch 150/173] avg loss 3.77002e-06, throughput 5.99256K wps
Begin Testing...
[Epoch 46] train avg loss 3.72683e-06, test acc 0.7417, test avg loss 1.54194, throughput 6.01669K wps
[Epoch 47 Batch 30/173] avg loss 5.18797e-06, throughput 6.12553K wps
[Epoch 47 Batch 60/173] avg loss 4.43217e-06, throughput 5.98591K wps
[Epoch 47 Batch 90/173] avg loss 4.40901e-06, throughput 5.98165K wps
[Epoch 47 Batch 120/173] avg loss 4.62837e-06, throughput 5.98923K wps
[Epoch 47 Batch 150/173] avg loss 3.35835e-06, throughput 5.97845K wps
Begin Testing...
[Epoch 47] train avg loss 4.5563e-06, test acc 0.7406, test avg loss 1.55495, throughput 6.00662K wps
[Epoch 48 Batch 30/173] avg loss 3.16396e-06, throughput 6.12691K wps
[Epoch 48 Batch 60/173] avg loss 2.67815e-06, throughput 6.00356K wps
[Epoch 48 Batch 90/173] avg loss 3.58998e-06, throughput 6.00397K wps
[Epoch 48 Batch 120/173] avg loss 2.8887e-06, throughput 6.00163K wps
[Epoch 48 Batch 150/173] avg loss 3.15171e-06, throughput 5.98685K wps
Begin Testing...
[Epoch 48] train avg loss 3.19803e-06, test acc 0.7344, test avg loss 1.56904, throughput 6.02179K wps
[Epoch 49 Batch 30/173] avg loss 2.83168e-06, throughput 6.13934K wps
[Epoch 49 Batch 60/173] avg loss 2.60087e-06, throughput 6.00404K wps
[Epoch 49 Batch 90/173] avg loss 2.77223e-06, throughput 5.98239K wps
[Epoch 49 Batch 120/173] avg loss 3.8156e-06, throughput 5.98664K wps
[Epoch 49 Batch 150/173] avg loss 1.88099e-06, throughput 5.99471K wps
Begin Testing...
[Epoch 49] train avg loss 2.82048e-06, test acc 0.7354, test avg loss 1.59004, throughput 6.01716K wps
[Epoch 50 Batch 30/173] avg loss 1.9184e-06, throughput 6.1354K wps
[Epoch 50 Batch 60/173] avg loss 2.45422e-06, throughput 6.00376K wps
[Epoch 50 Batch 90/173] avg loss 3.12771e-06, throughput 5.98001K wps
[Epoch 50 Batch 120/173] avg loss 1.81983e-06, throughput 5.97428K wps
[Epoch 50 Batch 150/173] avg loss 2.26386e-06, throughput 5.96787K wps
Begin Testing...
[Epoch 50] train avg loss 2.42408e-06, test acc 0.7365, test avg loss 1.59773, throughput 6.00647K wps
[Epoch 51 Batch 30/173] avg loss 1.90887e-06, throughput 6.13143K wps
[Epoch 51 Batch 60/173] avg loss 2.22563e-06, throughput 5.97954K wps
[Epoch 51 Batch 90/173] avg loss 1.9402e-06, throughput 5.98356K wps
[Epoch 51 Batch 120/173] avg loss 1.84529e-06, throughput 5.99123K wps
[Epoch 51 Batch 150/173] avg loss 2.09761e-06, throughput 5.99388K wps
Begin Testing...
[Epoch 51] train avg loss 2.03398e-06, test acc 0.7354, test avg loss 1.61495, throughput 6.01177K wps
[Epoch 52 Batch 30/173] avg loss 2.12218e-06, throughput 6.1297K wps
[Epoch 52 Batch 60/173] avg loss 2.85592e-06, throughput 5.98115K wps
[Epoch 52 Batch 90/173] avg loss 2.55497e-06, throughput 5.98002K wps
[Epoch 52 Batch 120/173] avg loss 2.19346e-06, throughput 5.98066K wps
[Epoch 52 Batch 150/173] avg loss 2.33371e-06, throughput 5.96943K wps
Begin Testing...
[Epoch 52] train avg loss 2.43034e-06, test acc 0.7354, test avg loss 1.63356, throughput 6.00476K wps
[Epoch 53 Batch 30/173] avg loss 2.88707e-06, throughput 6.13175K wps
[Epoch 53 Batch 60/173] avg loss 2.48417e-06, throughput 5.98046K wps
[Epoch 53 Batch 90/173] avg loss 2.13069e-06, throughput 5.97962K wps
[Epoch 53 Batch 120/173] avg loss 2.55335e-06, throughput 5.98776K wps
[Epoch 53 Batch 150/173] avg loss 2.31313e-06, throughput 5.98332K wps
Begin Testing...
[Epoch 53] train avg loss 2.46226e-06, test acc 0.7323, test avg loss 1.66624, throughput 6.00859K wps
[Epoch 54 Batch 30/173] avg loss 2.25267e-06, throughput 6.13294K wps
[Epoch 54 Batch 60/173] avg loss 2.56347e-06, throughput 5.9956K wps
[Epoch 54 Batch 90/173] avg loss 2.15803e-06, throughput 5.99208K wps
[Epoch 54 Batch 120/173] avg loss 2.58851e-06, throughput 5.99588K wps
[Epoch 54 Batch 150/173] avg loss 1.64345e-06, throughput 5.99214K wps
Begin Testing...
[Epoch 54] train avg loss 2.25062e-06, test acc 0.7271, test avg loss 1.66936, throughput 6.01604K wps
[Epoch 55 Batch 30/173] avg loss 1.55421e-06, throughput 6.14892K wps
[Epoch 55 Batch 60/173] avg loss 1.34238e-06, throughput 5.97864K wps
[Epoch 55 Batch 90/173] avg loss 2.32142e-06, throughput 5.98998K wps
[Epoch 55 Batch 120/173] avg loss 1.24041e-06, throughput 5.97872K wps
[Epoch 55 Batch 150/173] avg loss 2.06561e-06, throughput 5.97202K wps
Begin Testing...
[Epoch 55] train avg loss 1.8405e-06, test acc 0.7292, test avg loss 1.69749, throughput 6.01074K wps
[Epoch 56 Batch 30/173] avg loss 1.75933e-06, throughput 6.12827K wps
[Epoch 56 Batch 60/173] avg loss 1.16919e-06, throughput 5.99531K wps
[Epoch 56 Batch 90/173] avg loss 1.67521e-06, throughput 5.99248K wps
[Epoch 56 Batch 120/173] avg loss 1.63068e-06, throughput 5.97803K wps
[Epoch 56 Batch 150/173] avg loss 2.01596e-06, throughput 5.99222K wps
Begin Testing...
[Epoch 56] train avg loss 1.68144e-06, test acc 0.7302, test avg loss 1.71444, throughput 6.01275K wps
[Epoch 57 Batch 30/173] avg loss 1.10946e-06, throughput 6.13536K wps
[Epoch 57 Batch 60/173] avg loss 6.47543e-06, throughput 5.9901K wps
[Epoch 57 Batch 90/173] avg loss 3.26799e-06, throughput 5.97917K wps
[Epoch 57 Batch 120/173] avg loss 2.098e-06, throughput 5.98409K wps
[Epoch 57 Batch 150/173] avg loss 2.72815e-06, throughput 5.98521K wps
Begin Testing...
[Epoch 57] train avg loss 2.86954e-06, test acc 0.7385, test avg loss 1.73796, throughput 6.0128K wps
[Epoch 58 Batch 30/173] avg loss 1.47272e-06, throughput 6.11713K wps
[Epoch 58 Batch 60/173] avg loss 1.32544e-06, throughput 5.95897K wps
[Epoch 58 Batch 90/173] avg loss 1.33085e-06, throughput 5.95968K wps
[Epoch 58 Batch 120/173] avg loss 1.22043e-06, throughput 5.97724K wps
[Epoch 58 Batch 150/173] avg loss 1.1366e-06, throughput 5.98781K wps
Begin Testing...
[Epoch 58] train avg loss 1.31755e-06, test acc 0.7396, test avg loss 1.75176, throughput 5.99738K wps
[Epoch 59 Batch 30/173] avg loss 9.80314e-07, throughput 6.13069K wps
[Epoch 59 Batch 60/173] avg loss 1.14807e-06, throughput 5.97912K wps
[Epoch 59 Batch 90/173] avg loss 1.27253e-06, throughput 5.98277K wps
[Epoch 59 Batch 120/173] avg loss 1.35867e-06, throughput 5.9723K wps
[Epoch 59 Batch 150/173] avg loss 1.08743e-06, throughput 5.97238K wps
Begin Testing...
[Epoch 59] train avg loss 1.15682e-06, test acc 0.7333, test avg loss 1.74433, throughput 6.0059K wps
Test loss 0.50488, test acc 0.7664
Total time cost 358.10s
9596 1066
[Epoch 0 Batch 30/173] avg loss 0.0138736, throughput 5.76323K wps
[Epoch 0 Batch 60/173] avg loss 0.0138124, throughput 5.9722K wps
[Epoch 0 Batch 90/173] avg loss 0.0138385, throughput 5.97792K wps
[Epoch 0 Batch 120/173] avg loss 0.0137827, throughput 5.97068K wps
[Epoch 0 Batch 150/173] avg loss 0.0137797, throughput 5.9848K wps
Begin Testing...
[Epoch 0] train avg loss 0.0138301, test acc 0.5573, test avg loss 0.686689, throughput 5.93994K wps
Observed Improvement.
Begin Testing...
[Epoch 1 Batch 30/173] avg loss 0.0134921, throughput 6.12377K wps
[Epoch 1 Batch 60/173] avg loss 0.0134882, throughput 5.98254K wps
[Epoch 1 Batch 90/173] avg loss 0.0134218, throughput 5.97916K wps
[Epoch 1 Batch 120/173] avg loss 0.0134087, throughput 5.97314K wps
[Epoch 1 Batch 150/173] avg loss 0.013324, throughput 5.99422K wps
Begin Testing...
[Epoch 1] train avg loss 0.0134261, test acc 0.6646, test avg loss 0.67238, throughput 6.0069K wps
Observed Improvement.
Begin Testing...
[Epoch 2 Batch 30/173] avg loss 0.0128734, throughput 6.12859K wps
[Epoch 2 Batch 60/173] avg loss 0.0128246, throughput 5.97748K wps
[Epoch 2 Batch 90/173] avg loss 0.0127151, throughput 5.98456K wps
[Epoch 2 Batch 120/173] avg loss 0.0126095, throughput 5.98972K wps
[Epoch 2 Batch 150/173] avg loss 0.0124503, throughput 5.98324K wps
Begin Testing...
[Epoch 2] train avg loss 0.0126593, test acc 0.7052, test avg loss 0.633422, throughput 6.00999K wps
Observed Improvement.
Begin Testing...
[Epoch 3 Batch 30/173] avg loss 0.0116709, throughput 6.12868K wps
[Epoch 3 Batch 60/173] avg loss 0.0112628, throughput 5.98245K wps
[Epoch 3 Batch 90/173] avg loss 0.0111275, throughput 5.97558K wps
[Epoch 3 Batch 120/173] avg loss 0.0106071, throughput 5.98075K wps
[Epoch 3 Batch 150/173] avg loss 0.0105447, throughput 5.98329K wps
Begin Testing...
[Epoch 3] train avg loss 0.0109727, test acc 0.7281, test avg loss 0.563194, throughput 6.00638K wps
Observed Improvement.
Begin Testing...
[Epoch 4 Batch 30/173] avg loss 0.00899543, throughput 6.11738K wps
[Epoch 4 Batch 60/173] avg loss 0.00873845, throughput 5.97901K wps
[Epoch 4 Batch 90/173] avg loss 0.00858693, throughput 5.9815K wps
[Epoch 4 Batch 120/173] avg loss 0.00822122, throughput 5.98971K wps
[Epoch 4 Batch 150/173] avg loss 0.00827927, throughput 5.97801K wps
Begin Testing...
[Epoch 4] train avg loss 0.00846558, test acc 0.7604, test avg loss 0.50155, throughput 6.00373K wps
Observed Improvement.
Begin Testing...
[Epoch 5 Batch 30/173] avg loss 0.00651719, throughput 6.14809K wps
[Epoch 5 Batch 60/173] avg loss 0.00613757, throughput 5.98664K wps
[Epoch 5 Batch 90/173] avg loss 0.00595768, throughput 5.98517K wps
[Epoch 5 Batch 120/173] avg loss 0.0059446, throughput 5.98938K wps
[Epoch 5 Batch 150/173] avg loss 0.00568907, throughput 5.98554K wps
Begin Testing...
[Epoch 5] train avg loss 0.0060695, test acc 0.7667, test avg loss 0.485384, throughput 6.01553K wps
Observed Improvement.
Begin Testing...
[Epoch 6 Batch 30/173] avg loss 0.00477656, throughput 6.13242K wps
[Epoch 6 Batch 60/173] avg loss 0.00417011, throughput 5.98121K wps
[Epoch 6 Batch 90/173] avg loss 0.00433206, throughput 5.98472K wps
[Epoch 6 Batch 120/173] avg loss 0.00416033, throughput 5.99721K wps
[Epoch 6 Batch 150/173] avg loss 0.00399114, throughput 5.9896K wps
Begin Testing...
[Epoch 6] train avg loss 0.00428231, test acc 0.7781, test avg loss 0.491979, throughput 6.01263K wps
Observed Improvement.
Begin Testing...
[Epoch 7 Batch 30/173] avg loss 0.00296037, throughput 6.13343K wps
[Epoch 7 Batch 60/173] avg loss 0.0030804, throughput 5.98346K wps
[Epoch 7 Batch 90/173] avg loss 0.00297615, throughput 5.99413K wps
[Epoch 7 Batch 120/173] avg loss 0.00288993, throughput 5.9929K wps
[Epoch 7 Batch 150/173] avg loss 0.00303542, throughput 5.98136K wps
Begin Testing...
[Epoch 7] train avg loss 0.00298566, test acc 0.7760, test avg loss 0.514381, throughput 6.01243K wps
[Epoch 8 Batch 30/173] avg loss 0.00217078, throughput 6.12898K wps
[Epoch 8 Batch 60/173] avg loss 0.0021435, throughput 5.99427K wps
[Epoch 8 Batch 90/173] avg loss 0.00194387, throughput 5.99026K wps
[Epoch 8 Batch 120/173] avg loss 0.00227191, throughput 5.99268K wps
[Epoch 8 Batch 150/173] avg loss 0.0019959, throughput 5.98875K wps
Begin Testing...
[Epoch 8] train avg loss 0.00210602, test acc 0.7740, test avg loss 0.546986, throughput 6.01527K wps
[Epoch 9 Batch 30/173] avg loss 0.00156608, throughput 6.13647K wps
[Epoch 9 Batch 60/173] avg loss 0.00152342, throughput 6.00394K wps
[Epoch 9 Batch 90/173] avg loss 0.00148715, throughput 5.98068K wps
[Epoch 9 Batch 120/173] avg loss 0.00148912, throughput 5.98761K wps
[Epoch 9 Batch 150/173] avg loss 0.00142896, throughput 5.99811K wps
Begin Testing...
[Epoch 9] train avg loss 0.00147784, test acc 0.7656, test avg loss 0.589363, throughput 6.01578K wps
[Epoch 10 Batch 30/173] avg loss 0.000981228, throughput 6.14427K wps
[Epoch 10 Batch 60/173] avg loss 0.00104568, throughput 6.00996K wps
[Epoch 10 Batch 90/173] avg loss 0.00106742, throughput 6.00279K wps
[Epoch 10 Batch 120/173] avg loss 0.00122133, throughput 5.99458K wps
[Epoch 10 Batch 150/173] avg loss 0.0010053, throughput 5.99294K wps
Begin Testing...
[Epoch 10] train avg loss 0.00106014, test acc 0.7562, test avg loss 0.631042, throughput 6.02471K wps
[Epoch 11 Batch 30/173] avg loss 0.000748341, throughput 6.13359K wps
[Epoch 11 Batch 60/173] avg loss 0.000721945, throughput 5.98718K wps
[Epoch 11 Batch 90/173] avg loss 0.000726721, throughput 5.98576K wps
[Epoch 11 Batch 120/173] avg loss 0.000831549, throughput 5.97981K wps
[Epoch 11 Batch 150/173] avg loss 0.000709471, throughput 5.99292K wps
Begin Testing...
[Epoch 11] train avg loss 0.000772638, test acc 0.7604, test avg loss 0.669524, throughput 6.01437K wps
[Epoch 12 Batch 30/173] avg loss 0.000576346, throughput 6.12211K wps
[Epoch 12 Batch 60/173] avg loss 0.000609396, throughput 5.98696K wps
[Epoch 12 Batch 90/173] avg loss 0.000550079, throughput 5.99503K wps
[Epoch 12 Batch 120/173] avg loss 0.000552794, throughput 5.98962K wps
[Epoch 12 Batch 150/173] avg loss 0.000575434, throughput 5.99684K wps
Begin Testing...
[Epoch 12] train avg loss 0.000584418, test acc 0.7552, test avg loss 0.709196, throughput 6.01503K wps
[Epoch 13 Batch 30/173] avg loss 0.000420168, throughput 6.12917K wps
[Epoch 13 Batch 60/173] avg loss 0.000447542, throughput 5.97695K wps
[Epoch 13 Batch 90/173] avg loss 0.00047321, throughput 5.9908K wps
[Epoch 13 Batch 120/173] avg loss 0.000467442, throughput 5.97804K wps
[Epoch 13 Batch 150/173] avg loss 0.000407074, throughput 5.98276K wps
Begin Testing...
[Epoch 13] train avg loss 0.000447629, test acc 0.7562, test avg loss 0.74197, throughput 6.00733K wps
[Epoch 14 Batch 30/173] avg loss 0.000342822, throughput 6.15178K wps
[Epoch 14 Batch 60/173] avg loss 0.000347373, throughput 5.99969K wps
[Epoch 14 Batch 90/173] avg loss 0.000294939, throughput 5.99108K wps
[Epoch 14 Batch 120/173] avg loss 0.000313677, throughput 5.9844K wps
[Epoch 14 Batch 150/173] avg loss 0.000355665, throughput 5.98585K wps
Begin Testing...
[Epoch 14] train avg loss 0.000329124, test acc 0.7562, test avg loss 0.782392, throughput 6.01632K wps
[Epoch 15 Batch 30/173] avg loss 0.000256517, throughput 6.14978K wps
[Epoch 15 Batch 60/173] avg loss 0.000230935, throughput 5.9987K wps
[Epoch 15 Batch 90/173] avg loss 0.000261226, throughput 6.00155K wps
[Epoch 15 Batch 120/173] avg loss 0.000283534, throughput 5.99618K wps
[Epoch 15 Batch 150/173] avg loss 0.000273048, throughput 6.00727K wps
Begin Testing...
[Epoch 15] train avg loss 0.00026713, test acc 0.7510, test avg loss 0.814894, throughput 6.02623K wps
[Epoch 16 Batch 30/173] avg loss 0.000188615, throughput 6.13373K wps
[Epoch 16 Batch 60/173] avg loss 0.000201221, throughput 5.99145K wps
[Epoch 16 Batch 90/173] avg loss 0.000218198, throughput 5.99285K wps
[Epoch 16 Batch 120/173] avg loss 0.000260726, throughput 5.99022K wps
[Epoch 16 Batch 150/173] avg loss 0.00020932, throughput 5.98846K wps
Begin Testing...
[Epoch 16] train avg loss 0.000209157, test acc 0.7531, test avg loss 0.847386, throughput 6.01592K wps
[Epoch 17 Batch 30/173] avg loss 0.000154404, throughput 6.13089K wps
[Epoch 17 Batch 60/173] avg loss 0.000176729, throughput 5.98553K wps
[Epoch 17 Batch 90/173] avg loss 0.000153682, throughput 5.99161K wps
[Epoch 17 Batch 120/173] avg loss 0.000175932, throughput 5.98921K wps
[Epoch 17 Batch 150/173] avg loss 0.00015919, throughput 5.98812K wps
Begin Testing...
[Epoch 17] train avg loss 0.000170659, test acc 0.7500, test avg loss 0.886913, throughput 6.01431K wps
[Epoch 18 Batch 30/173] avg loss 0.000139527, throughput 6.12237K wps
[Epoch 18 Batch 60/173] avg loss 0.000162104, throughput 5.97699K wps
[Epoch 18 Batch 90/173] avg loss 0.000145474, throughput 5.97688K wps
[Epoch 18 Batch 120/173] avg loss 0.000139858, throughput 5.99195K wps
[Epoch 18 Batch 150/173] avg loss 0.000140377, throughput 5.98167K wps
Begin Testing...
[Epoch 18] train avg loss 0.000146703, test acc 0.7521, test avg loss 0.909476, throughput 6.00743K wps
[Epoch 19 Batch 30/173] avg loss 0.000112747, throughput 6.13093K wps
[Epoch 19 Batch 60/173] avg loss 0.00010569, throughput 5.98111K wps
[Epoch 19 Batch 90/173] avg loss 0.000149336, throughput 5.98135K wps
[Epoch 19 Batch 120/173] avg loss 0.000101651, throughput 5.97136K wps
[Epoch 19 Batch 150/173] avg loss 0.000107213, throughput 5.99153K wps
Begin Testing...
