Permalink
Find file Copy path
Fetching contributors…
Cannot retrieve contributors at this time
4478 lines (4477 sloc) 280 KB
Namespace(batch_size=50, data_name='MR', dropout=0.5, epochs=60, gpu=0, log_interval=30, lr=0.0001, model_mode='multichannel', save_prefix='sa-model')
Use gpu0
2320
56
Done! Tokenizing Time=1.03s, #Sentences=10662
SentimentNet(
(embedding): Embedding(18768 -> 300, float32)
(embedding_extend): Embedding(18768 -> 300, float32)
(encoder): ConvolutionalEncoder(
(_convs): HybridConcurrent(
(0): HybridSequential(
(0): Conv1D(600 -> 100, kernel_size=(3,), stride=(1,))
(1): Activation(relu)
(2): HybridLambda(<lambda>)
)
(1): HybridSequential(
(0): Conv1D(600 -> 100, kernel_size=(4,), stride=(1,))
(1): Activation(relu)
(2): HybridLambda(<lambda>)
)
(2): HybridSequential(
(0): Conv1D(600 -> 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 0.0157471, throughput 2.60015K wps
[Epoch 0 Batch 60/173] avg loss 0.0147572, throughput 4.00481K wps
[Epoch 0 Batch 90/173] avg loss 0.0145813, throughput 3.99603K wps
[Epoch 0 Batch 120/173] avg loss 0.0138326, throughput 4.00244K wps
[Epoch 0 Batch 150/173] avg loss 0.014164, throughput 4.00125K wps
Begin Testing...
[Epoch 0] train avg loss 0.0144574, test acc 0.6031, test avg loss 0.660847, throughput 3.47795K wps
Observed Improvement.
Begin Testing...
[Epoch 1 Batch 30/173] avg loss 0.0132752, throughput 4.09976K wps
[Epoch 1 Batch 60/173] avg loss 0.0134795, throughput 3.98857K wps
[Epoch 1 Batch 90/173] avg loss 0.0130566, throughput 3.97307K wps
[Epoch 1 Batch 120/173] avg loss 0.0124578, throughput 3.99991K wps
[Epoch 1 Batch 150/173] avg loss 0.0129509, throughput 4.00383K wps
Begin Testing...
[Epoch 1] train avg loss 0.0130193, test acc 0.6510, test avg loss 0.636842, throughput 4.01131K wps
Observed Improvement.
Begin Testing...
[Epoch 2 Batch 30/173] avg loss 0.0126215, throughput 4.04459K wps
[Epoch 2 Batch 60/173] avg loss 0.0122645, throughput 4.00278K wps
[Epoch 2 Batch 90/173] avg loss 0.0120211, throughput 3.97523K wps
[Epoch 2 Batch 120/173] avg loss 0.0118471, throughput 3.98782K wps
[Epoch 2 Batch 150/173] avg loss 0.0121468, throughput 3.97706K wps
Begin Testing...
[Epoch 2] train avg loss 0.0121428, test acc 0.6229, test avg loss 0.624933, throughput 3.9963K wps
[Epoch 3 Batch 30/173] avg loss 0.0114316, throughput 4.08667K wps
[Epoch 3 Batch 60/173] avg loss 0.0113349, throughput 3.98737K wps
[Epoch 3 Batch 90/173] avg loss 0.011123, throughput 3.99952K wps
[Epoch 3 Batch 120/173] avg loss 0.0113423, throughput 3.99375K wps
[Epoch 3 Batch 150/173] avg loss 0.0110177, throughput 3.99135K wps
Begin Testing...
[Epoch 3] train avg loss 0.0112388, test acc 0.7177, test avg loss 0.572575, throughput 4.0097K wps
Observed Improvement.
Begin Testing...
[Epoch 4 Batch 30/173] avg loss 0.0104647, throughput 4.04248K wps
[Epoch 4 Batch 60/173] avg loss 0.0103388, throughput 3.97828K wps
[Epoch 4 Batch 90/173] avg loss 0.0101457, throughput 3.9809K wps
[Epoch 4 Batch 120/173] avg loss 0.0100336, throughput 3.98302K wps
[Epoch 4 Batch 150/173] avg loss 0.0102693, throughput 3.98251K wps
Begin Testing...