[Epoch 19] train avg loss 0.000114681, test acc 0.7510, test avg loss 0.936209, throughput 6.00736K wps
[Epoch 20 Batch 30/173] avg loss 0.000104781, throughput 6.13546K wps
[Epoch 20 Batch 60/173] avg loss 0.000102698, throughput 5.99044K wps
[Epoch 20 Batch 90/173] avg loss 0.000113567, throughput 5.98804K wps
[Epoch 20 Batch 120/173] avg loss 8.48489e-05, throughput 5.99098K wps
[Epoch 20 Batch 150/173] avg loss 8.37387e-05, throughput 5.99608K wps
Begin Testing...
[Epoch 20] train avg loss 0.000102882, test acc 0.7448, test avg loss 0.970083, throughput 6.01622K wps
[Epoch 21 Batch 30/173] avg loss 8.53529e-05, throughput 6.14189K wps
[Epoch 21 Batch 60/173] avg loss 8.08329e-05, throughput 5.9779K wps
[Epoch 21 Batch 90/173] avg loss 7.81841e-05, throughput 5.982K wps
[Epoch 21 Batch 120/173] avg loss 9.73412e-05, throughput 5.98511K wps
[Epoch 21 Batch 150/173] avg loss 8.42669e-05, throughput 5.98106K wps
Begin Testing...
[Epoch 21] train avg loss 8.33726e-05, test acc 0.7427, test avg loss 0.994608, throughput 6.01104K wps
[Epoch 22 Batch 30/173] avg loss 8.31495e-05, throughput 6.14697K wps
[Epoch 22 Batch 60/173] avg loss 5.93724e-05, throughput 5.99713K wps
[Epoch 22 Batch 90/173] avg loss 7.51828e-05, throughput 6.00042K wps
[Epoch 22 Batch 120/173] avg loss 5.80802e-05, throughput 5.99922K wps
[Epoch 22 Batch 150/173] avg loss 5.9176e-05, throughput 5.98579K wps
Begin Testing...
[Epoch 22] train avg loss 6.64512e-05, test acc 0.7490, test avg loss 1.01274, throughput 6.02049K wps
[Epoch 23 Batch 30/173] avg loss 5.68003e-05, throughput 6.1339K wps
[Epoch 23 Batch 60/173] avg loss 7.10986e-05, throughput 5.99704K wps
[Epoch 23 Batch 90/173] avg loss 6.74624e-05, throughput 5.99553K wps
[Epoch 23 Batch 120/173] avg loss 5.79227e-05, throughput 5.9836K wps
[Epoch 23 Batch 150/173] avg loss 7.20693e-05, throughput 5.99152K wps
Begin Testing...
[Epoch 23] train avg loss 6.54319e-05, test acc 0.7490, test avg loss 1.0374, throughput 6.01577K wps
[Epoch 24 Batch 30/173] avg loss 4.89492e-05, throughput 6.15188K wps
[Epoch 24 Batch 60/173] avg loss 5.92816e-05, throughput 6.00109K wps
[Epoch 24 Batch 90/173] avg loss 4.74981e-05, throughput 5.99112K wps
[Epoch 24 Batch 120/173] avg loss 5.81121e-05, throughput 5.99564K wps
[Epoch 24 Batch 150/173] avg loss 5.75769e-05, throughput 5.97618K wps
Begin Testing...
[Epoch 24] train avg loss 5.37033e-05, test acc 0.7479, test avg loss 1.06188, throughput 6.01674K wps
[Epoch 25 Batch 30/173] avg loss 5.26127e-05, throughput 6.13914K wps
[Epoch 25 Batch 60/173] avg loss 4.47976e-05, throughput 5.97258K wps
[Epoch 25 Batch 90/173] avg loss 4.64228e-05, throughput 5.98459K wps
[Epoch 25 Batch 120/173] avg loss 5.51569e-05, throughput 5.97983K wps
[Epoch 25 Batch 150/173] avg loss 4.24624e-05, throughput 5.99871K wps
Begin Testing...
[Epoch 25] train avg loss 4.6576e-05, test acc 0.7490, test avg loss 1.07376, throughput 6.01448K wps
[Epoch 26 Batch 30/173] avg loss 3.95386e-05, throughput 6.13964K wps
[Epoch 26 Batch 60/173] avg loss 3.33561e-05, throughput 5.9993K wps
[Epoch 26 Batch 90/173] avg loss 4.27793e-05, throughput 5.99406K wps
[Epoch 26 Batch 120/173] avg loss 3.43688e-05, throughput 5.99636K wps
[Epoch 26 Batch 150/173] avg loss 4.96109e-05, throughput 5.98476K wps
Begin Testing...
[Epoch 26] train avg loss 3.91585e-05, test acc 0.7469, test avg loss 1.1099, throughput 6.01935K wps
[Epoch 27 Batch 30/173] avg loss 2.82612e-05, throughput 6.13291K wps
[Epoch 27 Batch 60/173] avg loss 3.27077e-05, throughput 5.99358K wps
[Epoch 27 Batch 90/173] avg loss 5.2966e-05, throughput 6.00174K wps
[Epoch 27 Batch 120/173] avg loss 2.8574e-05, throughput 5.9918K wps
[Epoch 27 Batch 150/173] avg loss 2.76405e-05, throughput 5.98277K wps
Begin Testing...
[Epoch 27] train avg loss 3.30647e-05, test acc 0.7479, test avg loss 1.12924, throughput 6.01494K wps
[Epoch 28 Batch 30/173] avg loss 2.46642e-05, throughput 6.1197K wps
[Epoch 28 Batch 60/173] avg loss 3.29791e-05, throughput 5.92688K wps
[Epoch 28 Batch 90/173] avg loss 2.60617e-05, throughput 5.96582K wps
[Epoch 28 Batch 120/173] avg loss 2.55998e-05, throughput 5.9899K wps
[Epoch 28 Batch 150/173] avg loss 4.41726e-05, throughput 5.99781K wps
Begin Testing...
[Epoch 28] train avg loss 2.99973e-05, test acc 0.7458, test avg loss 1.1442, throughput 5.99949K wps
[Epoch 29 Batch 30/173] avg loss 2.25695e-05, throughput 6.14562K wps
[Epoch 29 Batch 60/173] avg loss 2.8361e-05, throughput 5.99191K wps
[Epoch 29 Batch 90/173] avg loss 2.65245e-05, throughput 5.99153K wps
[Epoch 29 Batch 120/173] avg loss 2.02506e-05, throughput 5.99687K wps
[Epoch 29 Batch 150/173] avg loss 2.4353e-05, throughput 6.00362K wps
Begin Testing...
[Epoch 29] train avg loss 2.63063e-05, test acc 0.7490, test avg loss 1.17505, throughput 6.02235K wps
[Epoch 30 Batch 30/173] avg loss 2.93797e-05, throughput 6.13361K wps
[Epoch 30 Batch 60/173] avg loss 2.79739e-05, throughput 5.9858K wps
[Epoch 30 Batch 90/173] avg loss 2.08077e-05, throughput 5.98458K wps
[Epoch 30 Batch 120/173] avg loss 1.86336e-05, throughput 5.98072K wps
[Epoch 30 Batch 150/173] avg loss 1.98776e-05, throughput 5.985K wps
Begin Testing...
[Epoch 30] train avg loss 2.29544e-05, test acc 0.7438, test avg loss 1.19959, throughput 6.01173K wps
[Epoch 31 Batch 30/173] avg loss 1.75405e-05, throughput 6.14863K wps
[Epoch 31 Batch 60/173] avg loss 2.58552e-05, throughput 5.98345K wps
[Epoch 31 Batch 90/173] avg loss 1.53367e-05, throughput 5.9779K wps
[Epoch 31 Batch 120/173] avg loss 2.69776e-05, throughput 5.98484K wps
[Epoch 31 Batch 150/173] avg loss 1.79728e-05, throughput 5.99172K wps
Begin Testing...
[Epoch 31] train avg loss 2.08568e-05, test acc 0.7438, test avg loss 1.21907, throughput 6.01508K wps
[Epoch 32 Batch 30/173] avg loss 2.02412e-05, throughput 6.13577K wps
[Epoch 32 Batch 60/173] avg loss 1.47559e-05, throughput 5.98255K wps
[Epoch 32 Batch 90/173] avg loss 3.2697e-05, throughput 5.98026K wps
[Epoch 32 Batch 120/173] avg loss 1.40068e-05, throughput 5.97991K wps
[Epoch 32 Batch 150/173] avg loss 1.77468e-05, throughput 5.97615K wps
Begin Testing...
[Epoch 32] train avg loss 1.98032e-05, test acc 0.7438, test avg loss 1.23378, throughput 6.008K wps
[Epoch 33 Batch 30/173] avg loss 1.47014e-05, throughput 6.13415K wps
[Epoch 33 Batch 60/173] avg loss 1.50328e-05, throughput 5.98571K wps
[Epoch 33 Batch 90/173] avg loss 1.29347e-05, throughput 5.98892K wps
[Epoch 33 Batch 120/173] avg loss 2.33838e-05, throughput 5.98374K wps
[Epoch 33 Batch 150/173] avg loss 1.81417e-05, throughput 5.9937K wps
Begin Testing...
[Epoch 33] train avg loss 1.65448e-05, test acc 0.7406, test avg loss 1.26066, throughput 6.01359K wps
[Epoch 34 Batch 30/173] avg loss 1.09144e-05, throughput 6.12469K wps
[Epoch 34 Batch 60/173] avg loss 1.38974e-05, throughput 5.97779K wps
[Epoch 34 Batch 90/173] avg loss 1.21255e-05, throughput 5.9892K wps
[Epoch 34 Batch 120/173] avg loss 2.1586e-05, throughput 5.97736K wps
[Epoch 34 Batch 150/173] avg loss 1.43369e-05, throughput 5.98262K wps
Begin Testing...
[Epoch 34] train avg loss 1.52882e-05, test acc 0.7427, test avg loss 1.27648, throughput 6.00766K wps
[Epoch 35 Batch 30/173] avg loss 1.26746e-05, throughput 6.11449K wps
[Epoch 35 Batch 60/173] avg loss 1.0834e-05, throughput 5.97435K wps
[Epoch 35 Batch 90/173] avg loss 2.37328e-05, throughput 5.98607K wps
[Epoch 35 Batch 120/173] avg loss 1.804e-05, throughput 5.98795K wps
[Epoch 35 Batch 150/173] avg loss 1.37435e-05, throughput 5.98939K wps
Begin Testing...
[Epoch 35] train avg loss 1.50917e-05, test acc 0.7406, test avg loss 1.29292, throughput 6.00868K wps
[Epoch 36 Batch 30/173] avg loss 1.15465e-05, throughput 6.13516K wps
[Epoch 36 Batch 60/173] avg loss 1.15618e-05, throughput 5.98783K wps
[Epoch 36 Batch 90/173] avg loss 1.05521e-05, throughput 5.99181K wps
[Epoch 36 Batch 120/173] avg loss 1.94036e-05, throughput 5.98619K wps
[Epoch 36 Batch 150/173] avg loss 1.07624e-05, throughput 5.99584K wps
Begin Testing...
[Epoch 36] train avg loss 1.24382e-05, test acc 0.7417, test avg loss 1.32251, throughput 6.01579K wps
[Epoch 37 Batch 30/173] avg loss 8.21761e-06, throughput 6.14795K wps
[Epoch 37 Batch 60/173] avg loss 9.12384e-06, throughput 5.98077K wps
[Epoch 37 Batch 90/173] avg loss 8.78049e-06, throughput 5.99641K wps
[Epoch 37 Batch 120/173] avg loss 9.60527e-06, throughput 5.99812K wps
[Epoch 37 Batch 150/173] avg loss 2.22648e-05, throughput 5.98269K wps
Begin Testing...
[Epoch 37] train avg loss 1.1619e-05, test acc 0.7406, test avg loss 1.34548, throughput 6.01692K wps
[Epoch 38 Batch 30/173] avg loss 7.89958e-06, throughput 6.14237K wps
[Epoch 38 Batch 60/173] avg loss 8.07136e-06, throughput 5.99862K wps
[Epoch 38 Batch 90/173] avg loss 1.86282e-05, throughput 5.98388K wps
[Epoch 38 Batch 120/173] avg loss 7.02375e-06, throughput 5.99676K wps
[Epoch 38 Batch 150/173] avg loss 8.19263e-06, throughput 5.97374K wps
Begin Testing...
[Epoch 38] train avg loss 9.60678e-06, test acc 0.7385, test avg loss 1.35421, throughput 6.01395K wps
[Epoch 39 Batch 30/173] avg loss 8.46544e-06, throughput 6.13516K wps
[Epoch 39 Batch 60/173] avg loss 1.61929e-05, throughput 5.97588K wps
[Epoch 39 Batch 90/173] avg loss 1.03699e-05, throughput 5.9907K wps
[Epoch 39 Batch 120/173] avg loss 8.08482e-06, throughput 5.98751K wps
[Epoch 39 Batch 150/173] avg loss 7.65265e-06, throughput 5.98179K wps
Begin Testing...
[Epoch 39] train avg loss 9.93768e-06, test acc 0.7417, test avg loss 1.36833, throughput 6.01114K wps
[Epoch 40 Batch 30/173] avg loss 9.12862e-06, throughput 6.1274K wps
[Epoch 40 Batch 60/173] avg loss 6.48672e-06, throughput 5.97817K wps
[Epoch 40 Batch 90/173] avg loss 6.00839e-06, throughput 5.9672K wps
[Epoch 40 Batch 120/173] avg loss 1.10858e-05, throughput 5.97654K wps
[Epoch 40 Batch 150/173] avg loss 1.82655e-05, throughput 5.97681K wps
Begin Testing...
[Epoch 40] train avg loss 1.01332e-05, test acc 0.7375, test avg loss 1.40227, throughput 6.00339K wps
[Epoch 41 Batch 30/173] avg loss 5.01571e-06, throughput 6.13367K wps
[Epoch 41 Batch 60/173] avg loss 6.11128e-06, throughput 5.9755K wps
[Epoch 41 Batch 90/173] avg loss 6.42799e-06, throughput 5.98285K wps
[Epoch 41 Batch 120/173] avg loss 1.60059e-05, throughput 5.978K wps
[Epoch 41 Batch 150/173] avg loss 7.71011e-06, throughput 5.96224K wps
Begin Testing...
[Epoch 41] train avg loss 8.13905e-06, test acc 0.7385, test avg loss 1.42175, throughput 6.00318K wps
[Epoch 42 Batch 30/173] avg loss 5.52029e-06, throughput 6.12268K wps
[Epoch 42 Batch 60/173] avg loss 6.74243e-06, throughput 5.97544K wps
[Epoch 42 Batch 90/173] avg loss 5.76699e-06, throughput 5.998K wps
[Epoch 42 Batch 120/173] avg loss 1.48332e-05, throughput 5.98243K wps
[Epoch 42 Batch 150/173] avg loss 6.83578e-06, throughput 5.99048K wps
Begin Testing...
[Epoch 42] train avg loss 7.62181e-06, test acc 0.7396, test avg loss 1.44258, throughput 6.01157K wps
[Epoch 43 Batch 30/173] avg loss 4.87087e-06, throughput 6.13721K wps
[Epoch 43 Batch 60/173] avg loss 4.85431e-06, throughput 5.97964K wps
[Epoch 43 Batch 90/173] avg loss 4.40852e-06, throughput 5.97494K wps
[Epoch 43 Batch 120/173] avg loss 6.52416e-06, throughput 5.97065K wps
[Epoch 43 Batch 150/173] avg loss 5.28322e-06, throughput 5.97845K wps
Begin Testing...
[Epoch 43] train avg loss 6.73312e-06, test acc 0.7427, test avg loss 1.4678, throughput 6.00726K wps
[Epoch 44 Batch 30/173] avg loss 3.21338e-06, throughput 6.14437K wps
[Epoch 44 Batch 60/173] avg loss 4.94844e-06, throughput 5.99553K wps
[Epoch 44 Batch 90/173] avg loss 4.44424e-06, throughput 5.98499K wps
[Epoch 44 Batch 120/173] avg loss 7.74226e-06, throughput 5.9856K wps
[Epoch 44 Batch 150/173] avg loss 1.40722e-05, throughput 5.99685K wps
Begin Testing...
[Epoch 44] train avg loss 6.6585e-06, test acc 0.7323, test avg loss 1.49023, throughput 6.01624K wps
[Epoch 45 Batch 30/173] avg loss 3.85665e-06, throughput 6.13779K wps
[Epoch 45 Batch 60/173] avg loss 8.64833e-06, throughput 5.98201K wps
[Epoch 45 Batch 90/173] avg loss 4.24466e-06, throughput 5.99782K wps
[Epoch 45 Batch 120/173] avg loss 4.03501e-06, throughput 5.99931K wps
[Epoch 45 Batch 150/173] avg loss 4.55762e-06, throughput 5.98009K wps
Begin Testing...
[Epoch 45] train avg loss 6.3883e-06, test acc 0.7406, test avg loss 1.50936, throughput 6.01474K wps
[Epoch 46 Batch 30/173] avg loss 3.11075e-06, throughput 6.12755K wps
[Epoch 46 Batch 60/173] avg loss 3.01313e-06, throughput 5.98776K wps
[Epoch 46 Batch 90/173] avg loss 6.57026e-06, throughput 5.97841K wps
[Epoch 46 Batch 120/173] avg loss 2.92839e-06, throughput 5.98322K wps
[Epoch 46 Batch 150/173] avg loss 1.34336e-05, throughput 5.98673K wps
Begin Testing...
[Epoch 46] train avg loss 5.5082e-06, test acc 0.7396, test avg loss 1.51541, throughput 6.00886K wps
[Epoch 47 Batch 30/173] avg loss 1.14731e-05, throughput 6.13737K wps
[Epoch 47 Batch 60/173] avg loss 3.7889e-06, throughput 5.98715K wps
[Epoch 47 Batch 90/173] avg loss 4.04425e-06, throughput 6.01131K wps
[Epoch 47 Batch 120/173] avg loss 4.17606e-06, throughput 5.99865K wps
[Epoch 47 Batch 150/173] avg loss 3.90423e-06, throughput 5.99275K wps
Begin Testing...
[Epoch 47] train avg loss 5.12705e-06, test acc 0.7375, test avg loss 1.54611, throughput 6.02086K wps
[Epoch 48 Batch 30/173] avg loss 1.17646e-05, throughput 6.14615K wps
[Epoch 48 Batch 60/173] avg loss 3.37327e-06, throughput 5.98569K wps
[Epoch 48 Batch 90/173] avg loss 2.31972e-06, throughput 5.99423K wps
[Epoch 48 Batch 120/173] avg loss 3.23178e-06, throughput 5.98623K wps
[Epoch 48 Batch 150/173] avg loss 4.24322e-06, throughput 6.00281K wps
Begin Testing...
[Epoch 48] train avg loss 4.76285e-06, test acc 0.7406, test avg loss 1.56629, throughput 6.01869K wps
[Epoch 49 Batch 30/173] avg loss 2.8836e-06, throughput 6.13203K wps
[Epoch 49 Batch 60/173] avg loss 1.11617e-05, throughput 6.00128K wps
[Epoch 49 Batch 90/173] avg loss 2.45692e-06, throughput 5.9961K wps
[Epoch 49 Batch 120/173] avg loss 3.03704e-06, throughput 5.98367K wps
[Epoch 49 Batch 150/173] avg loss 3.01778e-06, throughput 5.99642K wps
Begin Testing...
[Epoch 49] train avg loss 5.26399e-06, test acc 0.7323, test avg loss 1.60461, throughput 6.01815K wps
[Epoch 50 Batch 30/173] avg loss 4.9883e-06, throughput 6.1231K wps
[Epoch 50 Batch 60/173] avg loss 2.4139e-06, throughput 5.98508K wps
[Epoch 50 Batch 90/173] avg loss 1.23929e-05, throughput 5.97525K wps
[Epoch 50 Batch 120/173] avg loss 2.24394e-06, throughput 5.96756K wps
[Epoch 50 Batch 150/173] avg loss 2.76766e-06, throughput 5.99621K wps
Begin Testing...
[Epoch 50] train avg loss 5.18243e-06, test acc 0.7385, test avg loss 1.60629, throughput 6.00739K wps
[Epoch 51 Batch 30/173] avg loss 3.07522e-06, throughput 6.13647K wps
[Epoch 51 Batch 60/173] avg loss 1.75496e-06, throughput 5.98389K wps
[Epoch 51 Batch 90/173] avg loss 2.98338e-06, throughput 6.00385K wps
[Epoch 51 Batch 120/173] avg loss 1.98605e-06, throughput 5.99679K wps
[Epoch 51 Batch 150/173] avg loss 1.1598e-05, throughput 5.9854K wps
Begin Testing...
[Epoch 51] train avg loss 3.95139e-06, test acc 0.7354, test avg loss 1.60901, throughput 6.01528K wps
[Epoch 52 Batch 30/173] avg loss 2.24178e-06, throughput 6.13732K wps
[Epoch 52 Batch 60/173] avg loss 2.2258e-06, throughput 5.9745K wps
[Epoch 52 Batch 90/173] avg loss 2.28652e-06, throughput 5.98856K wps
[Epoch 52 Batch 120/173] avg loss 2.28186e-06, throughput 5.98529K wps
[Epoch 52 Batch 150/173] avg loss 1.3908e-06, throughput 5.97602K wps
Begin Testing...