[Epoch 4] train avg loss 0.0102156, test acc 0.7490, test avg loss 0.537008, throughput 3.99267K wps
Observed Improvement.
Begin Testing...
[Epoch 5 Batch 30/173] avg loss 0.00949295, throughput 4.03773K wps
[Epoch 5 Batch 60/173] avg loss 0.00901308, throughput 3.96987K wps
[Epoch 5 Batch 90/173] avg loss 0.00907549, throughput 3.98388K wps
[Epoch 5 Batch 120/173] avg loss 0.00879587, throughput 3.97716K wps
[Epoch 5 Batch 150/173] avg loss 0.00923361, throughput 3.99149K wps
Begin Testing...
[Epoch 5] train avg loss 0.00907776, test acc 0.7688, test avg loss 0.499666, throughput 3.99132K wps
Observed Improvement.
Begin Testing...
[Epoch 6 Batch 30/173] avg loss 0.00816854, throughput 4.03226K wps
[Epoch 6 Batch 60/173] avg loss 0.00819699, throughput 3.96805K wps
[Epoch 6 Batch 90/173] avg loss 0.00808233, throughput 3.98161K wps
[Epoch 6 Batch 120/173] avg loss 0.00808742, throughput 3.98213K wps
[Epoch 6 Batch 150/173] avg loss 0.00794661, throughput 4.00132K wps
Begin Testing...
[Epoch 6] train avg loss 0.00810661, test acc 0.7792, test avg loss 0.468191, throughput 3.9937K wps
Observed Improvement.
Begin Testing...
[Epoch 7 Batch 30/173] avg loss 0.00704023, throughput 4.0353K wps
[Epoch 7 Batch 60/173] avg loss 0.00726033, throughput 3.96911K wps
[Epoch 7 Batch 90/173] avg loss 0.00700074, throughput 3.98207K wps
[Epoch 7 Batch 120/173] avg loss 0.00711214, throughput 3.9666K wps
[Epoch 7 Batch 150/173] avg loss 0.00682218, throughput 3.98488K wps
Begin Testing...
[Epoch 7] train avg loss 0.00705041, test acc 0.7792, test avg loss 0.452758, throughput 3.98766K wps
Observed Improvement.
Begin Testing...
[Epoch 8 Batch 30/173] avg loss 0.00679779, throughput 4.04953K wps
[Epoch 8 Batch 60/173] avg loss 0.00592811, throughput 3.96338K wps
[Epoch 8 Batch 90/173] avg loss 0.00604165, throughput 3.97478K wps
[Epoch 8 Batch 120/173] avg loss 0.00618495, throughput 3.98647K wps
[Epoch 8 Batch 150/173] avg loss 0.00598045, throughput 3.97903K wps
Begin Testing...
[Epoch 8] train avg loss 0.00617674, test acc 0.7885, test avg loss 0.446116, throughput 3.99008K wps
Observed Improvement.
Begin Testing...
[Epoch 9 Batch 30/173] avg loss 0.00560612, throughput 4.04461K wps
[Epoch 9 Batch 60/173] avg loss 0.00520654, throughput 3.984K wps
[Epoch 9 Batch 90/173] avg loss 0.00530604, throughput 3.97318K wps
[Epoch 9 Batch 120/173] avg loss 0.00551145, throughput 3.98192K wps
[Epoch 9 Batch 150/173] avg loss 0.00517933, throughput 3.94171K wps
Begin Testing...
[Epoch 9] train avg loss 0.00538616, test acc 0.7969, test avg loss 0.429637, throughput 3.97891K wps
Observed Improvement.
Begin Testing...
[Epoch 10 Batch 30/173] avg loss 0.00468569, throughput 4.01623K wps
[Epoch 10 Batch 60/173] avg loss 0.00461581, throughput 3.9516K wps
[Epoch 10 Batch 90/173] avg loss 0.00475527, throughput 3.92419K wps
[Epoch 10 Batch 120/173] avg loss 0.00461951, throughput 3.93949K wps
[Epoch 10 Batch 150/173] avg loss 0.00467305, throughput 3.94695K wps
Begin Testing...
[Epoch 10] train avg loss 0.00469553, test acc 0.7906, test avg loss 0.42918, throughput 3.95156K wps
[Epoch 11 Batch 30/173] avg loss 0.00412019, throughput 4.03931K wps
[Epoch 11 Batch 60/173] avg loss 0.00404152, throughput 3.93644K wps
[Epoch 11 Batch 90/173] avg loss 0.00409582, throughput 3.94977K wps
[Epoch 11 Batch 120/173] avg loss 0.00365978, throughput 3.9475K wps
[Epoch 11 Batch 150/173] avg loss 0.00414771, throughput 3.92706K wps
Begin Testing...
[Epoch 11] train avg loss 0.00403528, test acc 0.7833, test avg loss 0.441626, throughput 3.95814K wps
[Epoch 12 Batch 30/173] avg loss 0.00354411, throughput 4.02232K wps
[Epoch 12 Batch 60/173] avg loss 0.00335672, throughput 3.93958K wps
[Epoch 12 Batch 90/173] avg loss 0.00345916, throughput 3.92081K wps
[Epoch 12 Batch 120/173] avg loss 0.00335828, throughput 3.95005K wps
[Epoch 12 Batch 150/173] avg loss 0.00350461, throughput 3.9172K wps
Begin Testing...
[Epoch 12] train avg loss 0.00348563, test acc 0.8021, test avg loss 0.431681, throughput 3.94697K wps
Observed Improvement.
Begin Testing...
[Epoch 13 Batch 30/173] avg loss 0.00298986, throughput 4.0281K wps
[Epoch 13 Batch 60/173] avg loss 0.00288677, throughput 3.93187K wps
[Epoch 13 Batch 90/173] avg loss 0.00285996, throughput 3.92442K wps
[Epoch 13 Batch 120/173] avg loss 0.00295405, throughput 3.94967K wps
[Epoch 13 Batch 150/173] avg loss 0.00285471, throughput 3.91138K wps
Begin Testing...
[Epoch 13] train avg loss 0.00291457, test acc 0.7875, test avg loss 0.442311, throughput 3.94787K wps
[Epoch 14 Batch 30/173] avg loss 0.00226747, throughput 4.02601K wps
[Epoch 14 Batch 60/173] avg loss 0.00259424, throughput 3.93864K wps
[Epoch 14 Batch 90/173] avg loss 0.00245398, throughput 3.91341K wps
[Epoch 14 Batch 120/173] avg loss 0.00238256, throughput 3.94381K wps
[Epoch 14 Batch 150/173] avg loss 0.00269251, throughput 3.90625K wps
Begin Testing...
[Epoch 14] train avg loss 0.00246088, test acc 0.7990, test avg loss 0.448475, throughput 3.93242K wps
[Epoch 15 Batch 30/173] avg loss 0.0020959, throughput 4.00575K wps
[Epoch 15 Batch 60/173] avg loss 0.00203262, throughput 3.93347K wps
[Epoch 15 Batch 90/173] avg loss 0.00214802, throughput 3.9076K wps
[Epoch 15 Batch 120/173] avg loss 0.00201078, throughput 3.9289K wps
[Epoch 15 Batch 150/173] avg loss 0.00210687, throughput 3.90076K wps
Begin Testing...
[Epoch 15] train avg loss 0.00208825, test acc 0.7885, test avg loss 0.462507, throughput 3.93472K wps
[Epoch 16 Batch 30/173] avg loss 0.00173112, throughput 4.02604K wps
[Epoch 16 Batch 60/173] avg loss 0.00179449, throughput 3.93416K wps
[Epoch 16 Batch 90/173] avg loss 0.00177504, throughput 3.91258K wps
[Epoch 16 Batch 120/173] avg loss 0.00168162, throughput 3.9337K wps
[Epoch 16 Batch 150/173] avg loss 0.00179378, throughput 3.90775K wps
Begin Testing...