[Epoch 52] train avg loss 3.84747e-06, test acc 0.7344, test avg loss 1.6369, throughput 6.00938K wps
[Epoch 53 Batch 30/173] avg loss 5.56701e-06, throughput 6.14454K wps
[Epoch 53 Batch 60/173] avg loss 2.23686e-06, throughput 5.99737K wps
[Epoch 53 Batch 90/173] avg loss 1.86825e-06, throughput 6.00001K wps
[Epoch 53 Batch 120/173] avg loss 1.93046e-06, throughput 5.98537K wps
[Epoch 53 Batch 150/173] avg loss 1.96032e-06, throughput 6.00174K wps
Begin Testing...
[Epoch 53] train avg loss 4.19405e-06, test acc 0.7385, test avg loss 1.64058, throughput 6.02257K wps
[Epoch 54 Batch 30/173] avg loss 1.63499e-06, throughput 6.14306K wps
[Epoch 54 Batch 60/173] avg loss 1.69317e-06, throughput 5.99123K wps
[Epoch 54 Batch 90/173] avg loss 1.13475e-05, throughput 5.99076K wps
[Epoch 54 Batch 120/173] avg loss 1.72774e-06, throughput 5.98094K wps
[Epoch 54 Batch 150/173] avg loss 1.86665e-06, throughput 5.98371K wps
Begin Testing...
[Epoch 54] train avg loss 3.33451e-06, test acc 0.7438, test avg loss 1.64667, throughput 6.01391K wps
[Epoch 55 Batch 30/173] avg loss 1.20807e-06, throughput 6.12656K wps
[Epoch 55 Batch 60/173] avg loss 1.97294e-06, throughput 5.98054K wps
[Epoch 55 Batch 90/173] avg loss 1.45183e-06, throughput 5.97036K wps
[Epoch 55 Batch 120/173] avg loss 1.77859e-06, throughput 5.97559K wps
[Epoch 55 Batch 150/173] avg loss 1.02612e-05, throughput 5.98704K wps
Begin Testing...
[Epoch 55] train avg loss 3.077e-06, test acc 0.7448, test avg loss 1.67673, throughput 6.00544K wps
[Epoch 56 Batch 30/173] avg loss 1.23685e-06, throughput 6.13933K wps
[Epoch 56 Batch 60/173] avg loss 1.40859e-06, throughput 5.98232K wps
[Epoch 56 Batch 90/173] avg loss 1.61742e-06, throughput 5.97521K wps
[Epoch 56 Batch 120/173] avg loss 1.03955e-05, throughput 5.97621K wps
[Epoch 56 Batch 150/173] avg loss 1.93737e-06, throughput 5.99996K wps
Begin Testing...
[Epoch 56] train avg loss 3.11816e-06, test acc 0.7417, test avg loss 1.69565, throughput 6.01158K wps
[Epoch 57 Batch 30/173] avg loss 1.51425e-06, throughput 6.13986K wps
[Epoch 57 Batch 60/173] avg loss 1.59763e-06, throughput 5.98884K wps
[Epoch 57 Batch 90/173] avg loss 1.14086e-06, throughput 5.98378K wps
[Epoch 57 Batch 120/173] avg loss 1.04694e-05, throughput 5.99017K wps
[Epoch 57 Batch 150/173] avg loss 1.97258e-06, throughput 5.99753K wps
Begin Testing...
[Epoch 57] train avg loss 3.06747e-06, test acc 0.7417, test avg loss 1.71281, throughput 6.01693K wps
[Epoch 58 Batch 30/173] avg loss 1.38869e-06, throughput 6.14797K wps
[Epoch 58 Batch 60/173] avg loss 1.01007e-05, throughput 6.00101K wps
[Epoch 58 Batch 90/173] avg loss 9.90468e-07, throughput 5.98238K wps
[Epoch 58 Batch 120/173] avg loss 9.96205e-07, throughput 5.9368K wps
[Epoch 58 Batch 150/173] avg loss 9.23764e-07, throughput 5.98504K wps
Begin Testing...
[Epoch 58] train avg loss 2.63323e-06, test acc 0.7406, test avg loss 1.73615, throughput 6.00735K wps
[Epoch 59 Batch 30/173] avg loss 1.22512e-06, throughput 6.1415K wps
[Epoch 59 Batch 60/173] avg loss 1.85519e-06, throughput 5.99405K wps
[Epoch 59 Batch 90/173] avg loss 9.29724e-07, throughput 5.98895K wps
[Epoch 59 Batch 120/173] avg loss 8.98241e-06, throughput 5.99759K wps
[Epoch 59 Batch 150/173] avg loss 9.5136e-07, throughput 5.99936K wps
Begin Testing...
[Epoch 59] train avg loss 2.66669e-06, test acc 0.7406, test avg loss 1.762, throughput 6.02306K wps
Test loss 0.488674, test acc 0.7739
Total time cost 358.43s
9596 1066
[Epoch 0 Batch 30/173] avg loss 0.0138581, throughput 5.80181K wps
[Epoch 0 Batch 60/173] avg loss 0.0138388, throughput 5.9871K wps
[Epoch 0 Batch 90/173] avg loss 0.0138316, throughput 5.99243K wps
[Epoch 0 Batch 120/173] avg loss 0.013798, throughput 6.00108K wps
[Epoch 0 Batch 150/173] avg loss 0.0137259, throughput 6.00464K wps
Begin Testing...
[Epoch 0] train avg loss 0.0138245, test acc 0.6490, test avg loss 0.686592, throughput 5.96386K wps
Observed Improvement.
Begin Testing...
[Epoch 1 Batch 30/173] avg loss 0.0134742, throughput 6.12764K wps
[Epoch 1 Batch 60/173] avg loss 0.0134597, throughput 5.99056K wps
[Epoch 1 Batch 90/173] avg loss 0.0134222, throughput 5.99191K wps
[Epoch 1 Batch 120/173] avg loss 0.0133545, throughput 6.00178K wps
[Epoch 1 Batch 150/173] avg loss 0.0133355, throughput 6.00221K wps
Begin Testing...
[Epoch 1] train avg loss 0.0134098, test acc 0.6729, test avg loss 0.671692, throughput 6.01888K wps
Observed Improvement.
Begin Testing...
[Epoch 2 Batch 30/173] avg loss 0.0129351, throughput 6.13876K wps
[Epoch 2 Batch 60/173] avg loss 0.0128053, throughput 5.98769K wps
[Epoch 2 Batch 90/173] avg loss 0.0126683, throughput 5.98285K wps
[Epoch 2 Batch 120/173] avg loss 0.0124557, throughput 6.00079K wps
[Epoch 2 Batch 150/173] avg loss 0.0124076, throughput 5.98637K wps
Begin Testing...
[Epoch 2] train avg loss 0.012613, test acc 0.6854, test avg loss 0.632697, throughput 6.01595K wps
Observed Improvement.
Begin Testing...
[Epoch 3 Batch 30/173] avg loss 0.0116695, throughput 6.13526K wps
[Epoch 3 Batch 60/173] avg loss 0.0113354, throughput 5.9854K wps
[Epoch 3 Batch 90/173] avg loss 0.011021, throughput 5.98018K wps
[Epoch 3 Batch 120/173] avg loss 0.0107911, throughput 5.98614K wps
[Epoch 3 Batch 150/173] avg loss 0.0105536, throughput 5.99693K wps
Begin Testing...
[Epoch 3] train avg loss 0.0109895, test acc 0.7396, test avg loss 0.564485, throughput 6.0135K wps
Observed Improvement.
Begin Testing...
[Epoch 4 Batch 30/173] avg loss 0.00930597, throughput 6.13141K wps
[Epoch 4 Batch 60/173] avg loss 0.00887498, throughput 5.98858K wps
[Epoch 4 Batch 90/173] avg loss 0.00879953, throughput 5.98062K wps
[Epoch 4 Batch 120/173] avg loss 0.0081349, throughput 5.97274K wps
[Epoch 4 Batch 150/173] avg loss 0.008226, throughput 5.97764K wps
Begin Testing...
[Epoch 4] train avg loss 0.00856136, test acc 0.7531, test avg loss 0.5016, throughput 6.00729K wps
Observed Improvement.
Begin Testing...
[Epoch 5 Batch 30/173] avg loss 0.00641252, throughput 6.13928K wps
[Epoch 5 Batch 60/173] avg loss 0.00642763, throughput 5.99262K wps
[Epoch 5 Batch 90/173] avg loss 0.00604683, throughput 5.9932K wps
[Epoch 5 Batch 120/173] avg loss 0.00612354, throughput 5.99484K wps
[Epoch 5 Batch 150/173] avg loss 0.00599615, throughput 5.9898K wps
Begin Testing...
[Epoch 5] train avg loss 0.00620961, test acc 0.7594, test avg loss 0.479232, throughput 6.01775K wps
Observed Improvement.
Begin Testing...
[Epoch 6 Batch 30/173] avg loss 0.00458632, throughput 6.12931K wps
[Epoch 6 Batch 60/173] avg loss 0.00451505, throughput 5.9856K wps
[Epoch 6 Batch 90/173] avg loss 0.00417264, throughput 5.98941K wps
[Epoch 6 Batch 120/173] avg loss 0.00399805, throughput 5.98562K wps
[Epoch 6 Batch 150/173] avg loss 0.00426647, throughput 5.99287K wps
Begin Testing...
[Epoch 6] train avg loss 0.0043164, test acc 0.7594, test avg loss 0.492507, throughput 6.01287K wps
Observed Improvement.
Begin Testing...
[Epoch 7 Batch 30/173] avg loss 0.00309604, throughput 6.139K wps
[Epoch 7 Batch 60/173] avg loss 0.00287735, throughput 5.99258K wps
[Epoch 7 Batch 90/173] avg loss 0.00312089, throughput 5.98462K wps
[Epoch 7 Batch 120/173] avg loss 0.00283196, throughput 5.99649K wps
[Epoch 7 Batch 150/173] avg loss 0.00324466, throughput 6.00269K wps
Begin Testing...
[Epoch 7] train avg loss 0.00302614, test acc 0.7583, test avg loss 0.52766, throughput 6.01902K wps
[Epoch 8 Batch 30/173] avg loss 0.00219029, throughput 6.12585K wps
[Epoch 8 Batch 60/173] avg loss 0.00212418, throughput 5.99145K wps
[Epoch 8 Batch 90/173] avg loss 0.00210101, throughput 5.98024K wps
[Epoch 8 Batch 120/173] avg loss 0.00215232, throughput 5.99122K wps
[Epoch 8 Batch 150/173] avg loss 0.00207579, throughput 5.98362K wps
Begin Testing...
[Epoch 8] train avg loss 0.0021197, test acc 0.7500, test avg loss 0.571467, throughput 6.01283K wps
[Epoch 9 Batch 30/173] avg loss 0.00132048, throughput 6.13162K wps
[Epoch 9 Batch 60/173] avg loss 0.00147143, throughput 6.00072K wps
[Epoch 9 Batch 90/173] avg loss 0.00146768, throughput 5.99718K wps
[Epoch 9 Batch 120/173] avg loss 0.00155563, throughput 5.98597K wps
[Epoch 9 Batch 150/173] avg loss 0.00152971, throughput 5.98244K wps
Begin Testing...
[Epoch 9] train avg loss 0.00145036, test acc 0.7448, test avg loss 0.618518, throughput 6.01714K wps
[Epoch 10 Batch 30/173] avg loss 0.00102803, throughput 6.14203K wps
[Epoch 10 Batch 60/173] avg loss 0.000996931, throughput 5.99768K wps
[Epoch 10 Batch 90/173] avg loss 0.000977487, throughput 5.97899K wps
[Epoch 10 Batch 120/173] avg loss 0.00104613, throughput 6.0078K wps
[Epoch 10 Batch 150/173] avg loss 0.000985252, throughput 6.00036K wps
Begin Testing...
[Epoch 10] train avg loss 0.00100651, test acc 0.7396, test avg loss 0.672639, throughput 6.0223K wps
[Epoch 11 Batch 30/173] avg loss 0.000686497, throughput 6.14807K wps
[Epoch 11 Batch 60/173] avg loss 0.000685929, throughput 5.9972K wps
[Epoch 11 Batch 90/173] avg loss 0.000823224, throughput 6.0082K wps
[Epoch 11 Batch 120/173] avg loss 0.000784096, throughput 5.99423K wps
[Epoch 11 Batch 150/173] avg loss 0.000762411, throughput 6.00231K wps
Begin Testing...
[Epoch 11] train avg loss 0.000732443, test acc 0.7365, test avg loss 0.719713, throughput 6.02638K wps
[Epoch 12 Batch 30/173] avg loss 0.000523056, throughput 6.13814K wps
[Epoch 12 Batch 60/173] avg loss 0.000622008, throughput 5.98853K wps
[Epoch 12 Batch 90/173] avg loss 0.000481362, throughput 6.0074K wps
[Epoch 12 Batch 120/173] avg loss 0.00056282, throughput 5.98864K wps
[Epoch 12 Batch 150/173] avg loss 0.000523004, throughput 6.00348K wps
Begin Testing...
[Epoch 12] train avg loss 0.00054091, test acc 0.7292, test avg loss 0.77265, throughput 6.02094K wps
[Epoch 13 Batch 30/173] avg loss 0.000433486, throughput 6.12034K wps
[Epoch 13 Batch 60/173] avg loss 0.000415298, throughput 5.98703K wps
[Epoch 13 Batch 90/173] avg loss 0.000398538, throughput 6.00342K wps
[Epoch 13 Batch 120/173] avg loss 0.000450692, throughput 5.99268K wps
[Epoch 13 Batch 150/173] avg loss 0.00037504, throughput 6.00127K wps
Begin Testing...
[Epoch 13] train avg loss 0.000412153, test acc 0.7198, test avg loss 0.814484, throughput 6.01767K wps
[Epoch 14 Batch 30/173] avg loss 0.00027812, throughput 6.15021K wps
[Epoch 14 Batch 60/173] avg loss 0.000285291, throughput 5.99272K wps
[Epoch 14 Batch 90/173] avg loss 0.000301484, throughput 5.97494K wps
[Epoch 14 Batch 120/173] avg loss 0.000329037, throughput 5.98732K wps
[Epoch 14 Batch 150/173] avg loss 0.000311581, throughput 5.97844K wps
Begin Testing...
[Epoch 14] train avg loss 0.000303025, test acc 0.7250, test avg loss 0.856727, throughput 6.01239K wps
[Epoch 15 Batch 30/173] avg loss 0.000216275, throughput 6.14117K wps
[Epoch 15 Batch 60/173] avg loss 0.000240394, throughput 5.98044K wps
[Epoch 15 Batch 90/173] avg loss 0.000281218, throughput 5.99258K wps
[Epoch 15 Batch 120/173] avg loss 0.000241655, throughput 6.00367K wps
[Epoch 15 Batch 150/173] avg loss 0.000224189, throughput 5.99179K wps
Begin Testing...
[Epoch 15] train avg loss 0.000248306, test acc 0.7260, test avg loss 0.888811, throughput 6.0177K wps
[Epoch 16 Batch 30/173] avg loss 0.000181709, throughput 6.15563K wps
[Epoch 16 Batch 60/173] avg loss 0.000189283, throughput 5.98754K wps
[Epoch 16 Batch 90/173] avg loss 0.000206904, throughput 5.98149K wps
[Epoch 16 Batch 120/173] avg loss 0.000201577, throughput 5.99451K wps
[Epoch 16 Batch 150/173] avg loss 0.000192738, throughput 6.01484K wps
Begin Testing...
[Epoch 16] train avg loss 0.000191871, test acc 0.7281, test avg loss 0.929677, throughput 6.02339K wps
[Epoch 17 Batch 30/173] avg loss 0.000186164, throughput 6.1283K wps
[Epoch 17 Batch 60/173] avg loss 0.000136291, throughput 5.99208K wps
[Epoch 17 Batch 90/173] avg loss 0.000130061, throughput 5.98921K wps
[Epoch 17 Batch 120/173] avg loss 0.000166879, throughput 5.99047K wps
[Epoch 17 Batch 150/173] avg loss 0.000162337, throughput 5.98248K wps
Begin Testing...
[Epoch 17] train avg loss 0.00015845, test acc 0.7208, test avg loss 0.967821, throughput 6.01386K wps
[Epoch 18 Batch 30/173] avg loss 0.000119569, throughput 6.1467K wps
[Epoch 18 Batch 60/173] avg loss 0.00012959, throughput 6.0014K wps
[Epoch 18 Batch 90/173] avg loss 0.000124211, throughput 6.0041K wps
[Epoch 18 Batch 120/173] avg loss 0.000133896, throughput 6.00283K wps
[Epoch 18 Batch 150/173] avg loss 0.00012685, throughput 5.99178K wps
Begin Testing...
[Epoch 18] train avg loss 0.000127993, test acc 0.7240, test avg loss 0.996875, throughput 6.02355K wps
[Epoch 19 Batch 30/173] avg loss 0.000125159, throughput 6.1448K wps
[Epoch 19 Batch 60/173] avg loss 0.000102728, throughput 6.00448K wps
[Epoch 19 Batch 90/173] avg loss 0.000107086, throughput 5.97969K wps
[Epoch 19 Batch 120/173] avg loss 0.000110754, throughput 5.98979K wps
[Epoch 19 Batch 150/173] avg loss 0.00011962, throughput 5.99737K wps
Begin Testing...
[Epoch 19] train avg loss 0.000112126, test acc 0.7188, test avg loss 1.03351, throughput 6.02054K wps
[Epoch 20 Batch 30/173] avg loss 9.1155e-05, throughput 6.1402K wps
[Epoch 20 Batch 60/173] avg loss 0.000115797, throughput 5.98432K wps
[Epoch 20 Batch 90/173] avg loss 8.95928e-05, throughput 5.98049K wps
[Epoch 20 Batch 120/173] avg loss 0.000100672, throughput 5.98117K wps
[Epoch 20 Batch 150/173] avg loss 8.62249e-05, throughput 5.98685K wps
Begin Testing...
[Epoch 20] train avg loss 9.74183e-05, test acc 0.7125, test avg loss 1.06572, throughput 6.01006K wps
[Epoch 21 Batch 30/173] avg loss 6.87464e-05, throughput 6.12232K wps
[Epoch 21 Batch 60/173] avg loss 7.10771e-05, throughput 6.00415K wps
[Epoch 21 Batch 90/173] avg loss 7.5425e-05, throughput 6.00375K wps
[Epoch 21 Batch 120/173] avg loss 8.85548e-05, throughput 6.00494K wps
[Epoch 21 Batch 150/173] avg loss 6.84724e-05, throughput 5.99816K wps
Begin Testing...
[Epoch 21] train avg loss 7.63615e-05, test acc 0.7219, test avg loss 1.08824, throughput 6.02123K wps
[Epoch 22 Batch 30/173] avg loss 6.37113e-05, throughput 6.14104K wps
[Epoch 22 Batch 60/173] avg loss 6.30499e-05, throughput 5.99856K wps
[Epoch 22 Batch 90/173] avg loss 6.14184e-05, throughput 5.98534K wps
[Epoch 22 Batch 120/173] avg loss 6.56844e-05, throughput 5.98886K wps
[Epoch 22 Batch 150/173] avg loss 8.50339e-05, throughput 5.98646K wps
Begin Testing...
[Epoch 22] train avg loss 6.69317e-05, test acc 0.7208, test avg loss 1.12158, throughput 6.01545K wps
[Epoch 23 Batch 30/173] avg loss 4.94784e-05, throughput 6.13149K wps
[Epoch 23 Batch 60/173] avg loss 5.80565e-05, throughput 5.99692K wps
[Epoch 23 Batch 90/173] avg loss 4.93704e-05, throughput 5.99173K wps
[Epoch 23 Batch 120/173] avg loss 5.58751e-05, throughput 5.98524K wps
[Epoch 23 Batch 150/173] avg loss 5.29602e-05, throughput 5.98457K wps
Begin Testing...
[Epoch 23] train avg loss 5.56082e-05, test acc 0.7208, test avg loss 1.14903, throughput 6.01429K wps
[Epoch 24 Batch 30/173] avg loss 3.89327e-05, throughput 6.06606K wps
[Epoch 24 Batch 60/173] avg loss 4.48968e-05, throughput 5.99338K wps
[Epoch 24 Batch 90/173] avg loss 6.31438e-05, throughput 5.99942K wps
[Epoch 24 Batch 120/173] avg loss 4.69524e-05, throughput 6.00024K wps
[Epoch 24 Batch 150/173] avg loss 5.34437e-05, throughput 5.98611K wps
Begin Testing...
[Epoch 24] train avg loss 5.00113e-05, test acc 0.7177, test avg loss 1.17785, throughput 6.00852K wps
[Epoch 25 Batch 30/173] avg loss 4.14357e-05, throughput 6.13758K wps
[Epoch 25 Batch 60/173] avg loss 3.31554e-05, throughput 5.97929K wps
[Epoch 25 Batch 90/173] avg loss 4.25476e-05, throughput 5.98369K wps
[Epoch 25 Batch 120/173] avg loss 3.83027e-05, throughput 6.00365K wps
[Epoch 25 Batch 150/173] avg loss 4.45313e-05, throughput 5.99606K wps
Begin Testing...
[Epoch 25] train avg loss 4.15873e-05, test acc 0.7208, test avg loss 1.20138, throughput 6.01748K wps
[Epoch 26 Batch 30/173] avg loss 2.75932e-05, throughput 6.15325K wps
[Epoch 26 Batch 60/173] avg loss 4.29323e-05, throughput 5.99575K wps
[Epoch 26 Batch 90/173] avg loss 4.48111e-05, throughput 5.98797K wps
[Epoch 26 Batch 120/173] avg loss 3.44741e-05, throughput 6.00961K wps
[Epoch 26 Batch 150/173] avg loss 3.8276e-05, throughput 6.005K wps
Begin Testing...
[Epoch 26] train avg loss 3.84587e-05, test acc 0.7240, test avg loss 1.23302, throughput 6.02654K wps
[Epoch 27 Batch 30/173] avg loss 2.47385e-05, throughput 6.14154K wps
[Epoch 27 Batch 60/173] avg loss 2.92298e-05, throughput 5.98586K wps
[Epoch 27 Batch 90/173] avg loss 2.72276e-05, throughput 5.98758K wps
[Epoch 27 Batch 120/173] avg loss 3.11694e-05, throughput 5.9981K wps
[Epoch 27 Batch 150/173] avg loss 2.89878e-05, throughput 5.98608K wps
Begin Testing...
[Epoch 27] train avg loss 3.05886e-05, test acc 0.7219, test avg loss 1.2539, throughput 6.01491K wps
[Epoch 28 Batch 30/173] avg loss 2.90155e-05, throughput 6.14277K wps
[Epoch 28 Batch 60/173] avg loss 3.87853e-05, throughput 6.00405K wps
[Epoch 28 Batch 90/173] avg loss 2.41498e-05, throughput 5.92848K wps
[Epoch 28 Batch 120/173] avg loss 2.74228e-05, throughput 5.94493K wps
[Epoch 28 Batch 150/173] avg loss 2.92717e-05, throughput 5.977K wps
Begin Testing...
[Epoch 28] train avg loss 3.0013e-05, test acc 0.7260, test avg loss 1.27844, throughput 6.00024K wps
[Epoch 29 Batch 30/173] avg loss 2.6781e-05, throughput 6.14668K wps
[Epoch 29 Batch 60/173] avg loss 1.98395e-05, throughput 6.00369K wps
[Epoch 29 Batch 90/173] avg loss 3.15107e-05, throughput 5.98762K wps
[Epoch 29 Batch 120/173] avg loss 2.0211e-05, throughput 5.98504K wps
[Epoch 29 Batch 150/173] avg loss 2.30062e-05, throughput 5.98191K wps
Begin Testing...
[Epoch 29] train avg loss 2.46519e-05, test acc 0.7260, test avg loss 1.29869, throughput 6.01779K wps
[Epoch 30 Batch 30/173] avg loss 2.13976e-05, throughput 6.13772K wps
[Epoch 30 Batch 60/173] avg loss 2.08724e-05, throughput 5.98293K wps
[Epoch 30 Batch 90/173] avg loss 1.78014e-05, throughput 5.98638K wps
[Epoch 30 Batch 120/173] avg loss 2.0902e-05, throughput 5.97872K wps
[Epoch 30 Batch 150/173] avg loss 3.99448e-05, throughput 5.98941K wps
Begin Testing...
[Epoch 30] train avg loss 2.32369e-05, test acc 0.7271, test avg loss 1.32385, throughput 6.0125K wps
[Epoch 31 Batch 30/173] avg loss 1.59915e-05, throughput 6.13494K wps
[Epoch 31 Batch 60/173] avg loss 2.4632e-05, throughput 5.99161K wps
[Epoch 31 Batch 90/173] avg loss 1.51921e-05, throughput 5.98619K wps
[Epoch 31 Batch 120/173] avg loss 1.86504e-05, throughput 5.97938K wps
[Epoch 31 Batch 150/173] avg loss 1.74705e-05, throughput 5.98969K wps
Begin Testing...
[Epoch 31] train avg loss 1.83385e-05, test acc 0.7250, test avg loss 1.34895, throughput 6.01218K wps
[Epoch 32 Batch 30/173] avg loss 2.36342e-05, throughput 6.14207K wps
[Epoch 32 Batch 60/173] avg loss 1.40464e-05, throughput 5.98548K wps
[Epoch 32 Batch 90/173] avg loss 1.35721e-05, throughput 6.00034K wps
[Epoch 32 Batch 120/173] avg loss 2.12294e-05, throughput 6.00317K wps
[Epoch 32 Batch 150/173] avg loss 1.46601e-05, throughput 5.98996K wps
Begin Testing...
[Epoch 32] train avg loss 1.76382e-05, test acc 0.7323, test avg loss 1.36941, throughput 6.02077K wps
[Epoch 33 Batch 30/173] avg loss 1.44088e-05, throughput 6.12956K wps
[Epoch 33 Batch 60/173] avg loss 1.16258e-05, throughput 5.98699K wps
[Epoch 33 Batch 90/173] avg loss 1.58093e-05, throughput 5.98983K wps
[Epoch 33 Batch 120/173] avg loss 1.55772e-05, throughput 5.98283K wps
[Epoch 33 Batch 150/173] avg loss 2.02612e-05, throughput 5.98205K wps
Begin Testing...
[Epoch 33] train avg loss 1.53314e-05, test acc 0.7333, test avg loss 1.38951, throughput 6.01215K wps
[Epoch 34 Batch 30/173] avg loss 1.44001e-05, throughput 6.12443K wps
[Epoch 34 Batch 60/173] avg loss 1.5174e-05, throughput 5.98223K wps
[Epoch 34 Batch 90/173] avg loss 1.14526e-05, throughput 5.99776K wps
[Epoch 34 Batch 120/173] avg loss 1.19348e-05, throughput 6.00153K wps
[Epoch 34 Batch 150/173] avg loss 1.62328e-05, throughput 5.99748K wps
Begin Testing...
[Epoch 34] train avg loss 1.51574e-05, test acc 0.7302, test avg loss 1.41105, throughput 6.01602K wps
[Epoch 35 Batch 30/173] avg loss 1.27051e-05, throughput 6.14215K wps
[Epoch 35 Batch 60/173] avg loss 1.14384e-05, throughput 5.98231K wps
[Epoch 35 Batch 90/173] avg loss 1.8643e-05, throughput 5.97761K wps
[Epoch 35 Batch 120/173] avg loss 1.14251e-05, throughput 5.97151K wps
[Epoch 35 Batch 150/173] avg loss 1.10639e-05, throughput 5.98223K wps
Begin Testing...
[Epoch 35] train avg loss 1.30314e-05, test acc 0.7302, test avg loss 1.43242, throughput 6.01085K wps
[Epoch 36 Batch 30/173] avg loss 7.86418e-06, throughput 6.13947K wps
[Epoch 36 Batch 60/173] avg loss 9.75293e-06, throughput 5.97789K wps
[Epoch 36 Batch 90/173] avg loss 1.81906e-05, throughput 5.98558K wps
[Epoch 36 Batch 120/173] avg loss 8.7299e-06, throughput 5.98161K wps
[Epoch 36 Batch 150/173] avg loss 1.06002e-05, throughput 5.97477K wps
Begin Testing...
[Epoch 36] train avg loss 1.04339e-05, test acc 0.7312, test avg loss 1.45486, throughput 6.00865K wps
[Epoch 37 Batch 30/173] avg loss 7.065e-06, throughput 6.13868K wps
[Epoch 37 Batch 60/173] avg loss 7.4581e-06, throughput 5.98943K wps
[Epoch 37 Batch 90/173] avg loss 8.74416e-06, throughput 5.98467K wps
[Epoch 37 Batch 120/173] avg loss 9.59655e-06, throughput 5.98342K wps
[Epoch 37 Batch 150/173] avg loss 1.59415e-05, throughput 5.98734K wps
Begin Testing...
[Epoch 37] train avg loss 9.54104e-06, test acc 0.7333, test avg loss 1.47634, throughput 6.01476K wps
[Epoch 38 Batch 30/173] avg loss 7.91142e-06, throughput 6.13068K wps
[Epoch 38 Batch 60/173] avg loss 8.62791e-06, throughput 5.98322K wps
[Epoch 38 Batch 90/173] avg loss 7.07691e-06, throughput 5.98686K wps
[Epoch 38 Batch 120/173] avg loss 1.3902e-05, throughput 5.98015K wps
[Epoch 38 Batch 150/173] avg loss 7.23279e-06, throughput 5.96722K wps
Begin Testing...
[Epoch 38] train avg loss 8.51143e-06, test acc 0.7333, test avg loss 1.4966, throughput 6.00522K wps
[Epoch 39 Batch 30/173] avg loss 6.83816e-06, throughput 6.14652K wps
[Epoch 39 Batch 60/173] avg loss 5.54544e-06, throughput 5.98426K wps
[Epoch 39 Batch 90/173] avg loss 2.04524e-05, throughput 5.98751K wps
[Epoch 39 Batch 120/173] avg loss 5.99808e-06, throughput 5.98311K wps
[Epoch 39 Batch 150/173] avg loss 1.31071e-05, throughput 5.98579K wps
Begin Testing...
[Epoch 39] train avg loss 9.99994e-06, test acc 0.7271, test avg loss 1.52891, throughput 6.0119K wps
[Epoch 40 Batch 30/173] avg loss 6.844e-06, throughput 6.14429K wps
[Epoch 40 Batch 60/173] avg loss 7.02955e-06, throughput 5.98123K wps
[Epoch 40 Batch 90/173] avg loss 7.32996e-06, throughput 5.98736K wps
[Epoch 40 Batch 120/173] avg loss 5.46487e-06, throughput 5.97424K wps
[Epoch 40 Batch 150/173] avg loss 1.26582e-05, throughput 5.9905K wps
Begin Testing...
[Epoch 40] train avg loss 7.92651e-06, test acc 0.7260, test avg loss 1.5346, throughput 6.01319K wps
[Epoch 41 Batch 30/173] avg loss 7.35088e-06, throughput 6.11499K wps
[Epoch 41 Batch 60/173] avg loss 5.94709e-06, throughput 5.99177K wps
[Epoch 41 Batch 90/173] avg loss 6.65772e-06, throughput 5.98341K wps
[Epoch 41 Batch 120/173] avg loss 8.99943e-06, throughput 5.99711K wps
[Epoch 41 Batch 150/173] avg loss 9.39054e-06, throughput 5.99631K wps
Begin Testing...
[Epoch 41] train avg loss 7.56277e-06, test acc 0.7292, test avg loss 1.55204, throughput 6.01174K wps
[Epoch 42 Batch 30/173] avg loss 5.74196e-06, throughput 6.13808K wps
[Epoch 42 Batch 60/173] avg loss 6.20524e-06, throughput 5.99716K wps
[Epoch 42 Batch 90/173] avg loss 4.88336e-06, throughput 5.98683K wps
[Epoch 42 Batch 120/173] avg loss 9.93284e-06, throughput 5.98677K wps
[Epoch 42 Batch 150/173] avg loss 4.92268e-06, throughput 5.98964K wps
Begin Testing...
[Epoch 42] train avg loss 6.41866e-06, test acc 0.7271, test avg loss 1.57449, throughput 6.0158K wps
[Epoch 43 Batch 30/173] avg loss 6.65683e-06, throughput 6.12818K wps
[Epoch 43 Batch 60/173] avg loss 4.27794e-06, throughput 5.98102K wps
[Epoch 43 Batch 90/173] avg loss 6.93859e-06, throughput 5.9768K wps
[Epoch 43 Batch 120/173] avg loss 3.9135e-06, throughput 5.97498K wps
[Epoch 43 Batch 150/173] avg loss 4.15945e-06, throughput 5.97631K wps
Begin Testing...
[Epoch 43] train avg loss 5.1186e-06, test acc 0.7292, test avg loss 1.59503, throughput 6.00461K wps
[Epoch 44 Batch 30/173] avg loss 3.9264e-06, throughput 6.13637K wps
[Epoch 44 Batch 60/173] avg loss 4.32276e-06, throughput 5.99284K wps
[Epoch 44 Batch 90/173] avg loss 3.94375e-06, throughput 5.98129K wps
[Epoch 44 Batch 120/173] avg loss 3.99597e-06, throughput 5.97733K wps
[Epoch 44 Batch 150/173] avg loss 3.5349e-06, throughput 5.9715K wps
Begin Testing...
[Epoch 44] train avg loss 4.1729e-06, test acc 0.7312, test avg loss 1.60954, throughput 6.00823K wps
[Epoch 45 Batch 30/173] avg loss 3.80617e-06, throughput 6.12928K wps
[Epoch 45 Batch 60/173] avg loss 4.15783e-06, throughput 5.97535K wps
[Epoch 45 Batch 90/173] avg loss 5.35633e-06, throughput 5.9778K wps
[Epoch 45 Batch 120/173] avg loss 3.73396e-06, throughput 5.99561K wps
[Epoch 45 Batch 150/173] avg loss 5.39394e-06, throughput 5.98157K wps
Begin Testing...
[Epoch 45] train avg loss 4.36285e-06, test acc 0.7312, test avg loss 1.62603, throughput 6.00876K wps
[Epoch 46 Batch 30/173] avg loss 3.25763e-06, throughput 6.13811K wps
[Epoch 46 Batch 60/173] avg loss 2.80068e-06, throughput 5.98396K wps
[Epoch 46 Batch 90/173] avg loss 3.69638e-06, throughput 5.97615K wps
[Epoch 46 Batch 120/173] avg loss 3.04738e-06, throughput 5.98163K wps
[Epoch 46 Batch 150/173] avg loss 4.75122e-06, throughput 5.98578K wps
Begin Testing...
[Epoch 46] train avg loss 3.55189e-06, test acc 0.7312, test avg loss 1.65164, throughput 6.01K wps
[Epoch 47 Batch 30/173] avg loss 2.81598e-06, throughput 6.13456K wps
[Epoch 47 Batch 60/173] avg loss 2.89429e-06, throughput 5.98788K wps
[Epoch 47 Batch 90/173] avg loss 4.05725e-06, throughput 5.98522K wps
[Epoch 47 Batch 120/173] avg loss 2.86975e-06, throughput 5.9831K wps
[Epoch 47 Batch 150/173] avg loss 3.6331e-06, throughput 5.98442K wps
Begin Testing...
[Epoch 47] train avg loss 3.41572e-06, test acc 0.7292, test avg loss 1.68471, throughput 6.01027K wps
[Epoch 48 Batch 30/173] avg loss 3.96736e-06, throughput 6.13943K wps
[Epoch 48 Batch 60/173] avg loss 3.49233e-06, throughput 5.98903K wps
[Epoch 48 Batch 90/173] avg loss 2.75909e-06, throughput 5.99929K wps
[Epoch 48 Batch 120/173] avg loss 3.15808e-06, throughput 5.98766K wps
[Epoch 48 Batch 150/173] avg loss 2.60807e-06, throughput 5.98376K wps
Begin Testing...
[Epoch 48] train avg loss 3.22487e-06, test acc 0.7292, test avg loss 1.68896, throughput 6.0166K wps
[Epoch 49 Batch 30/173] avg loss 3.31477e-06, throughput 6.13167K wps
[Epoch 49 Batch 60/173] avg loss 2.70423e-06, throughput 5.97998K wps
[Epoch 49 Batch 90/173] avg loss 2.71523e-06, throughput 5.98342K wps
[Epoch 49 Batch 120/173] avg loss 3.08309e-06, throughput 5.9898K wps
[Epoch 49 Batch 150/173] avg loss 2.42602e-06, throughput 5.98805K wps
Begin Testing...
[Epoch 49] train avg loss 3.04945e-06, test acc 0.7292, test avg loss 1.70424, throughput 6.00996K wps
[Epoch 50 Batch 30/173] avg loss 1.9033e-06, throughput 6.12848K wps
[Epoch 50 Batch 60/173] avg loss 2.3362e-06, throughput 5.99424K wps
[Epoch 50 Batch 90/173] avg loss 2.76743e-06, throughput 5.98501K wps
[Epoch 50 Batch 120/173] avg loss 2.99615e-06, throughput 5.9837K wps
[Epoch 50 Batch 150/173] avg loss 2.71909e-06, throughput 5.98006K wps
Begin Testing...
[Epoch 50] train avg loss 2.4849e-06, test acc 0.7344, test avg loss 1.72557, throughput 6.01071K wps
[Epoch 51 Batch 30/173] avg loss 2.18744e-06, throughput 6.14228K wps
[Epoch 51 Batch 60/173] avg loss 2.36385e-06, throughput 5.97862K wps
[Epoch 51 Batch 90/173] avg loss 2.46211e-06, throughput 5.99565K wps
[Epoch 51 Batch 120/173] avg loss 2.89507e-06, throughput 5.98554K wps
[Epoch 51 Batch 150/173] avg loss 1.65069e-06, throughput 5.97616K wps
Begin Testing...
[Epoch 51] train avg loss 2.25797e-06, test acc 0.7292, test avg loss 1.74713, throughput 6.01269K wps
[Epoch 52 Batch 30/173] avg loss 1.54129e-06, throughput 6.14128K wps
[Epoch 52 Batch 60/173] avg loss 2.72149e-06, throughput 5.98924K wps
[Epoch 52 Batch 90/173] avg loss 1.60584e-06, throughput 5.98358K wps
[Epoch 52 Batch 120/173] avg loss 1.94147e-06, throughput 5.99179K wps
[Epoch 52 Batch 150/173] avg loss 1.49858e-06, throughput 6.00129K wps
Begin Testing...
[Epoch 52] train avg loss 1.86081e-06, test acc 0.7271, test avg loss 1.76851, throughput 6.01892K wps
[Epoch 53 Batch 30/173] avg loss 1.52476e-06, throughput 6.13916K wps
[Epoch 53 Batch 60/173] avg loss 1.97465e-06, throughput 5.99425K wps
[Epoch 53 Batch 90/173] avg loss 3.02768e-06, throughput 5.99859K wps
[Epoch 53 Batch 120/173] avg loss 1.767e-06, throughput 5.98434K wps
[Epoch 53 Batch 150/173] avg loss 1.57487e-06, throughput 5.98889K wps
Begin Testing...
[Epoch 53] train avg loss 1.98784e-06, test acc 0.7302, test avg loss 1.78459, throughput 6.01735K wps
[Epoch 54 Batch 30/173] avg loss 1.42349e-06, throughput 6.14671K wps
[Epoch 54 Batch 60/173] avg loss 2.31968e-06, throughput 5.98692K wps
[Epoch 54 Batch 90/173] avg loss 1.91881e-06, throughput 5.98796K wps
[Epoch 54 Batch 120/173] avg loss 1.328e-06, throughput 5.99868K wps
[Epoch 54 Batch 150/173] avg loss 2.75652e-06, throughput 5.9883K wps
Begin Testing...
[Epoch 54] train avg loss 1.89457e-06, test acc 0.7219, test avg loss 1.81046, throughput 6.01963K wps
[Epoch 55 Batch 30/173] avg loss 1.29239e-06, throughput 6.12688K wps
[Epoch 55 Batch 60/173] avg loss 1.99789e-06, throughput 5.99293K wps
[Epoch 55 Batch 90/173] avg loss 1.40516e-06, throughput 5.99457K wps
[Epoch 55 Batch 120/173] avg loss 1.84526e-06, throughput 5.98719K wps
[Epoch 55 Batch 150/173] avg loss 1.51577e-06, throughput 5.98187K wps
Begin Testing...
[Epoch 55] train avg loss 1.57328e-06, test acc 0.7271, test avg loss 1.82701, throughput 6.01608K wps
[Epoch 56 Batch 30/173] avg loss 1.25485e-06, throughput 6.13647K wps
[Epoch 56 Batch 60/173] avg loss 1.0867e-06, throughput 6.0002K wps
[Epoch 56 Batch 90/173] avg loss 2.01311e-06, throughput 5.98128K wps
[Epoch 56 Batch 120/173] avg loss 1.25311e-06, throughput 5.99571K wps
[Epoch 56 Batch 150/173] avg loss 2.08564e-06, throughput 5.99009K wps
Begin Testing...
[Epoch 56] train avg loss 1.46108e-06, test acc 0.7333, test avg loss 1.84328, throughput 6.01637K wps
[Epoch 57 Batch 30/173] avg loss 1.56328e-06, throughput 6.12194K wps
[Epoch 57 Batch 60/173] avg loss 1.29717e-06, throughput 5.98984K wps
[Epoch 57 Batch 90/173] avg loss 1.07565e-06, throughput 5.99207K wps
[Epoch 57 Batch 120/173] avg loss 1.16233e-06, throughput 5.98744K wps
[Epoch 57 Batch 150/173] avg loss 1.07496e-06, throughput 5.9955K wps
Begin Testing...
[Epoch 57] train avg loss 1.21417e-06, test acc 0.7323, test avg loss 1.85966, throughput 6.01409K wps
[Epoch 58 Batch 30/173] avg loss 1.02962e-06, throughput 6.13778K wps
[Epoch 58 Batch 60/173] avg loss 1.22551e-06, throughput 5.9796K wps
[Epoch 58 Batch 90/173] avg loss 1.22907e-06, throughput 5.98567K wps
[Epoch 58 Batch 120/173] avg loss 9.96043e-07, throughput 5.99194K wps
[Epoch 58 Batch 150/173] avg loss 1.43533e-06, throughput 5.97838K wps
Begin Testing...
[Epoch 58] train avg loss 1.18e-06, test acc 0.7302, test avg loss 1.87859, throughput 6.00958K wps
[Epoch 59 Batch 30/173] avg loss 9.72038e-07, throughput 6.13605K wps
[Epoch 59 Batch 60/173] avg loss 1.11272e-06, throughput 5.98965K wps
[Epoch 59 Batch 90/173] avg loss 7.9458e-07, throughput 5.98K wps
[Epoch 59 Batch 120/173] avg loss 1.99416e-06, throughput 5.98407K wps
[Epoch 59 Batch 150/173] avg loss 1.37681e-06, throughput 5.99715K wps
Begin Testing...