[Epoch 16] train avg loss 0.00174725, test acc 0.7854, test avg loss 0.478523, throughput 3.93866K wps
[Epoch 17 Batch 30/173] avg loss 0.00141243, throughput 4.00965K wps
[Epoch 17 Batch 60/173] avg loss 0.00144864, throughput 3.9025K wps
[Epoch 17 Batch 90/173] avg loss 0.00146766, throughput 3.89724K wps
[Epoch 17 Batch 120/173] avg loss 0.0014642, throughput 3.90065K wps
[Epoch 17 Batch 150/173] avg loss 0.00143211, throughput 3.89861K wps
Begin Testing...
[Epoch 17] train avg loss 0.00145784, test acc 0.7812, test avg loss 0.492459, throughput 3.91653K wps
[Epoch 18 Batch 30/173] avg loss 0.00120367, throughput 3.98547K wps
[Epoch 18 Batch 60/173] avg loss 0.00118504, throughput 3.8998K wps
[Epoch 18 Batch 90/173] avg loss 0.00126789, throughput 3.91871K wps
[Epoch 18 Batch 120/173] avg loss 0.00128304, throughput 3.8852K wps
[Epoch 18 Batch 150/173] avg loss 0.00111382, throughput 3.89992K wps
Begin Testing...
[Epoch 18] train avg loss 0.00120912, test acc 0.7927, test avg loss 0.507358, throughput 3.91801K wps
[Epoch 19 Batch 30/173] avg loss 0.0010373, throughput 4.02357K wps
[Epoch 19 Batch 60/173] avg loss 0.000912748, throughput 3.93067K wps
[Epoch 19 Batch 90/173] avg loss 0.00102424, throughput 3.905K wps
[Epoch 19 Batch 120/173] avg loss 0.00103714, throughput 3.9166K wps
[Epoch 19 Batch 150/173] avg loss 0.00106993, throughput 3.91699K wps
Begin Testing...
[Epoch 19] train avg loss 0.00102971, test acc 0.7917, test avg loss 0.522362, throughput 3.93402K wps
[Epoch 20 Batch 30/173] avg loss 0.000874012, throughput 3.98711K wps
[Epoch 20 Batch 60/173] avg loss 0.000925783, throughput 3.91171K wps
[Epoch 20 Batch 90/173] avg loss 0.000868624, throughput 3.90047K wps
[Epoch 20 Batch 120/173] avg loss 0.000793412, throughput 3.88794K wps
[Epoch 20 Batch 150/173] avg loss 0.000896021, throughput 3.92325K wps
Begin Testing...
[Epoch 20] train avg loss 0.000875003, test acc 0.7906, test avg loss 0.540063, throughput 3.91466K wps
[Epoch 21 Batch 30/173] avg loss 0.000752554, throughput 4.00522K wps
[Epoch 21 Batch 60/173] avg loss 0.000675514, throughput 3.88533K wps
[Epoch 21 Batch 90/173] avg loss 0.000855909, throughput 3.88018K wps
[Epoch 21 Batch 120/173] avg loss 0.000719481, throughput 3.8974K wps
[Epoch 21 Batch 150/173] avg loss 0.000742425, throughput 3.90889K wps
Begin Testing...
[Epoch 21] train avg loss 0.000744425, test acc 0.7917, test avg loss 0.555513, throughput 3.91643K wps
[Epoch 22 Batch 30/173] avg loss 0.000632064, throughput 4.00557K wps
[Epoch 22 Batch 60/173] avg loss 0.0005907, throughput 3.90999K wps
[Epoch 22 Batch 90/173] avg loss 0.000601495, throughput 3.89867K wps
[Epoch 22 Batch 120/173] avg loss 0.000693787, throughput 3.90872K wps
[Epoch 22 Batch 150/173] avg loss 0.000711021, throughput 3.88414K wps
Begin Testing...