[Epoch 59] train avg loss 1.21963e-06, test acc 0.7302, test avg loss 1.89753, throughput 6.01494K wps
Test loss 0.49, test acc 0.7674
Total time cost 358.32s
9596 1066
[Epoch 0 Batch 30/173] avg loss 0.0138433, throughput 5.79821K wps
[Epoch 0 Batch 60/173] avg loss 0.0138323, throughput 5.98223K wps
[Epoch 0 Batch 90/173] avg loss 0.0138176, throughput 5.99107K wps
[Epoch 0 Batch 120/173] avg loss 0.0137714, throughput 5.98898K wps
[Epoch 0 Batch 150/173] avg loss 0.013744, throughput 5.9915K wps
Begin Testing...
[Epoch 0] train avg loss 0.0138116, test acc 0.6323, test avg loss 0.685456, throughput 5.95431K wps
Observed Improvement.
Begin Testing...
[Epoch 1 Batch 30/173] avg loss 0.0134814, throughput 6.12796K wps
[Epoch 1 Batch 60/173] avg loss 0.0134413, throughput 5.98425K wps
[Epoch 1 Batch 90/173] avg loss 0.0133911, throughput 5.99584K wps
[Epoch 1 Batch 120/173] avg loss 0.0133388, throughput 5.98598K wps
[Epoch 1 Batch 150/173] avg loss 0.0132841, throughput 5.99891K wps
Begin Testing...
[Epoch 1] train avg loss 0.013372, test acc 0.6813, test avg loss 0.668097, throughput 6.01325K wps
Observed Improvement.
Begin Testing...
[Epoch 2 Batch 30/173] avg loss 0.012821, throughput 6.13602K wps
[Epoch 2 Batch 60/173] avg loss 0.0126818, throughput 6.00544K wps
[Epoch 2 Batch 90/173] avg loss 0.0125009, throughput 6.00612K wps
[Epoch 2 Batch 120/173] avg loss 0.0123551, throughput 6.00567K wps
[Epoch 2 Batch 150/173] avg loss 0.0122499, throughput 6.00548K wps
Begin Testing...
[Epoch 2] train avg loss 0.0124803, test acc 0.7125, test avg loss 0.622159, throughput 6.02628K wps
Observed Improvement.
Begin Testing...
[Epoch 3 Batch 30/173] avg loss 0.0113183, throughput 6.15062K wps
[Epoch 3 Batch 60/173] avg loss 0.0110575, throughput 5.98812K wps
[Epoch 3 Batch 90/173] avg loss 0.0106133, throughput 5.98976K wps
[Epoch 3 Batch 120/173] avg loss 0.0104751, throughput 5.99475K wps
[Epoch 3 Batch 150/173] avg loss 0.0104545, throughput 5.99721K wps
Begin Testing...
[Epoch 3] train avg loss 0.0107102, test acc 0.7562, test avg loss 0.550947, throughput 6.01834K wps
Observed Improvement.
Begin Testing...
[Epoch 4 Batch 30/173] avg loss 0.00891323, throughput 6.14899K wps
[Epoch 4 Batch 60/173] avg loss 0.00849604, throughput 5.99168K wps
[Epoch 4 Batch 90/173] avg loss 0.00827081, throughput 5.98152K wps
[Epoch 4 Batch 120/173] avg loss 0.00803469, throughput 5.97454K wps
[Epoch 4 Batch 150/173] avg loss 0.00782824, throughput 5.98412K wps
Begin Testing...
[Epoch 4] train avg loss 0.00825821, test acc 0.7812, test avg loss 0.498745, throughput 6.01252K wps
Observed Improvement.
Begin Testing...
[Epoch 5 Batch 30/173] avg loss 0.00652781, throughput 6.13127K wps
[Epoch 5 Batch 60/173] avg loss 0.00616077, throughput 5.99043K wps
[Epoch 5 Batch 90/173] avg loss 0.00583149, throughput 5.99498K wps
[Epoch 5 Batch 120/173] avg loss 0.00583714, throughput 5.99272K wps
[Epoch 5 Batch 150/173] avg loss 0.00574762, throughput 6.00329K wps
Begin Testing...
[Epoch 5] train avg loss 0.00595066, test acc 0.7792, test avg loss 0.481479, throughput 6.01876K wps
[Epoch 6 Batch 30/173] avg loss 0.00429705, throughput 6.14043K wps
[Epoch 6 Batch 60/173] avg loss 0.00423309, throughput 6.00613K wps
[Epoch 6 Batch 90/173] avg loss 0.00435747, throughput 6.00234K wps
[Epoch 6 Batch 120/173] avg loss 0.0043001, throughput 5.99058K wps
[Epoch 6 Batch 150/173] avg loss 0.00404478, throughput 5.99507K wps
Begin Testing...
[Epoch 6] train avg loss 0.00420361, test acc 0.7792, test avg loss 0.494532, throughput 6.02195K wps
[Epoch 7 Batch 30/173] avg loss 0.00279551, throughput 6.1279K wps
[Epoch 7 Batch 60/173] avg loss 0.00306257, throughput 5.99332K wps
[Epoch 7 Batch 90/173] avg loss 0.00271707, throughput 5.99373K wps
[Epoch 7 Batch 120/173] avg loss 0.00298232, throughput 5.99023K wps
[Epoch 7 Batch 150/173] avg loss 0.00275233, throughput 5.98944K wps
Begin Testing...
[Epoch 7] train avg loss 0.00290865, test acc 0.7750, test avg loss 0.519692, throughput 6.01625K wps
[Epoch 8 Batch 30/173] avg loss 0.00193278, throughput 6.14364K wps
[Epoch 8 Batch 60/173] avg loss 0.00185855, throughput 5.98632K wps
[Epoch 8 Batch 90/173] avg loss 0.00186939, throughput 5.99796K wps
[Epoch 8 Batch 120/173] avg loss 0.00215294, throughput 5.99048K wps
[Epoch 8 Batch 150/173] avg loss 0.00209032, throughput 6.00285K wps
Begin Testing...
[Epoch 8] train avg loss 0.00199188, test acc 0.7719, test avg loss 0.559981, throughput 6.02174K wps
[Epoch 9 Batch 30/173] avg loss 0.00132673, throughput 6.12964K wps
[Epoch 9 Batch 60/173] avg loss 0.00123993, throughput 5.98975K wps
[Epoch 9 Batch 90/173] avg loss 0.00150017, throughput 5.98046K wps
[Epoch 9 Batch 120/173] avg loss 0.0014832, throughput 5.98295K wps
[Epoch 9 Batch 150/173] avg loss 0.00146653, throughput 5.98842K wps
Begin Testing...
[Epoch 9] train avg loss 0.00139623, test acc 0.7688, test avg loss 0.599769, throughput 6.01222K wps
[Epoch 10 Batch 30/173] avg loss 0.000988405, throughput 6.13515K wps
[Epoch 10 Batch 60/173] avg loss 0.000983142, throughput 5.99371K wps
[Epoch 10 Batch 90/173] avg loss 0.000920489, throughput 6.00529K wps
[Epoch 10 Batch 120/173] avg loss 0.000931339, throughput 5.99961K wps
[Epoch 10 Batch 150/173] avg loss 0.00104779, throughput 5.9924K wps
Begin Testing...
[Epoch 10] train avg loss 0.000998583, test acc 0.7729, test avg loss 0.642928, throughput 6.02026K wps
[Epoch 11 Batch 30/173] avg loss 0.000731138, throughput 6.16016K wps
[Epoch 11 Batch 60/173] avg loss 0.000761287, throughput 6.00846K wps
[Epoch 11 Batch 90/173] avg loss 0.000738461, throughput 5.99474K wps
[Epoch 11 Batch 120/173] avg loss 0.000801469, throughput 5.99624K wps
[Epoch 11 Batch 150/173] avg loss 0.000633929, throughput 5.99688K wps
Begin Testing...
[Epoch 11] train avg loss 0.000736651, test acc 0.7594, test avg loss 0.688336, throughput 6.0257K wps
[Epoch 12 Batch 30/173] avg loss 0.000480723, throughput 6.13466K wps
[Epoch 12 Batch 60/173] avg loss 0.000536141, throughput 5.99323K wps
[Epoch 12 Batch 90/173] avg loss 0.000536551, throughput 5.99345K wps
[Epoch 12 Batch 120/173] avg loss 0.000528912, throughput 6.00475K wps
[Epoch 12 Batch 150/173] avg loss 0.000522415, throughput 5.99279K wps
Begin Testing...
[Epoch 12] train avg loss 0.000538313, test acc 0.7688, test avg loss 0.725708, throughput 6.01786K wps
[Epoch 13 Batch 30/173] avg loss 0.000406921, throughput 6.14982K wps
[Epoch 13 Batch 60/173] avg loss 0.000428065, throughput 6.00475K wps
[Epoch 13 Batch 90/173] avg loss 0.00037095, throughput 5.9968K wps
[Epoch 13 Batch 120/173] avg loss 0.000418833, throughput 5.99484K wps
[Epoch 13 Batch 150/173] avg loss 0.000373942, throughput 5.99825K wps
Begin Testing...
[Epoch 13] train avg loss 0.000402073, test acc 0.7625, test avg loss 0.768806, throughput 6.0262K wps
[Epoch 14 Batch 30/173] avg loss 0.000293804, throughput 6.14255K wps
[Epoch 14 Batch 60/173] avg loss 0.000266031, throughput 5.99249K wps
[Epoch 14 Batch 90/173] avg loss 0.000355984, throughput 5.97682K wps
[Epoch 14 Batch 120/173] avg loss 0.000311307, throughput 5.9836K wps
[Epoch 14 Batch 150/173] avg loss 0.000298226, throughput 5.99146K wps
Begin Testing...
[Epoch 14] train avg loss 0.000313302, test acc 0.7604, test avg loss 0.807047, throughput 6.01154K wps
[Epoch 15 Batch 30/173] avg loss 0.000258732, throughput 6.13162K wps
[Epoch 15 Batch 60/173] avg loss 0.000223075, throughput 5.98668K wps
[Epoch 15 Batch 90/173] avg loss 0.000240169, throughput 5.99769K wps
[Epoch 15 Batch 120/173] avg loss 0.000251875, throughput 5.99114K wps
[Epoch 15 Batch 150/173] avg loss 0.000213861, throughput 5.99261K wps
Begin Testing...
[Epoch 15] train avg loss 0.000239175, test acc 0.7615, test avg loss 0.844372, throughput 6.01716K wps
[Epoch 16 Batch 30/173] avg loss 0.000213292, throughput 6.12986K wps
[Epoch 16 Batch 60/173] avg loss 0.000206662, throughput 5.98778K wps
[Epoch 16 Batch 90/173] avg loss 0.00016164, throughput 5.99498K wps
[Epoch 16 Batch 120/173] avg loss 0.000187868, throughput 5.99507K wps
[Epoch 16 Batch 150/173] avg loss 0.000200841, throughput 5.99218K wps
Begin Testing...
[Epoch 16] train avg loss 0.000197526, test acc 0.7604, test avg loss 0.876948, throughput 6.014K wps
[Epoch 17 Batch 30/173] avg loss 0.000178176, throughput 6.12835K wps
[Epoch 17 Batch 60/173] avg loss 0.000150402, throughput 5.97762K wps
[Epoch 17 Batch 90/173] avg loss 0.000125777, throughput 5.98406K wps
[Epoch 17 Batch 120/173] avg loss 0.000159787, throughput 5.98579K wps
[Epoch 17 Batch 150/173] avg loss 0.000154039, throughput 5.99398K wps
Begin Testing...
[Epoch 17] train avg loss 0.000153738, test acc 0.7594, test avg loss 0.912888, throughput 6.0107K wps
[Epoch 18 Batch 30/173] avg loss 0.000126336, throughput 6.13878K wps
[Epoch 18 Batch 60/173] avg loss 0.000111909, throughput 5.99319K wps
[Epoch 18 Batch 90/173] avg loss 0.000120956, throughput 5.9813K wps
[Epoch 18 Batch 120/173] avg loss 0.000138702, throughput 5.98245K wps
[Epoch 18 Batch 150/173] avg loss 0.000143759, throughput 5.98621K wps
Begin Testing...
[Epoch 18] train avg loss 0.000132541, test acc 0.7594, test avg loss 0.943988, throughput 6.01293K wps
[Epoch 19 Batch 30/173] avg loss 0.000115209, throughput 6.13706K wps
[Epoch 19 Batch 60/173] avg loss 0.00010889, throughput 6.00025K wps
[Epoch 19 Batch 90/173] avg loss 0.000103213, throughput 5.97623K wps
[Epoch 19 Batch 120/173] avg loss 8.67704e-05, throughput 5.98734K wps
[Epoch 19 Batch 150/173] avg loss 0.000126246, throughput 6.00319K wps
Begin Testing...
[Epoch 19] train avg loss 0.00010616, test acc 0.7573, test avg loss 0.974811, throughput 6.0184K wps
[Epoch 20 Batch 30/173] avg loss 8.03695e-05, throughput 6.15124K wps
[Epoch 20 Batch 60/173] avg loss 8.37648e-05, throughput 6.0029K wps
[Epoch 20 Batch 90/173] avg loss 8.08184e-05, throughput 6.01212K wps
[Epoch 20 Batch 120/173] avg loss 0.000105769, throughput 5.99685K wps
[Epoch 20 Batch 150/173] avg loss 0.000106385, throughput 5.98369K wps
Begin Testing...
[Epoch 20] train avg loss 9.37285e-05, test acc 0.7594, test avg loss 1.00439, throughput 6.02336K wps
[Epoch 21 Batch 30/173] avg loss 8.75599e-05, throughput 6.13399K wps
[Epoch 21 Batch 60/173] avg loss 8.61692e-05, throughput 5.98635K wps
[Epoch 21 Batch 90/173] avg loss 8.41542e-05, throughput 5.97438K wps
[Epoch 21 Batch 120/173] avg loss 7.08282e-05, throughput 5.97617K wps
[Epoch 21 Batch 150/173] avg loss 8.26127e-05, throughput 5.99094K wps
Begin Testing...
[Epoch 21] train avg loss 7.98668e-05, test acc 0.7531, test avg loss 1.03785, throughput 6.0081K wps
[Epoch 22 Batch 30/173] avg loss 5.65244e-05, throughput 6.14186K wps
[Epoch 22 Batch 60/173] avg loss 6.7693e-05, throughput 5.98593K wps
[Epoch 22 Batch 90/173] avg loss 6.52948e-05, throughput 5.97649K wps
[Epoch 22 Batch 120/173] avg loss 7.34602e-05, throughput 5.99485K wps
[Epoch 22 Batch 150/173] avg loss 6.0231e-05, throughput 5.97236K wps
Begin Testing...
[Epoch 22] train avg loss 6.48517e-05, test acc 0.7500, test avg loss 1.0624, throughput 6.01085K wps
[Epoch 23 Batch 30/173] avg loss 5.1055e-05, throughput 6.12245K wps
[Epoch 23 Batch 60/173] avg loss 5.09696e-05, throughput 5.98217K wps
[Epoch 23 Batch 90/173] avg loss 5.46546e-05, throughput 5.98734K wps
[Epoch 23 Batch 120/173] avg loss 5.75015e-05, throughput 5.98228K wps
[Epoch 23 Batch 150/173] avg loss 4.91724e-05, throughput 5.98857K wps
Begin Testing...
[Epoch 23] train avg loss 5.45815e-05, test acc 0.7552, test avg loss 1.0842, throughput 6.01074K wps
[Epoch 24 Batch 30/173] avg loss 4.18679e-05, throughput 6.13364K wps
[Epoch 24 Batch 60/173] avg loss 3.88313e-05, throughput 5.97987K wps
[Epoch 24 Batch 90/173] avg loss 4.85991e-05, throughput 5.98066K wps
[Epoch 24 Batch 120/173] avg loss 4.96609e-05, throughput 5.98705K wps
[Epoch 24 Batch 150/173] avg loss 4.66742e-05, throughput 5.97767K wps
Begin Testing...
[Epoch 24] train avg loss 4.67111e-05, test acc 0.7531, test avg loss 1.11092, throughput 6.00896K wps
[Epoch 25 Batch 30/173] avg loss 3.82311e-05, throughput 6.13604K wps
[Epoch 25 Batch 60/173] avg loss 3.96691e-05, throughput 5.97976K wps
[Epoch 25 Batch 90/173] avg loss 3.6366e-05, throughput 5.97813K wps
[Epoch 25 Batch 120/173] avg loss 4.18472e-05, throughput 5.98149K wps
[Epoch 25 Batch 150/173] avg loss 3.96179e-05, throughput 5.97455K wps
Begin Testing...
[Epoch 25] train avg loss 4.23704e-05, test acc 0.7542, test avg loss 1.13976, throughput 6.0066K wps
[Epoch 26 Batch 30/173] avg loss 3.33443e-05, throughput 6.13032K wps
[Epoch 26 Batch 60/173] avg loss 3.11928e-05, throughput 5.9853K wps
[Epoch 26 Batch 90/173] avg loss 4.4034e-05, throughput 5.97625K wps
[Epoch 26 Batch 120/173] avg loss 4.44787e-05, throughput 5.98363K wps
[Epoch 26 Batch 150/173] avg loss 3.14691e-05, throughput 5.99107K wps
Begin Testing...
[Epoch 26] train avg loss 3.77719e-05, test acc 0.7542, test avg loss 1.16797, throughput 6.0111K wps
[Epoch 27 Batch 30/173] avg loss 3.213e-05, throughput 6.13881K wps
[Epoch 27 Batch 60/173] avg loss 2.96755e-05, throughput 5.99993K wps
[Epoch 27 Batch 90/173] avg loss 2.56384e-05, throughput 5.98701K wps
[Epoch 27 Batch 120/173] avg loss 2.99165e-05, throughput 5.98361K wps
[Epoch 27 Batch 150/173] avg loss 3.60033e-05, throughput 5.98637K wps
Begin Testing...
[Epoch 27] train avg loss 3.24618e-05, test acc 0.7552, test avg loss 1.19482, throughput 6.0159K wps
[Epoch 28 Batch 30/173] avg loss 3.35664e-05, throughput 6.14057K wps
[Epoch 28 Batch 60/173] avg loss 2.33576e-05, throughput 5.98165K wps
[Epoch 28 Batch 90/173] avg loss 2.34392e-05, throughput 5.98896K wps
[Epoch 28 Batch 120/173] avg loss 3.02e-05, throughput 5.99315K wps
[Epoch 28 Batch 150/173] avg loss 2.49926e-05, throughput 5.98258K wps
Begin Testing...
[Epoch 28] train avg loss 2.6612e-05, test acc 0.7594, test avg loss 1.21527, throughput 6.01206K wps
[Epoch 29 Batch 30/173] avg loss 2.46498e-05, throughput 6.12414K wps
[Epoch 29 Batch 60/173] avg loss 2.35498e-05, throughput 5.98923K wps
[Epoch 29 Batch 90/173] avg loss 3.47958e-05, throughput 5.98413K wps
[Epoch 29 Batch 120/173] avg loss 2.32106e-05, throughput 5.98788K wps
[Epoch 29 Batch 150/173] avg loss 1.95509e-05, throughput 6.00146K wps
Begin Testing...
[Epoch 29] train avg loss 2.48322e-05, test acc 0.7531, test avg loss 1.23998, throughput 6.01648K wps
[Epoch 30 Batch 30/173] avg loss 2.13001e-05, throughput 6.14111K wps
[Epoch 30 Batch 60/173] avg loss 2.84186e-05, throughput 6.00004K wps
[Epoch 30 Batch 90/173] avg loss 1.64259e-05, throughput 5.99857K wps
[Epoch 30 Batch 120/173] avg loss 2.25821e-05, throughput 5.99504K wps
[Epoch 30 Batch 150/173] avg loss 1.94092e-05, throughput 5.97952K wps
Begin Testing...
[Epoch 30] train avg loss 2.19511e-05, test acc 0.7573, test avg loss 1.25987, throughput 6.01923K wps
[Epoch 31 Batch 30/173] avg loss 1.68458e-05, throughput 6.13165K wps
[Epoch 31 Batch 60/173] avg loss 2.01118e-05, throughput 5.99054K wps
[Epoch 31 Batch 90/173] avg loss 2.42978e-05, throughput 5.98965K wps
[Epoch 31 Batch 120/173] avg loss 1.71349e-05, throughput 5.9813K wps
[Epoch 31 Batch 150/173] avg loss 1.77415e-05, throughput 5.98355K wps
Begin Testing...
[Epoch 31] train avg loss 1.93692e-05, test acc 0.7542, test avg loss 1.28149, throughput 6.01333K wps
[Epoch 32 Batch 30/173] avg loss 1.56648e-05, throughput 6.14103K wps
[Epoch 32 Batch 60/173] avg loss 1.4968e-05, throughput 5.98999K wps
[Epoch 32 Batch 90/173] avg loss 2.09773e-05, throughput 5.99903K wps
[Epoch 32 Batch 120/173] avg loss 1.50187e-05, throughput 5.99998K wps
[Epoch 32 Batch 150/173] avg loss 2.00342e-05, throughput 5.99739K wps
Begin Testing...
[Epoch 32] train avg loss 1.91614e-05, test acc 0.7521, test avg loss 1.31146, throughput 6.02064K wps
[Epoch 33 Batch 30/173] avg loss 1.46717e-05, throughput 6.13322K wps
[Epoch 33 Batch 60/173] avg loss 1.63654e-05, throughput 5.99305K wps
[Epoch 33 Batch 90/173] avg loss 2.18691e-05, throughput 5.99273K wps
[Epoch 33 Batch 120/173] avg loss 1.7815e-05, throughput 5.98006K wps
[Epoch 33 Batch 150/173] avg loss 1.21226e-05, throughput 5.9855K wps
Begin Testing...