[Epoch 22] train avg loss 0.000644759, test acc 0.7865, test avg loss 0.574636, throughput 3.91591K wps
[Epoch 23 Batch 30/173] avg loss 0.000513464, throughput 3.97134K wps
[Epoch 23 Batch 60/173] avg loss 0.000568148, throughput 3.88135K wps
[Epoch 23 Batch 90/173] avg loss 0.000558574, throughput 3.88947K wps
[Epoch 23 Batch 120/173] avg loss 0.000444473, throughput 3.90217K wps
[Epoch 23 Batch 150/173] avg loss 0.000588, throughput 3.89057K wps
Begin Testing...
[Epoch 23] train avg loss 0.000531237, test acc 0.7969, test avg loss 0.589529, throughput 3.90277K wps
[Epoch 24 Batch 30/173] avg loss 0.000449188, throughput 3.97491K wps
[Epoch 24 Batch 60/173] avg loss 0.00044116, throughput 3.87551K wps
[Epoch 24 Batch 90/173] avg loss 0.00045155, throughput 3.90446K wps
[Epoch 24 Batch 120/173] avg loss 0.000472358, throughput 3.92064K wps
[Epoch 24 Batch 150/173] avg loss 0.000425667, throughput 3.88906K wps
Begin Testing...
[Epoch 24] train avg loss 0.000459015, test acc 0.7906, test avg loss 0.611665, throughput 3.91488K wps
[Epoch 25 Batch 30/173] avg loss 0.00038595, throughput 4.02366K wps
[Epoch 25 Batch 60/173] avg loss 0.000343941, throughput 3.89809K wps
[Epoch 25 Batch 90/173] avg loss 0.000415869, throughput 3.88874K wps
[Epoch 25 Batch 120/173] avg loss 0.000436876, throughput 3.91235K wps
[Epoch 25 Batch 150/173] avg loss 0.000425049, throughput 3.88942K wps
Begin Testing...
[Epoch 25] train avg loss 0.000411321, test acc 0.7854, test avg loss 0.632548, throughput 3.91518K wps
[Epoch 26 Batch 30/173] avg loss 0.000338878, throughput 3.98773K wps
[Epoch 26 Batch 60/173] avg loss 0.000340269, throughput 3.88979K wps
[Epoch 26 Batch 90/173] avg loss 0.000312563, throughput 3.91532K wps
[Epoch 26 Batch 120/173] avg loss 0.00034106, throughput 3.88494K wps
[Epoch 26 Batch 150/173] avg loss 0.00034927, throughput 3.87652K wps
Begin Testing...
[Epoch 26] train avg loss 0.000341014, test acc 0.7948, test avg loss 0.648108, throughput 3.90711K wps
[Epoch 27 Batch 30/173] avg loss 0.000272546, throughput 3.98137K wps
[Epoch 27 Batch 60/173] avg loss 0.000259413, throughput 3.91152K wps
[Epoch 27 Batch 90/173] avg loss 0.000302696, throughput 3.91194K wps
[Epoch 27 Batch 120/173] avg loss 0.000318228, throughput 3.90044K wps
[Epoch 27 Batch 150/173] avg loss 0.00028782, throughput 3.92853K wps
Begin Testing...
[Epoch 27] train avg loss 0.000291205, test acc 0.7865, test avg loss 0.670919, throughput 3.92237K wps
[Epoch 28 Batch 30/173] avg loss 0.000257371, throughput 3.98208K wps
[Epoch 28 Batch 60/173] avg loss 0.000251174, throughput 3.89086K wps
[Epoch 28 Batch 90/173] avg loss 0.000243167, throughput 3.91018K wps
[Epoch 28 Batch 120/173] avg loss 0.000279066, throughput 3.87387K wps
[Epoch 28 Batch 150/173] avg loss 0.000274613, throughput 3.88775K wps
Begin Testing...
[Epoch 28] train avg loss 0.000258582, test acc 0.7917, test avg loss 0.683674, throughput 3.91135K wps
[Epoch 29 Batch 30/173] avg loss 0.000206925, throughput 3.9922K wps
[Epoch 29 Batch 60/173] avg loss 0.000183059, throughput 3.8934K wps
[Epoch 29 Batch 90/173] avg loss 0.000216462, throughput 3.88229K wps
[Epoch 29 Batch 120/173] avg loss 0.000207379, throughput 3.87738K wps
[Epoch 29 Batch 150/173] avg loss 0.000218366, throughput 3.88383K wps
Begin Testing...