[Epoch 33] train avg loss 1.59921e-05, test acc 0.7510, test avg loss 1.33693, throughput 6.01348K wps
[Epoch 34 Batch 30/173] avg loss 1.74998e-05, throughput 6.13336K wps
[Epoch 34 Batch 60/173] avg loss 1.23307e-05, throughput 5.99902K wps
[Epoch 34 Batch 90/173] avg loss 2.43361e-05, throughput 5.9799K wps
[Epoch 34 Batch 120/173] avg loss 1.30172e-05, throughput 5.98023K wps
[Epoch 34 Batch 150/173] avg loss 1.38512e-05, throughput 5.99095K wps
Begin Testing...
[Epoch 34] train avg loss 1.65285e-05, test acc 0.7458, test avg loss 1.36653, throughput 6.01192K wps
[Epoch 35 Batch 30/173] avg loss 1.17289e-05, throughput 6.14502K wps
[Epoch 35 Batch 60/173] avg loss 1.03869e-05, throughput 5.99333K wps
[Epoch 35 Batch 90/173] avg loss 1.00675e-05, throughput 5.99032K wps
[Epoch 35 Batch 120/173] avg loss 1.37744e-05, throughput 6.00298K wps
[Epoch 35 Batch 150/173] avg loss 2.55097e-05, throughput 5.98722K wps
Begin Testing...
[Epoch 35] train avg loss 1.39633e-05, test acc 0.7521, test avg loss 1.38926, throughput 6.0192K wps
[Epoch 36 Batch 30/173] avg loss 8.03799e-06, throughput 6.13752K wps
[Epoch 36 Batch 60/173] avg loss 9.23562e-06, throughput 5.98972K wps
[Epoch 36 Batch 90/173] avg loss 1.02378e-05, throughput 5.98067K wps
[Epoch 36 Batch 120/173] avg loss 8.66112e-06, throughput 5.981K wps
[Epoch 36 Batch 150/173] avg loss 1.97789e-05, throughput 5.99065K wps
Begin Testing...
[Epoch 36] train avg loss 1.12109e-05, test acc 0.7510, test avg loss 1.41202, throughput 6.01321K wps
[Epoch 37 Batch 30/173] avg loss 9.33928e-06, throughput 6.1422K wps
[Epoch 37 Batch 60/173] avg loss 6.69723e-06, throughput 5.98758K wps
[Epoch 37 Batch 90/173] avg loss 1.65352e-05, throughput 5.98752K wps
[Epoch 37 Batch 120/173] avg loss 8.04818e-06, throughput 5.99664K wps
[Epoch 37 Batch 150/173] avg loss 9.53548e-06, throughput 5.99847K wps
Begin Testing...
[Epoch 37] train avg loss 9.80767e-06, test acc 0.7500, test avg loss 1.42906, throughput 6.01943K wps
[Epoch 38 Batch 30/173] avg loss 7.42462e-06, throughput 6.13778K wps
[Epoch 38 Batch 60/173] avg loss 8.84076e-06, throughput 5.98847K wps
[Epoch 38 Batch 90/173] avg loss 7.73525e-06, throughput 5.99404K wps
[Epoch 38 Batch 120/173] avg loss 6.89113e-06, throughput 5.98805K wps
[Epoch 38 Batch 150/173] avg loss 1.49714e-05, throughput 5.98488K wps
Begin Testing...
[Epoch 38] train avg loss 9.27323e-06, test acc 0.7521, test avg loss 1.44865, throughput 6.0163K wps
[Epoch 39 Batch 30/173] avg loss 1.4777e-05, throughput 6.13558K wps
[Epoch 39 Batch 60/173] avg loss 7.98163e-06, throughput 5.98848K wps
[Epoch 39 Batch 90/173] avg loss 6.94793e-06, throughput 5.99009K wps
[Epoch 39 Batch 120/173] avg loss 6.29434e-06, throughput 5.98807K wps
[Epoch 39 Batch 150/173] avg loss 8.52175e-06, throughput 5.99458K wps
Begin Testing...
[Epoch 39] train avg loss 8.65898e-06, test acc 0.7510, test avg loss 1.46835, throughput 6.01649K wps
[Epoch 40 Batch 30/173] avg loss 5.74837e-06, throughput 6.12244K wps
[Epoch 40 Batch 60/173] avg loss 5.23758e-06, throughput 5.99821K wps
[Epoch 40 Batch 90/173] avg loss 6.4492e-06, throughput 5.98447K wps
[Epoch 40 Batch 120/173] avg loss 7.58711e-06, throughput 5.97812K wps
[Epoch 40 Batch 150/173] avg loss 6.7919e-06, throughput 5.98158K wps
Begin Testing...
[Epoch 40] train avg loss 7.98568e-06, test acc 0.7531, test avg loss 1.48744, throughput 6.00828K wps
[Epoch 41 Batch 30/173] avg loss 6.18899e-06, throughput 6.14334K wps
[Epoch 41 Batch 60/173] avg loss 6.94435e-06, throughput 6.00448K wps
[Epoch 41 Batch 90/173] avg loss 5.91211e-06, throughput 5.98908K wps
[Epoch 41 Batch 120/173] avg loss 1.89982e-05, throughput 5.99332K wps
[Epoch 41 Batch 150/173] avg loss 6.21284e-06, throughput 5.9821K wps
Begin Testing...
[Epoch 41] train avg loss 8.77007e-06, test acc 0.7479, test avg loss 1.51597, throughput 6.01846K wps
[Epoch 42 Batch 30/173] avg loss 4.87644e-06, throughput 6.13221K wps
[Epoch 42 Batch 60/173] avg loss 4.23458e-06, throughput 5.97614K wps
[Epoch 42 Batch 90/173] avg loss 6.2669e-06, throughput 5.98145K wps
[Epoch 42 Batch 120/173] avg loss 1.38743e-05, throughput 5.99607K wps
[Epoch 42 Batch 150/173] avg loss 7.98505e-06, throughput 5.99427K wps
Begin Testing...
[Epoch 42] train avg loss 7.32228e-06, test acc 0.7510, test avg loss 1.5376, throughput 6.01321K wps
[Epoch 43 Batch 30/173] avg loss 3.8463e-06, throughput 6.12887K wps
[Epoch 43 Batch 60/173] avg loss 4.58056e-06, throughput 5.98735K wps
[Epoch 43 Batch 90/173] avg loss 3.86413e-06, throughput 5.99316K wps
[Epoch 43 Batch 120/173] avg loss 1.2582e-05, throughput 5.98501K wps
[Epoch 43 Batch 150/173] avg loss 6.75356e-06, throughput 5.98974K wps
Begin Testing...
[Epoch 43] train avg loss 6.14037e-06, test acc 0.7510, test avg loss 1.55167, throughput 6.01495K wps
[Epoch 44 Batch 30/173] avg loss 9.03806e-06, throughput 6.11303K wps
[Epoch 44 Batch 60/173] avg loss 3.57423e-06, throughput 5.98299K wps
[Epoch 44 Batch 90/173] avg loss 4.18317e-06, throughput 5.98792K wps
[Epoch 44 Batch 120/173] avg loss 3.85694e-06, throughput 5.98521K wps
[Epoch 44 Batch 150/173] avg loss 3.70698e-06, throughput 5.98192K wps
Begin Testing...
[Epoch 44] train avg loss 6.25056e-06, test acc 0.7500, test avg loss 1.57369, throughput 6.00712K wps
[Epoch 45 Batch 30/173] avg loss 3.78903e-06, throughput 6.14429K wps
[Epoch 45 Batch 60/173] avg loss 3.11276e-06, throughput 5.9976K wps
[Epoch 45 Batch 90/173] avg loss 1.24891e-05, throughput 5.99713K wps
[Epoch 45 Batch 120/173] avg loss 4.12245e-06, throughput 5.99875K wps
[Epoch 45 Batch 150/173] avg loss 3.48098e-06, throughput 5.9724K wps
Begin Testing...
[Epoch 45] train avg loss 5.26565e-06, test acc 0.7510, test avg loss 1.59194, throughput 6.01572K wps
[Epoch 46 Batch 30/173] avg loss 4.27068e-06, throughput 6.12365K wps
[Epoch 46 Batch 60/173] avg loss 1.10535e-05, throughput 5.97846K wps
[Epoch 46 Batch 90/173] avg loss 4.43747e-06, throughput 5.99084K wps
[Epoch 46 Batch 120/173] avg loss 4.91916e-06, throughput 5.99643K wps
[Epoch 46 Batch 150/173] avg loss 3.36201e-06, throughput 5.99124K wps
Begin Testing...
[Epoch 46] train avg loss 5.23838e-06, test acc 0.7479, test avg loss 1.60675, throughput 6.01312K wps
[Epoch 47 Batch 30/173] avg loss 3.72901e-06, throughput 6.13612K wps
[Epoch 47 Batch 60/173] avg loss 3.83252e-06, throughput 5.98423K wps
[Epoch 47 Batch 90/173] avg loss 2.54391e-06, throughput 5.97194K wps
[Epoch 47 Batch 120/173] avg loss 1.21668e-05, throughput 5.97762K wps
[Epoch 47 Batch 150/173] avg loss 2.6545e-06, throughput 5.99536K wps
Begin Testing...
[Epoch 47] train avg loss 4.63563e-06, test acc 0.7542, test avg loss 1.63382, throughput 6.01065K wps
[Epoch 48 Batch 30/173] avg loss 2.52082e-06, throughput 6.1283K wps
[Epoch 48 Batch 60/173] avg loss 2.96353e-06, throughput 5.9801K wps
[Epoch 48 Batch 90/173] avg loss 1.02084e-05, throughput 5.9866K wps
[Epoch 48 Batch 120/173] avg loss 1.98711e-06, throughput 5.98304K wps
[Epoch 48 Batch 150/173] avg loss 2.41937e-06, throughput 5.98321K wps
Begin Testing...
[Epoch 48] train avg loss 3.95044e-06, test acc 0.7531, test avg loss 1.65378, throughput 6.00924K wps
[Epoch 49 Batch 30/173] avg loss 1.34427e-05, throughput 6.14555K wps
[Epoch 49 Batch 60/173] avg loss 7.17504e-06, throughput 5.99135K wps
[Epoch 49 Batch 90/173] avg loss 3.16317e-06, throughput 5.99811K wps
[Epoch 49 Batch 120/173] avg loss 1.86635e-06, throughput 5.99126K wps
[Epoch 49 Batch 150/173] avg loss 2.48298e-06, throughput 5.98548K wps
Begin Testing...
[Epoch 49] train avg loss 5.18207e-06, test acc 0.7490, test avg loss 1.67447, throughput 6.01819K wps
[Epoch 50 Batch 30/173] avg loss 2.08221e-06, throughput 6.1523K wps
[Epoch 50 Batch 60/173] avg loss 2.46025e-06, throughput 5.99864K wps
[Epoch 50 Batch 90/173] avg loss 3.1769e-06, throughput 5.97885K wps
[Epoch 50 Batch 120/173] avg loss 2.51134e-06, throughput 6.0031K wps
[Epoch 50 Batch 150/173] avg loss 1.0167e-05, throughput 6.0002K wps
Begin Testing...
[Epoch 50] train avg loss 3.91295e-06, test acc 0.7469, test avg loss 1.70157, throughput 6.02055K wps
[Epoch 51 Batch 30/173] avg loss 2.07e-06, throughput 6.1274K wps
[Epoch 51 Batch 60/173] avg loss 2.00117e-06, throughput 5.98322K wps
[Epoch 51 Batch 90/173] avg loss 1.50947e-06, throughput 5.99042K wps
[Epoch 51 Batch 120/173] avg loss 2.11374e-06, throughput 5.99843K wps
[Epoch 51 Batch 150/173] avg loss 3.0261e-06, throughput 5.99404K wps
Begin Testing...
[Epoch 51] train avg loss 3.20827e-06, test acc 0.7490, test avg loss 1.72139, throughput 6.01609K wps
[Epoch 52 Batch 30/173] avg loss 2.84364e-06, throughput 6.14487K wps
[Epoch 52 Batch 60/173] avg loss 3.35898e-06, throughput 5.99429K wps
[Epoch 52 Batch 90/173] avg loss 9.0831e-06, throughput 5.99299K wps
[Epoch 52 Batch 120/173] avg loss 3.06474e-06, throughput 5.98218K wps
[Epoch 52 Batch 150/173] avg loss 1.60541e-06, throughput 5.99645K wps
Begin Testing...
[Epoch 52] train avg loss 3.75545e-06, test acc 0.7458, test avg loss 1.75231, throughput 6.02122K wps
[Epoch 53 Batch 30/173] avg loss 1.59416e-06, throughput 6.14222K wps
[Epoch 53 Batch 60/173] avg loss 4.24843e-06, throughput 5.98804K wps
[Epoch 53 Batch 90/173] avg loss 8.86941e-06, throughput 5.98817K wps
[Epoch 53 Batch 120/173] avg loss 1.93073e-06, throughput 5.98456K wps
[Epoch 53 Batch 150/173] avg loss 1.99252e-06, throughput 5.99929K wps
Begin Testing...
[Epoch 53] train avg loss 3.55769e-06, test acc 0.7448, test avg loss 1.77275, throughput 6.01689K wps
[Epoch 54 Batch 30/173] avg loss 7.79373e-06, throughput 6.13987K wps
[Epoch 54 Batch 60/173] avg loss 2.00368e-06, throughput 5.9875K wps
[Epoch 54 Batch 90/173] avg loss 1.99469e-06, throughput 5.98574K wps
[Epoch 54 Batch 120/173] avg loss 3.39082e-06, throughput 5.97281K wps
[Epoch 54 Batch 150/173] avg loss 2.17264e-06, throughput 5.98148K wps
Begin Testing...
[Epoch 54] train avg loss 3.30919e-06, test acc 0.7469, test avg loss 1.80093, throughput 6.00976K wps
[Epoch 55 Batch 30/173] avg loss 1.06305e-06, throughput 6.12992K wps
[Epoch 55 Batch 60/173] avg loss 1.89423e-06, throughput 5.98634K wps
[Epoch 55 Batch 90/173] avg loss 1.49203e-06, throughput 5.98988K wps
[Epoch 55 Batch 120/173] avg loss 3.20832e-05, throughput 5.99449K wps
[Epoch 55 Batch 150/173] avg loss 2.12065e-06, throughput 5.98755K wps
Begin Testing...
[Epoch 55] train avg loss 6.86323e-06, test acc 0.7396, test avg loss 1.81045, throughput 6.01594K wps
[Epoch 56 Batch 30/173] avg loss 2.33091e-06, throughput 6.12923K wps
[Epoch 56 Batch 60/173] avg loss 8.37425e-06, throughput 5.97896K wps
[Epoch 56 Batch 90/173] avg loss 1.5758e-06, throughput 5.98994K wps
[Epoch 56 Batch 120/173] avg loss 2.01785e-06, throughput 5.99965K wps
[Epoch 56 Batch 150/173] avg loss 1.43264e-06, throughput 5.98427K wps
Begin Testing...
[Epoch 56] train avg loss 3.08623e-06, test acc 0.7438, test avg loss 1.82723, throughput 6.0127K wps
[Epoch 57 Batch 30/173] avg loss 1.13963e-06, throughput 6.12946K wps
[Epoch 57 Batch 60/173] avg loss 1.39221e-06, throughput 5.98156K wps
[Epoch 57 Batch 90/173] avg loss 1.72963e-06, throughput 5.97604K wps
[Epoch 57 Batch 120/173] avg loss 2.17662e-06, throughput 5.99279K wps
[Epoch 57 Batch 150/173] avg loss 5.71217e-06, throughput 5.99487K wps
Begin Testing...
[Epoch 57] train avg loss 2.55293e-06, test acc 0.7427, test avg loss 1.84034, throughput 6.01205K wps
[Epoch 58 Batch 30/173] avg loss 3.36517e-06, throughput 6.13062K wps
[Epoch 58 Batch 60/173] avg loss 1.93945e-06, throughput 5.99376K wps
[Epoch 58 Batch 90/173] avg loss 1.06309e-06, throughput 6.00242K wps
[Epoch 58 Batch 120/173] avg loss 1.83176e-06, throughput 5.99355K wps
[Epoch 58 Batch 150/173] avg loss 1.68708e-06, throughput 5.97811K wps
Begin Testing...
[Epoch 58] train avg loss 1.85127e-06, test acc 0.7375, test avg loss 1.86792, throughput 6.01653K wps
[Epoch 59 Batch 30/173] avg loss 1.38209e-06, throughput 6.11775K wps
[Epoch 59 Batch 60/173] avg loss 1.18718e-06, throughput 5.96842K wps
[Epoch 59 Batch 90/173] avg loss 1.38938e-06, throughput 6.00818K wps
[Epoch 59 Batch 120/173] avg loss 1.20969e-06, throughput 5.9964K wps
[Epoch 59 Batch 150/173] avg loss 1.61441e-06, throughput 6.00012K wps
Begin Testing...
[Epoch 59] train avg loss 1.35269e-06, test acc 0.7448, test avg loss 1.87602, throughput 6.01373K wps
Test loss 0.505131, test acc 0.7495
Total time cost 357.93s
9596 1066
[Epoch 0 Batch 30/173] avg loss 0.013862, throughput 5.77148K wps
[Epoch 0 Batch 60/173] avg loss 0.0138564, throughput 5.99145K wps
[Epoch 0 Batch 90/173] avg loss 0.0138138, throughput 5.98422K wps
[Epoch 0 Batch 120/173] avg loss 0.0138065, throughput 5.98101K wps
[Epoch 0 Batch 150/173] avg loss 0.013751, throughput 5.97378K wps
Begin Testing...
[Epoch 0] train avg loss 0.0138322, test acc 0.6594, test avg loss 0.686328, throughput 5.94685K wps
Observed Improvement.
Begin Testing...
[Epoch 1 Batch 30/173] avg loss 0.0134929, throughput 6.13107K wps
[Epoch 1 Batch 60/173] avg loss 0.0134857, throughput 6.00509K wps
[Epoch 1 Batch 90/173] avg loss 0.0134661, throughput 5.99823K wps
[Epoch 1 Batch 120/173] avg loss 0.0134049, throughput 6.0074K wps
[Epoch 1 Batch 150/173] avg loss 0.0133539, throughput 5.99605K wps
Begin Testing...
[Epoch 1] train avg loss 0.0134522, test acc 0.6885, test avg loss 0.672324, throughput 6.02192K wps
Observed Improvement.
Begin Testing...
[Epoch 2 Batch 30/173] avg loss 0.0129412, throughput 6.14384K wps
[Epoch 2 Batch 60/173] avg loss 0.0129055, throughput 5.98531K wps
[Epoch 2 Batch 90/173] avg loss 0.012748, throughput 5.99214K wps
[Epoch 2 Batch 120/173] avg loss 0.0125897, throughput 5.99389K wps
[Epoch 2 Batch 150/173] avg loss 0.0125304, throughput 5.98541K wps
Begin Testing...
[Epoch 2] train avg loss 0.0127014, test acc 0.7115, test avg loss 0.635578, throughput 6.01686K wps
Observed Improvement.
Begin Testing...
[Epoch 3 Batch 30/173] avg loss 0.0118013, throughput 6.14594K wps
[Epoch 3 Batch 60/173] avg loss 0.011385, throughput 6.00402K wps
[Epoch 3 Batch 90/173] avg loss 0.0112321, throughput 5.99131K wps
[Epoch 3 Batch 120/173] avg loss 0.0109781, throughput 5.98348K wps
[Epoch 3 Batch 150/173] avg loss 0.0106, throughput 5.98035K wps
Begin Testing...
[Epoch 3] train avg loss 0.0111031, test acc 0.7385, test avg loss 0.565085, throughput 6.01617K wps
Observed Improvement.
Begin Testing...
[Epoch 4 Batch 30/173] avg loss 0.00933889, throughput 6.14061K wps
[Epoch 4 Batch 60/173] avg loss 0.00907198, throughput 6.0015K wps
[Epoch 4 Batch 90/173] avg loss 0.00872673, throughput 5.99021K wps
[Epoch 4 Batch 120/173] avg loss 0.00843562, throughput 5.97853K wps
[Epoch 4 Batch 150/173] avg loss 0.00828991, throughput 5.98423K wps
Begin Testing...
[Epoch 4] train avg loss 0.00863958, test acc 0.7667, test avg loss 0.499303, throughput 6.01461K wps
Observed Improvement.
Begin Testing...
[Epoch 5 Batch 30/173] avg loss 0.00664685, throughput 6.13267K wps
[Epoch 5 Batch 60/173] avg loss 0.0064358, throughput 5.99635K wps
[Epoch 5 Batch 90/173] avg loss 0.00608552, throughput 5.99845K wps
[Epoch 5 Batch 120/173] avg loss 0.00609608, throughput 5.99288K wps
[Epoch 5 Batch 150/173] avg loss 0.00589802, throughput 6.00742K wps
Begin Testing...
[Epoch 5] train avg loss 0.00622486, test acc 0.7677, test avg loss 0.469886, throughput 6.01998K wps
Observed Improvement.
Begin Testing...
[Epoch 6 Batch 30/173] avg loss 0.00466109, throughput 6.13046K wps
[Epoch 6 Batch 60/173] avg loss 0.00436754, throughput 5.98895K wps
[Epoch 6 Batch 90/173] avg loss 0.00440373, throughput 5.98143K wps
[Epoch 6 Batch 120/173] avg loss 0.00423767, throughput 5.99097K wps
[Epoch 6 Batch 150/173] avg loss 0.00447841, throughput 5.98939K wps
Begin Testing...
[Epoch 6] train avg loss 0.00442161, test acc 0.7594, test avg loss 0.474535, throughput 6.01282K wps
[Epoch 7 Batch 30/173] avg loss 0.00318234, throughput 6.13179K wps
[Epoch 7 Batch 60/173] avg loss 0.00306956, throughput 5.98988K wps
[Epoch 7 Batch 90/173] avg loss 0.00322488, throughput 5.9932K wps
[Epoch 7 Batch 120/173] avg loss 0.00312496, throughput 5.98877K wps
[Epoch 7 Batch 150/173] avg loss 0.00303006, throughput 5.98344K wps
Begin Testing...
[Epoch 7] train avg loss 0.0030877, test acc 0.7531, test avg loss 0.494368, throughput 6.0143K wps
[Epoch 8 Batch 30/173] avg loss 0.002093, throughput 6.12758K wps
[Epoch 8 Batch 60/173] avg loss 0.00234556, throughput 5.99529K wps
[Epoch 8 Batch 90/173] avg loss 0.00219899, throughput 5.99329K wps
[Epoch 8 Batch 120/173] avg loss 0.00219358, throughput 5.99771K wps
[Epoch 8 Batch 150/173] avg loss 0.00207823, throughput 5.98856K wps
Begin Testing...
[Epoch 8] train avg loss 0.00217742, test acc 0.7688, test avg loss 0.523789, throughput 6.01772K wps
Observed Improvement.
Begin Testing...
[Epoch 9 Batch 30/173] avg loss 0.00140162, throughput 6.14331K wps
[Epoch 9 Batch 60/173] avg loss 0.00138644, throughput 5.97602K wps
[Epoch 9 Batch 90/173] avg loss 0.00155819, throughput 5.97901K wps
[Epoch 9 Batch 120/173] avg loss 0.00153986, throughput 5.99647K wps
[Epoch 9 Batch 150/173] avg loss 0.00170614, throughput 5.99404K wps
Begin Testing...
[Epoch 9] train avg loss 0.00155715, test acc 0.7625, test avg loss 0.551078, throughput 6.01459K wps
[Epoch 10 Batch 30/173] avg loss 0.00104728, throughput 6.15479K wps
[Epoch 10 Batch 60/173] avg loss 0.00111338, throughput 5.99859K wps
[Epoch 10 Batch 90/173] avg loss 0.00106781, throughput 6.00281K wps
[Epoch 10 Batch 120/173] avg loss 0.000998078, throughput 5.97419K wps
[Epoch 10 Batch 150/173] avg loss 0.00122035, throughput 5.98334K wps
Begin Testing...
[Epoch 10] train avg loss 0.00110879, test acc 0.7594, test avg loss 0.589868, throughput 6.01816K wps
[Epoch 11 Batch 30/173] avg loss 0.000810338, throughput 6.12683K wps
[Epoch 11 Batch 60/173] avg loss 0.000760625, throughput 5.97446K wps
[Epoch 11 Batch 90/173] avg loss 0.000860036, throughput 5.9755K wps
[Epoch 11 Batch 120/173] avg loss 0.000842198, throughput 5.99335K wps
[Epoch 11 Batch 150/173] avg loss 0.000830871, throughput 5.98436K wps
Begin Testing...
[Epoch 11] train avg loss 0.000818057, test acc 0.7521, test avg loss 0.625732, throughput 6.0099K wps
[Epoch 12 Batch 30/173] avg loss 0.000520954, throughput 6.14176K wps
[Epoch 12 Batch 60/173] avg loss 0.000639575, throughput 5.98233K wps
[Epoch 12 Batch 90/173] avg loss 0.000658006, throughput 5.98053K wps
[Epoch 12 Batch 120/173] avg loss 0.000611424, throughput 5.99753K wps
[Epoch 12 Batch 150/173] avg loss 0.00063274, throughput 5.98238K wps
Begin Testing...
[Epoch 12] train avg loss 0.000612042, test acc 0.7417, test avg loss 0.667662, throughput 6.0133K wps
[Epoch 13 Batch 30/173] avg loss 0.000428616, throughput 6.14915K wps
[Epoch 13 Batch 60/173] avg loss 0.000474476, throughput 5.98258K wps
[Epoch 13 Batch 90/173] avg loss 0.000461998, throughput 5.98745K wps
[Epoch 13 Batch 120/173] avg loss 0.000499897, throughput 5.99341K wps
[Epoch 13 Batch 150/173] avg loss 0.000416888, throughput 5.98666K wps
Begin Testing...
[Epoch 13] train avg loss 0.000469085, test acc 0.7438, test avg loss 0.696299, throughput 6.01545K wps
[Epoch 14 Batch 30/173] avg loss 0.000354124, throughput 6.13387K wps
[Epoch 14 Batch 60/173] avg loss 0.000315231, throughput 5.99507K wps
[Epoch 14 Batch 90/173] avg loss 0.000381192, throughput 5.98341K wps
[Epoch 14 Batch 120/173] avg loss 0.000295255, throughput 5.98442K wps
[Epoch 14 Batch 150/173] avg loss 0.000390362, throughput 5.99402K wps
Begin Testing...
[Epoch 14] train avg loss 0.000346827, test acc 0.7427, test avg loss 0.727232, throughput 6.01299K wps
[Epoch 15 Batch 30/173] avg loss 0.000267986, throughput 6.14481K wps
[Epoch 15 Batch 60/173] avg loss 0.000301247, throughput 6.00322K wps
[Epoch 15 Batch 90/173] avg loss 0.000236536, throughput 6.00145K wps
[Epoch 15 Batch 120/173] avg loss 0.000297414, throughput 5.98392K wps
[Epoch 15 Batch 150/173] avg loss 0.000295768, throughput 5.98048K wps
Begin Testing...
[Epoch 15] train avg loss 0.000278573, test acc 0.7427, test avg loss 0.758634, throughput 6.01793K wps
[Epoch 16 Batch 30/173] avg loss 0.000201398, throughput 6.1451K wps
[Epoch 16 Batch 60/173] avg loss 0.00022705, throughput 6.00652K wps
[Epoch 16 Batch 90/173] avg loss 0.000234467, throughput 5.98381K wps
[Epoch 16 Batch 120/173] avg loss 0.000205363, throughput 5.98989K wps
[Epoch 16 Batch 150/173] avg loss 0.000193697, throughput 5.98585K wps
Begin Testing...
[Epoch 16] train avg loss 0.00020914, test acc 0.7406, test avg loss 0.787127, throughput 6.01762K wps
[Epoch 17 Batch 30/173] avg loss 0.000187699, throughput 6.13488K wps
[Epoch 17 Batch 60/173] avg loss 0.000172559, throughput 5.99887K wps
[Epoch 17 Batch 90/173] avg loss 0.000168258, throughput 6.0077K wps
[Epoch 17 Batch 120/173] avg loss 0.000179411, throughput 5.99327K wps
[Epoch 17 Batch 150/173] avg loss 0.00017341, throughput 5.98445K wps
Begin Testing...
[Epoch 17] train avg loss 0.000176434, test acc 0.7365, test avg loss 0.817068, throughput 6.01857K wps
[Epoch 18 Batch 30/173] avg loss 0.00014942, throughput 6.12002K wps
[Epoch 18 Batch 60/173] avg loss 0.000128992, throughput 5.99138K wps
[Epoch 18 Batch 90/173] avg loss 0.000132664, throughput 5.99258K wps
[Epoch 18 Batch 120/173] avg loss 0.000138312, throughput 5.9835K wps
[Epoch 18 Batch 150/173] avg loss 0.000112177, throughput 5.99292K wps
Begin Testing...
[Epoch 18] train avg loss 0.000136974, test acc 0.7385, test avg loss 0.842321, throughput 6.01412K wps
[Epoch 19 Batch 30/173] avg loss 0.000115575, throughput 6.12726K wps
[Epoch 19 Batch 60/173] avg loss 0.00013479, throughput 5.98013K wps
[Epoch 19 Batch 90/173] avg loss 0.000126693, throughput 5.98057K wps
[Epoch 19 Batch 120/173] avg loss 0.00012318, throughput 5.97971K wps
[Epoch 19 Batch 150/173] avg loss 0.000118738, throughput 5.9727K wps
Begin Testing...
[Epoch 19] train avg loss 0.000122357, test acc 0.7427, test avg loss 0.867439, throughput 6.00531K wps
[Epoch 20 Batch 30/173] avg loss 9.94808e-05, throughput 6.14794K wps
[Epoch 20 Batch 60/173] avg loss 8.51967e-05, throughput 6.00668K wps
[Epoch 20 Batch 90/173] avg loss 9.26463e-05, throughput 5.98511K wps
[Epoch 20 Batch 120/173] avg loss 0.000101262, throughput 5.99186K wps
[Epoch 20 Batch 150/173] avg loss 0.000126933, throughput 5.98478K wps
Begin Testing...
[Epoch 20] train avg loss 0.000102021, test acc 0.7396, test avg loss 0.898392, throughput 6.01969K wps
[Epoch 21 Batch 30/173] avg loss 6.93756e-05, throughput 6.11999K wps
[Epoch 21 Batch 60/173] avg loss 8.93126e-05, throughput 5.98249K wps
[Epoch 21 Batch 90/173] avg loss 9.79364e-05, throughput 5.98528K wps
[Epoch 21 Batch 120/173] avg loss 7.28821e-05, throughput 5.9819K wps
[Epoch 21 Batch 150/173] avg loss 8.78502e-05, throughput 5.98948K wps
Begin Testing...
[Epoch 21] train avg loss 8.29257e-05, test acc 0.7406, test avg loss 0.927792, throughput 6.00863K wps
[Epoch 22 Batch 30/173] avg loss 6.68489e-05, throughput 6.13491K wps
[Epoch 22 Batch 60/173] avg loss 5.9716e-05, throughput 5.99059K wps
[Epoch 22 Batch 90/173] avg loss 5.30979e-05, throughput 5.97537K wps
[Epoch 22 Batch 120/173] avg loss 6.93004e-05, throughput 5.99191K wps
[Epoch 22 Batch 150/173] avg loss 7.20643e-05, throughput 5.99022K wps
Begin Testing...
[Epoch 22] train avg loss 6.82441e-05, test acc 0.7427, test avg loss 0.942233, throughput 6.0126K wps
[Epoch 23 Batch 30/173] avg loss 5.39834e-05, throughput 6.13353K wps
[Epoch 23 Batch 60/173] avg loss 4.9577e-05, throughput 5.99776K wps
[Epoch 23 Batch 90/173] avg loss 6.10417e-05, throughput 5.97999K wps
[Epoch 23 Batch 120/173] avg loss 8.3254e-05, throughput 5.98694K wps
[Epoch 23 Batch 150/173] avg loss 5.9833e-05, throughput 5.98379K wps
Begin Testing...
[Epoch 23] train avg loss 6.00062e-05, test acc 0.7375, test avg loss 0.965724, throughput 6.01092K wps
[Epoch 24 Batch 30/173] avg loss 4.35053e-05, throughput 6.12654K wps
[Epoch 24 Batch 60/173] avg loss 6.99521e-05, throughput 5.97786K wps
[Epoch 24 Batch 90/173] avg loss 5.62912e-05, throughput 5.99337K wps
[Epoch 24 Batch 120/173] avg loss 4.27379e-05, throughput 5.98016K wps
[Epoch 24 Batch 150/173] avg loss 7.03655e-05, throughput 5.98339K wps
Begin Testing...
[Epoch 24] train avg loss 5.57325e-05, test acc 0.7344, test avg loss 0.993359, throughput 6.00948K wps
[Epoch 25 Batch 30/173] avg loss 4.82899e-05, throughput 6.13631K wps
[Epoch 25 Batch 60/173] avg loss 5.51936e-05, throughput 5.99605K wps
[Epoch 25 Batch 90/173] avg loss 4.31911e-05, throughput 5.98704K wps
[Epoch 25 Batch 120/173] avg loss 4.27955e-05, throughput 6.00084K wps
[Epoch 25 Batch 150/173] avg loss 4.57533e-05, throughput 5.98888K wps
Begin Testing...
[Epoch 25] train avg loss 4.66949e-05, test acc 0.7365, test avg loss 1.01721, throughput 6.01786K wps
[Epoch 26 Batch 30/173] avg loss 3.35956e-05, throughput 6.13877K wps
[Epoch 26 Batch 60/173] avg loss 3.32587e-05, throughput 5.99722K wps
[Epoch 26 Batch 90/173] avg loss 3.4726e-05, throughput 5.99209K wps
[Epoch 26 Batch 120/173] avg loss 6.0149e-05, throughput 5.98253K wps
[Epoch 26 Batch 150/173] avg loss 3.94698e-05, throughput 5.97951K wps
Begin Testing...
[Epoch 26] train avg loss 3.95876e-05, test acc 0.7365, test avg loss 1.03565, throughput 6.01387K wps
[Epoch 27 Batch 30/173] avg loss 3.26212e-05, throughput 6.14151K wps
[Epoch 27 Batch 60/173] avg loss 3.15216e-05, throughput 5.99495K wps
[Epoch 27 Batch 90/173] avg loss 4.21077e-05, throughput 6.00312K wps
[Epoch 27 Batch 120/173] avg loss 2.68807e-05, throughput 6.00039K wps
[Epoch 27 Batch 150/173] avg loss 4.82363e-05, throughput 5.9839K wps
Begin Testing...
[Epoch 27] train avg loss 3.58218e-05, test acc 0.7375, test avg loss 1.06059, throughput 6.0205K wps
[Epoch 28 Batch 30/173] avg loss 4.10188e-05, throughput 6.13265K wps
[Epoch 28 Batch 60/173] avg loss 2.4083e-05, throughput 5.98136K wps
[Epoch 28 Batch 90/173] avg loss 3.24352e-05, throughput 5.97788K wps
[Epoch 28 Batch 120/173] avg loss 2.95939e-05, throughput 5.99175K wps
[Epoch 28 Batch 150/173] avg loss 2.92666e-05, throughput 5.99514K wps
Begin Testing...
[Epoch 28] train avg loss 3.02169e-05, test acc 0.7323, test avg loss 1.08033, throughput 6.01242K wps
[Epoch 29 Batch 30/173] avg loss 3.32124e-05, throughput 6.10381K wps
[Epoch 29 Batch 60/173] avg loss 2.70822e-05, throughput 5.92608K wps
[Epoch 29 Batch 90/173] avg loss 2.26148e-05, throughput 5.9947K wps
[Epoch 29 Batch 120/173] avg loss 2.33434e-05, throughput 5.98847K wps
[Epoch 29 Batch 150/173] avg loss 3.12318e-05, throughput 5.98666K wps
Begin Testing...
[Epoch 29] train avg loss 2.77524e-05, test acc 0.7323, test avg loss 1.09935, throughput 5.99924K wps
[Epoch 30 Batch 30/173] avg loss 1.75202e-05, throughput 6.14344K wps
[Epoch 30 Batch 60/173] avg loss 3.65847e-05, throughput 5.9927K wps
[Epoch 30 Batch 90/173] avg loss 2.38593e-05, throughput 5.99968K wps
[Epoch 30 Batch 120/173] avg loss 2.34238e-05, throughput 6.00033K wps
[Epoch 30 Batch 150/173] avg loss 2.11172e-05, throughput 6.00286K wps
Begin Testing...
[Epoch 30] train avg loss 2.38375e-05, test acc 0.7333, test avg loss 1.11502, throughput 6.02202K wps
[Epoch 31 Batch 30/173] avg loss 2.04365e-05, throughput 6.14622K wps
[Epoch 31 Batch 60/173] avg loss 2.11759e-05, throughput 6.00351K wps
[Epoch 31 Batch 90/173] avg loss 2.82823e-05, throughput 5.98985K wps
[Epoch 31 Batch 120/173] avg loss 2.20884e-05, throughput 5.98717K wps
[Epoch 31 Batch 150/173] avg loss 1.97054e-05, throughput 5.97913K wps
Begin Testing...
[Epoch 31] train avg loss 2.25362e-05, test acc 0.7333, test avg loss 1.1335, throughput 6.01752K wps
[Epoch 32 Batch 30/173] avg loss 1.6323e-05, throughput 6.1482K wps
[Epoch 32 Batch 60/173] avg loss 2.48024e-05, throughput 5.98861K wps
[Epoch 32 Batch 90/173] avg loss 2.23278e-05, throughput 5.98442K wps
[Epoch 32 Batch 120/173] avg loss 1.57559e-05, throughput 5.98407K wps
[Epoch 32 Batch 150/173] avg loss 1.95686e-05, throughput 5.98998K wps
Begin Testing...
[Epoch 32] train avg loss 1.93634e-05, test acc 0.7333, test avg loss 1.15836, throughput 6.01618K wps
[Epoch 33 Batch 30/173] avg loss 1.71444e-05, throughput 6.13104K wps
[Epoch 33 Batch 60/173] avg loss 1.2793e-05, throughput 5.97936K wps
[Epoch 33 Batch 90/173] avg loss 1.6664e-05, throughput 5.9901K wps
[Epoch 33 Batch 120/173] avg loss 1.33533e-05, throughput 5.978K wps
[Epoch 33 Batch 150/173] avg loss 1.61018e-05, throughput 5.98131K wps
Begin Testing...
[Epoch 33] train avg loss 1.66845e-05, test acc 0.7323, test avg loss 1.17847, throughput 6.0069K wps
[Epoch 34 Batch 30/173] avg loss 1.32109e-05, throughput 6.13102K wps
[Epoch 34 Batch 60/173] avg loss 1.38271e-05, throughput 5.98254K wps
[Epoch 34 Batch 90/173] avg loss 1.64951e-05, throughput 5.98372K wps
[Epoch 34 Batch 120/173] avg loss 1.21953e-05, throughput 5.98266K wps
[Epoch 34 Batch 150/173] avg loss 1.64247e-05, throughput 5.99981K wps
Begin Testing...
[Epoch 34] train avg loss 1.55372e-05, test acc 0.7302, test avg loss 1.20013, throughput 6.01375K wps
[Epoch 35 Batch 30/173] avg loss 2.19256e-05, throughput 6.13085K wps
[Epoch 35 Batch 60/173] avg loss 9.50299e-06, throughput 5.97807K wps
[Epoch 35 Batch 90/173] avg loss 1.31837e-05, throughput 5.99583K wps
[Epoch 35 Batch 120/173] avg loss 1.06983e-05, throughput 5.98125K wps
[Epoch 35 Batch 150/173] avg loss 1.51495e-05, throughput 5.98624K wps
Begin Testing...
[Epoch 35] train avg loss 1.37816e-05, test acc 0.7271, test avg loss 1.22122, throughput 6.01312K wps
[Epoch 36 Batch 30/173] avg loss 2.468e-05, throughput 6.12954K wps
[Epoch 36 Batch 60/173] avg loss 9.80011e-06, throughput 5.97382K wps
[Epoch 36 Batch 90/173] avg loss 1.22629e-05, throughput 5.99369K wps
[Epoch 36 Batch 120/173] avg loss 9.81728e-06, throughput 6.00255K wps
[Epoch 36 Batch 150/173] avg loss 9.44393e-06, throughput 5.97826K wps
Begin Testing...
[Epoch 36] train avg loss 1.30351e-05, test acc 0.7271, test avg loss 1.24075, throughput 6.01112K wps
[Epoch 37 Batch 30/173] avg loss 1.07934e-05, throughput 6.12861K wps
[Epoch 37 Batch 60/173] avg loss 8.47339e-06, throughput 5.99602K wps
[Epoch 37 Batch 90/173] avg loss 1.04175e-05, throughput 5.98135K wps
[Epoch 37 Batch 120/173] avg loss 1.18755e-05, throughput 5.969K wps
[Epoch 37 Batch 150/173] avg loss 8.49064e-06, throughput 5.99364K wps
Begin Testing...
[Epoch 37] train avg loss 1.13215e-05, test acc 0.7260, test avg loss 1.25791, throughput 6.01105K wps
[Epoch 38 Batch 30/173] avg loss 1.5201e-05, throughput 6.14066K wps
[Epoch 38 Batch 60/173] avg loss 6.61787e-06, throughput 5.99289K wps
[Epoch 38 Batch 90/173] avg loss 1.02291e-05, throughput 5.99531K wps
[Epoch 38 Batch 120/173] avg loss 9.49105e-06, throughput 5.98768K wps
[Epoch 38 Batch 150/173] avg loss 8.04129e-06, throughput 5.98554K wps
Begin Testing...
[Epoch 38] train avg loss 9.80055e-06, test acc 0.7323, test avg loss 1.27973, throughput 6.01578K wps
[Epoch 39 Batch 30/173] avg loss 8.38502e-06, throughput 6.1296K wps
[Epoch 39 Batch 60/173] avg loss 7.15385e-06, throughput 5.9755K wps
[Epoch 39 Batch 90/173] avg loss 8.54196e-06, throughput 5.98752K wps
[Epoch 39 Batch 120/173] avg loss 1.43967e-05, throughput 6.00001K wps
[Epoch 39 Batch 150/173] avg loss 7.26434e-06, throughput 5.99658K wps
Begin Testing...
[Epoch 39] train avg loss 8.99725e-06, test acc 0.7333, test avg loss 1.29128, throughput 6.01606K wps
[Epoch 40 Batch 30/173] avg loss 5.401e-06, throughput 6.13475K wps
[Epoch 40 Batch 60/173] avg loss 8.55885e-06, throughput 5.98619K wps
[Epoch 40 Batch 90/173] avg loss 5.80164e-06, throughput 5.98049K wps
[Epoch 40 Batch 120/173] avg loss 6.90117e-06, throughput 5.97923K wps
[Epoch 40 Batch 150/173] avg loss 1.39612e-05, throughput 5.97723K wps
Begin Testing...
[Epoch 40] train avg loss 7.97301e-06, test acc 0.7271, test avg loss 1.31139, throughput 6.008K wps
[Epoch 41 Batch 30/173] avg loss 5.32035e-06, throughput 6.14846K wps
[Epoch 41 Batch 60/173] avg loss 5.84858e-06, throughput 5.99346K wps
[Epoch 41 Batch 90/173] avg loss 5.56208e-06, throughput 5.98242K wps
[Epoch 41 Batch 120/173] avg loss 1.42217e-05, throughput 5.99149K wps
[Epoch 41 Batch 150/173] avg loss 5.58788e-06, throughput 5.98276K wps
Begin Testing...
[Epoch 41] train avg loss 7.06873e-06, test acc 0.7292, test avg loss 1.32647, throughput 6.01475K wps
[Epoch 42 Batch 30/173] avg loss 4.67656e-06, throughput 6.14113K wps
[Epoch 42 Batch 60/173] avg loss 1.25306e-05, throughput 5.99741K wps
[Epoch 42 Batch 90/173] avg loss 5.76816e-06, throughput 5.99256K wps
[Epoch 42 Batch 120/173] avg loss 5.20311e-06, throughput 5.99006K wps
[Epoch 42 Batch 150/173] avg loss 5.00804e-06, throughput 5.99516K wps
Begin Testing...
[Epoch 42] train avg loss 6.49103e-06, test acc 0.7292, test avg loss 1.35019, throughput 6.02087K wps
[Epoch 43 Batch 30/173] avg loss 3.34485e-06, throughput 6.1328K wps
[Epoch 43 Batch 60/173] avg loss 4.50336e-06, throughput 5.99771K wps
[Epoch 43 Batch 90/173] avg loss 4.58723e-06, throughput 5.99191K wps
[Epoch 43 Batch 120/173] avg loss 5.34532e-06, throughput 5.99547K wps
[Epoch 43 Batch 150/173] avg loss 5.11357e-06, throughput 5.99799K wps
Begin Testing...
[Epoch 43] train avg loss 5.77412e-06, test acc 0.7302, test avg loss 1.36787, throughput 6.02083K wps
[Epoch 44 Batch 30/173] avg loss 3.30523e-06, throughput 6.13732K wps
[Epoch 44 Batch 60/173] avg loss 4.13046e-06, throughput 5.98412K wps
[Epoch 44 Batch 90/173] avg loss 1.02736e-05, throughput 5.98218K wps
[Epoch 44 Batch 120/173] avg loss 3.8974e-06, throughput 5.99239K wps
[Epoch 44 Batch 150/173] avg loss 3.89166e-06, throughput 5.98778K wps
Begin Testing...
[Epoch 44] train avg loss 5.0046e-06, test acc 0.7292, test avg loss 1.38145, throughput 6.01289K wps
[Epoch 45 Batch 30/173] avg loss 2.85759e-06, throughput 6.12744K wps
[Epoch 45 Batch 60/173] avg loss 3.88127e-06, throughput 5.9764K wps
[Epoch 45 Batch 90/173] avg loss 1.09538e-05, throughput 5.9759K wps
[Epoch 45 Batch 120/173] avg loss 3.80443e-06, throughput 5.99266K wps
[Epoch 45 Batch 150/173] avg loss 9.11603e-06, throughput 5.99963K wps
Begin Testing...
[Epoch 45] train avg loss 5.91163e-06, test acc 0.7302, test avg loss 1.40098, throughput 6.01232K wps
[Epoch 46 Batch 30/173] avg loss 4.53487e-06, throughput 6.1394K wps
[Epoch 46 Batch 60/173] avg loss 4.62613e-06, throughput 5.98126K wps
[Epoch 46 Batch 90/173] avg loss 1.17984e-05, throughput 5.99094K wps
[Epoch 46 Batch 120/173] avg loss 3.377e-06, throughput 5.98474K wps
[Epoch 46 Batch 150/173] avg loss 2.95759e-06, throughput 5.97599K wps
Begin Testing...
[Epoch 46] train avg loss 5.22423e-06, test acc 0.7312, test avg loss 1.40825, throughput 6.0107K wps
[Epoch 47 Batch 30/173] avg loss 2.26182e-06, throughput 6.13962K wps
[Epoch 47 Batch 60/173] avg loss 2.58619e-06, throughput 5.99351K wps
[Epoch 47 Batch 90/173] avg loss 3.29751e-06, throughput 5.98489K wps
[Epoch 47 Batch 120/173] avg loss 2.88363e-06, throughput 5.9739K wps
[Epoch 47 Batch 150/173] avg loss 2.77089e-06, throughput 5.97878K wps
Begin Testing...
[Epoch 47] train avg loss 3.97346e-06, test acc 0.7281, test avg loss 1.42704, throughput 6.01024K wps
[Epoch 48 Batch 30/173] avg loss 3.97276e-06, throughput 6.13947K wps
[Epoch 48 Batch 60/173] avg loss 1.10316e-05, throughput 6.00045K wps
[Epoch 48 Batch 90/173] avg loss 2.38015e-06, throughput 5.98528K wps
[Epoch 48 Batch 120/173] avg loss 4.05628e-06, throughput 5.98762K wps
[Epoch 48 Batch 150/173] avg loss 3.03998e-06, throughput 5.97927K wps
Begin Testing...
[Epoch 48] train avg loss 4.77326e-06, test acc 0.7271, test avg loss 1.44925, throughput 6.01224K wps
[Epoch 49 Batch 30/173] avg loss 2.5247e-06, throughput 6.14299K wps
[Epoch 49 Batch 60/173] avg loss 1.01787e-05, throughput 5.98964K wps
[Epoch 49 Batch 90/173] avg loss 2.58643e-06, throughput 5.98562K wps
[Epoch 49 Batch 120/173] avg loss 3.75093e-06, throughput 5.97546K wps
[Epoch 49 Batch 150/173] avg loss 2.30742e-06, throughput 5.97519K wps
Begin Testing...
[Epoch 49] train avg loss 4.0839e-06, test acc 0.7281, test avg loss 1.47234, throughput 6.01058K wps
[Epoch 50 Batch 30/173] avg loss 1.71859e-06, throughput 6.14312K wps
[Epoch 50 Batch 60/173] avg loss 2.27125e-06, throughput 5.99989K wps
[Epoch 50 Batch 90/173] avg loss 9.49035e-06, throughput 5.98845K wps
[Epoch 50 Batch 120/173] avg loss 4.17229e-06, throughput 5.98609K wps
[Epoch 50 Batch 150/173] avg loss 3.11062e-06, throughput 5.98966K wps
Begin Testing...
[Epoch 50] train avg loss 3.89734e-06, test acc 0.7312, test avg loss 1.48585, throughput 6.01981K wps
[Epoch 51 Batch 30/173] avg loss 1.97856e-06, throughput 6.12954K wps
[Epoch 51 Batch 60/173] avg loss 2.10147e-06, throughput 5.98432K wps
[Epoch 51 Batch 90/173] avg loss 2.13443e-06, throughput 5.99K wps
[Epoch 51 Batch 120/173] avg loss 7.25808e-06, throughput 5.99855K wps
[Epoch 51 Batch 150/173] avg loss 2.10468e-06, throughput 6.00098K wps
Begin Testing...
[Epoch 51] train avg loss 3.06929e-06, test acc 0.7250, test avg loss 1.50606, throughput 6.01689K wps
[Epoch 52 Batch 30/173] avg loss 6.86778e-06, throughput 6.13021K wps
[Epoch 52 Batch 60/173] avg loss 1.90131e-06, throughput 5.97232K wps
[Epoch 52 Batch 90/173] avg loss 1.49524e-06, throughput 5.97581K wps
[Epoch 52 Batch 120/173] avg loss 1.71438e-06, throughput 6.01476K wps
[Epoch 52 Batch 150/173] avg loss 1.96877e-06, throughput 6.0023K wps
Begin Testing...
[Epoch 52] train avg loss 2.76225e-06, test acc 0.7250, test avg loss 1.51882, throughput 6.0171K wps
[Epoch 53 Batch 30/173] avg loss 1.68768e-06, throughput 6.15358K wps
[Epoch 53 Batch 60/173] avg loss 2.17189e-06, throughput 5.98409K wps
[Epoch 53 Batch 90/173] avg loss 6.20011e-06, throughput 5.993K wps
[Epoch 53 Batch 120/173] avg loss 1.95336e-06, throughput 6.00625K wps
[Epoch 53 Batch 150/173] avg loss 3.14036e-06, throughput 6.00752K wps
Begin Testing...
[Epoch 53] train avg loss 2.90517e-06, test acc 0.7250, test avg loss 1.55595, throughput 6.02493K wps
[Epoch 54 Batch 30/173] avg loss 2.50157e-06, throughput 6.14407K wps
[Epoch 54 Batch 60/173] avg loss 1.8845e-06, throughput 6.007K wps
[Epoch 54 Batch 90/173] avg loss 1.89125e-06, throughput 5.98106K wps
[Epoch 54 Batch 120/173] avg loss 1.56939e-06, throughput 5.9882K wps
[Epoch 54 Batch 150/173] avg loss 1.9177e-06, throughput 5.99952K wps
Begin Testing...
[Epoch 54] train avg loss 2.93125e-06, test acc 0.7229, test avg loss 1.56437, throughput 6.02093K wps
[Epoch 55 Batch 30/173] avg loss 1.64303e-06, throughput 6.1341K wps
[Epoch 55 Batch 60/173] avg loss 1.71323e-06, throughput 5.98294K wps
[Epoch 55 Batch 90/173] avg loss 1.72008e-06, throughput 5.9899K wps
[Epoch 55 Batch 120/173] avg loss 1.36738e-06, throughput 5.9967K wps
[Epoch 55 Batch 150/173] avg loss 1.52043e-06, throughput 6.00643K wps
Begin Testing...
[Epoch 55] train avg loss 1.55008e-06, test acc 0.7292, test avg loss 1.57905, throughput 6.01727K wps
[Epoch 56 Batch 30/173] avg loss 1.12857e-06, throughput 6.14389K wps
[Epoch 56 Batch 60/173] avg loss 1.68796e-06, throughput 5.98951K wps
[Epoch 56 Batch 90/173] avg loss 1.43584e-06, throughput 5.99197K wps
[Epoch 56 Batch 120/173] avg loss 1.19475e-06, throughput 5.99142K wps
[Epoch 56 Batch 150/173] avg loss 2.21861e-06, throughput 5.98719K wps
Begin Testing...
[Epoch 56] train avg loss 1.62172e-06, test acc 0.7260, test avg loss 1.59711, throughput 6.0171K wps
[Epoch 57 Batch 30/173] avg loss 1.34883e-06, throughput 6.12991K wps
[Epoch 57 Batch 60/173] avg loss 1.17447e-06, throughput 5.98375K wps
[Epoch 57 Batch 90/173] avg loss 1.35501e-06, throughput 5.98739K wps
[Epoch 57 Batch 120/173] avg loss 1.38841e-06, throughput 5.98631K wps
[Epoch 57 Batch 150/173] avg loss 1.6213e-06, throughput 5.99012K wps
Begin Testing...
[Epoch 57] train avg loss 1.29408e-06, test acc 0.7333, test avg loss 1.60921, throughput 6.01379K wps
[Epoch 58 Batch 30/173] avg loss 1.35318e-06, throughput 6.14675K wps
[Epoch 58 Batch 60/173] avg loss 1.53029e-06, throughput 5.97928K wps
[Epoch 58 Batch 90/173] avg loss 9.58345e-07, throughput 5.97454K wps
[Epoch 58 Batch 120/173] avg loss 8.01843e-07, throughput 5.97607K wps
[Epoch 58 Batch 150/173] avg loss 1.53293e-06, throughput 5.98948K wps
Begin Testing...
[Epoch 58] train avg loss 1.17013e-06, test acc 0.7302, test avg loss 1.63369, throughput 6.00942K wps
[Epoch 59 Batch 30/173] avg loss 1.4173e-06, throughput 6.14155K wps
[Epoch 59 Batch 60/173] avg loss 1.71688e-06, throughput 5.98414K wps
[Epoch 59 Batch 90/173] avg loss 1.34202e-06, throughput 5.96882K wps
[Epoch 59 Batch 120/173] avg loss 1.27358e-06, throughput 5.98522K wps
[Epoch 59 Batch 150/173] avg loss 1.11399e-06, throughput 5.98594K wps
Begin Testing...
[Epoch 59] train avg loss 1.33887e-06, test acc 0.7271, test avg loss 1.65644, throughput 6.01005K wps
Test loss 0.529175, test acc 0.7552
Total time cost 358.36s
9596 1066
[Epoch 0 Batch 30/173] avg loss 0.0138654, throughput 5.79319K wps
[Epoch 0 Batch 60/173] avg loss 0.0138244, throughput 5.97902K wps
[Epoch 0 Batch 90/173] avg loss 0.0138155, throughput 5.99274K wps
[Epoch 0 Batch 120/173] avg loss 0.0137969, throughput 5.98651K wps
[Epoch 0 Batch 150/173] avg loss 0.0137696, throughput 5.99203K wps
Begin Testing...
[Epoch 0] train avg loss 0.0138309, test acc 0.6469, test avg loss 0.685975, throughput 5.95374K wps
Observed Improvement.
Begin Testing...
[Epoch 1 Batch 30/173] avg loss 0.0135153, throughput 6.1406K wps
[Epoch 1 Batch 60/173] avg loss 0.0134702, throughput 5.99362K wps
[Epoch 1 Batch 90/173] avg loss 0.0133891, throughput 5.99824K wps
[Epoch 1 Batch 120/173] avg loss 0.0133763, throughput 5.98188K wps
[Epoch 1 Batch 150/173] avg loss 0.0133189, throughput 5.99042K wps
Begin Testing...
[Epoch 1] train avg loss 0.0134085, test acc 0.6729, test avg loss 0.670058, throughput 6.0172K wps
Observed Improvement.
Begin Testing...
[Epoch 2 Batch 30/173] avg loss 0.0128856, throughput 6.14254K wps
[Epoch 2 Batch 60/173] avg loss 0.0128119, throughput 5.98241K wps
[Epoch 2 Batch 90/173] avg loss 0.0126502, throughput 5.99824K wps
[Epoch 2 Batch 120/173] avg loss 0.0123989, throughput 5.98459K wps
[Epoch 2 Batch 150/173] avg loss 0.0123961, throughput 5.99497K wps
Begin Testing...
[Epoch 2] train avg loss 0.012574, test acc 0.6969, test avg loss 0.625448, throughput 6.01651K wps
Observed Improvement.
Begin Testing...
[Epoch 3 Batch 30/173] avg loss 0.011594, throughput 6.12184K wps
[Epoch 3 Batch 60/173] avg loss 0.0111959, throughput 5.98195K wps
[Epoch 3 Batch 90/173] avg loss 0.0110117, throughput 5.98986K wps
[Epoch 3 Batch 120/173] avg loss 0.0107173, throughput 5.99528K wps
[Epoch 3 Batch 150/173] avg loss 0.0104139, throughput 5.98662K wps
Begin Testing...
[Epoch 3] train avg loss 0.0109169, test acc 0.7583, test avg loss 0.548371, throughput 6.01228K wps
Observed Improvement.
Begin Testing...
[Epoch 4 Batch 30/173] avg loss 0.00905487, throughput 6.14177K wps
[Epoch 4 Batch 60/173] avg loss 0.0086535, throughput 5.98801K wps
[Epoch 4 Batch 90/173] avg loss 0.00821954, throughput 5.98648K wps
[Epoch 4 Batch 120/173] avg loss 0.00839596, throughput 5.98307K wps
[Epoch 4 Batch 150/173] avg loss 0.00811223, throughput 5.99966K wps
Begin Testing...
[Epoch 4] train avg loss 0.00842054, test acc 0.7802, test avg loss 0.483506, throughput 6.0168K wps
Observed Improvement.
Begin Testing...
[Epoch 5 Batch 30/173] avg loss 0.00620303, throughput 6.14609K wps
[Epoch 5 Batch 60/173] avg loss 0.00637442, throughput 5.98034K wps
[Epoch 5 Batch 90/173] avg loss 0.00629995, throughput 5.97934K wps
[Epoch 5 Batch 120/173] avg loss 0.00583493, throughput 5.97969K wps
[Epoch 5 Batch 150/173] avg loss 0.00573021, throughput 5.98615K wps
Begin Testing...
[Epoch 5] train avg loss 0.00602764, test acc 0.7854, test avg loss 0.46186, throughput 6.01174K wps
Observed Improvement.
Begin Testing...
[Epoch 6 Batch 30/173] avg loss 0.00431438, throughput 6.12583K wps
[Epoch 6 Batch 60/173] avg loss 0.00424806, throughput 5.97478K wps
[Epoch 6 Batch 90/173] avg loss 0.00426467, throughput 5.9819K wps
[Epoch 6 Batch 120/173] avg loss 0.00422994, throughput 5.99759K wps
[Epoch 6 Batch 150/173] avg loss 0.00405429, throughput 5.99499K wps
Begin Testing...
[Epoch 6] train avg loss 0.00424713, test acc 0.7729, test avg loss 0.476367, throughput 6.00919K wps
[Epoch 7 Batch 30/173] avg loss 0.00308371, throughput 6.14679K wps
[Epoch 7 Batch 60/173] avg loss 0.00289438, throughput 5.97818K wps
[Epoch 7 Batch 90/173] avg loss 0.00287049, throughput 5.97498K wps
[Epoch 7 Batch 120/173] avg loss 0.00292354, throughput 6.00004K wps
[Epoch 7 Batch 150/173] avg loss 0.0028901, throughput 5.97574K wps
Begin Testing...
[Epoch 7] train avg loss 0.00295622, test acc 0.7750, test avg loss 0.502971, throughput 6.01308K wps
[Epoch 8 Batch 30/173] avg loss 0.00204111, throughput 6.12866K wps
[Epoch 8 Batch 60/173] avg loss 0.00211302, throughput 5.97414K wps
[Epoch 8 Batch 90/173] avg loss 0.00196412, throughput 5.99964K wps
[Epoch 8 Batch 120/173] avg loss 0.00215626, throughput 5.98324K wps
[Epoch 8 Batch 150/173] avg loss 0.0021261, throughput 6.00052K wps
Begin Testing...
[Epoch 8] train avg loss 0.00207234, test acc 0.7708, test avg loss 0.543587, throughput 6.01566K wps
[Epoch 9 Batch 30/173] avg loss 0.00144243, throughput 6.1278K wps
[Epoch 9 Batch 60/173] avg loss 0.00140404, throughput 5.97803K wps
[Epoch 9 Batch 90/173] avg loss 0.00142612, throughput 5.99394K wps
[Epoch 9 Batch 120/173] avg loss 0.00135572, throughput 5.9885K wps
[Epoch 9 Batch 150/173] avg loss 0.00152127, throughput 5.99591K wps
Begin Testing...
[Epoch 9] train avg loss 0.00143971, test acc 0.7656, test avg loss 0.587714, throughput 6.01298K wps
[Epoch 10 Batch 30/173] avg loss 0.00098477, throughput 6.12823K wps
[Epoch 10 Batch 60/173] avg loss 0.00106378, throughput 5.99322K wps
[Epoch 10 Batch 90/173] avg loss 0.00103737, throughput 5.98817K wps
[Epoch 10 Batch 120/173] avg loss 0.00109145, throughput 5.97401K wps
[Epoch 10 Batch 150/173] avg loss 0.000985303, throughput 5.98413K wps
Begin Testing...
[Epoch 10] train avg loss 0.00103224, test acc 0.7677, test avg loss 0.631876, throughput 6.00894K wps
[Epoch 11 Batch 30/173] avg loss 0.000727428, throughput 6.13055K wps
[Epoch 11 Batch 60/173] avg loss 0.00072372, throughput 5.98052K wps
[Epoch 11 Batch 90/173] avg loss 0.000823448, throughput 5.97664K wps
[Epoch 11 Batch 120/173] avg loss 0.000804825, throughput 5.97865K wps
[Epoch 11 Batch 150/173] avg loss 0.000794628, throughput 5.97733K wps
Begin Testing...
[Epoch 11] train avg loss 0.000760873, test acc 0.7667, test avg loss 0.67502, throughput 6.00523K wps
[Epoch 12 Batch 30/173] avg loss 0.000524852, throughput 6.13863K wps
[Epoch 12 Batch 60/173] avg loss 0.00053846, throughput 5.98215K wps
[Epoch 12 Batch 90/173] avg loss 0.000527538, throughput 5.98653K wps
[Epoch 12 Batch 120/173] avg loss 0.00057286, throughput 5.97315K wps
[Epoch 12 Batch 150/173] avg loss 0.000573407, throughput 5.97735K wps
Begin Testing...
[Epoch 12] train avg loss 0.000545407, test acc 0.7604, test avg loss 0.715989, throughput 6.00873K wps
[Epoch 13 Batch 30/173] avg loss 0.0004251, throughput 6.13366K wps
[Epoch 13 Batch 60/173] avg loss 0.00040567, throughput