[Epoch 29] train avg loss 0.00021386, test acc 0.7917, test avg loss 0.705349, throughput 3.90545K wps
[Epoch 30 Batch 30/173] avg loss 0.000189186, throughput 4.00179K wps
[Epoch 30 Batch 60/173] avg loss 0.000193284, throughput 3.90002K wps
[Epoch 30 Batch 90/173] avg loss 0.00019357, throughput 3.91913K wps
[Epoch 30 Batch 120/173] avg loss 0.000194609, throughput 3.89747K wps
[Epoch 30 Batch 150/173] avg loss 0.000172244, throughput 3.87833K wps
Begin Testing...
[Epoch 30] train avg loss 0.000189623, test acc 0.7917, test avg loss 0.72284, throughput 3.91803K wps
[Epoch 31 Batch 30/173] avg loss 0.000202806, throughput 3.99425K wps
[Epoch 31 Batch 60/173] avg loss 0.000165855, throughput 3.87887K wps
[Epoch 31 Batch 90/173] avg loss 0.000149838, throughput 3.88283K wps
[Epoch 31 Batch 120/173] avg loss 0.000133034, throughput 3.88339K wps
[Epoch 31 Batch 150/173] avg loss 0.000176121, throughput 3.88624K wps
Begin Testing...
[Epoch 31] train avg loss 0.000163476, test acc 0.7875, test avg loss 0.745572, throughput 3.90547K wps
[Epoch 32 Batch 30/173] avg loss 0.000138492, throughput 3.98548K wps
[Epoch 32 Batch 60/173] avg loss 0.000168043, throughput 3.87749K wps
[Epoch 32 Batch 90/173] avg loss 0.000124916, throughput 3.88147K wps
[Epoch 32 Batch 120/173] avg loss 0.000126785, throughput 3.88556K wps
[Epoch 32 Batch 150/173] avg loss 0.000142744, throughput 3.92141K wps
Begin Testing...
[Epoch 32] train avg loss 0.000141786, test acc 0.7927, test avg loss 0.761188, throughput 3.90602K wps
[Epoch 33 Batch 30/173] avg loss 0.000119109, throughput 3.98559K wps
[Epoch 33 Batch 60/173] avg loss 0.00011924, throughput 3.88829K wps
[Epoch 33 Batch 90/173] avg loss 0.000112608, throughput 3.90293K wps
[Epoch 33 Batch 120/173] avg loss 0.000122058, throughput 3.89567K wps
[Epoch 33 Batch 150/173] avg loss 0.000143235, throughput 3.87377K wps
Begin Testing...
[Epoch 33] train avg loss 0.000121838, test acc 0.7958, test avg loss 0.776342, throughput 3.90577K wps
[Epoch 34 Batch 30/173] avg loss 0.000135382, throughput 3.96372K wps
[Epoch 34 Batch 60/173] avg loss 0.000121098, throughput 3.87215K wps
[Epoch 34 Batch 90/173] avg loss 0.000113329, throughput 3.87198K wps
[Epoch 34 Batch 120/173] avg loss 9.79469e-05, throughput 3.87526K wps
[Epoch 34 Batch 150/173] avg loss 0.000113063, throughput 3.88768K wps
Begin Testing...
[Epoch 34] train avg loss 0.000116368, test acc 0.7927, test avg loss 0.804329, throughput 3.89277K wps
[Epoch 35 Batch 30/173] avg loss 9.77972e-05, throughput 3.98114K wps
[Epoch 35 Batch 60/173] avg loss 8.7724e-05, throughput 3.86961K wps
[Epoch 35 Batch 90/173] avg loss 0.000117905, throughput 3.86518K wps
[Epoch 35 Batch 120/173] avg loss 0.000121315, throughput 3.88817K wps
[Epoch 35 Batch 150/173] avg loss 0.000101978, throughput 3.88175K wps
Begin Testing...
[Epoch 35] train avg loss 0.00010483, test acc 0.7948, test avg loss 0.816635, throughput 3.90029K wps
[Epoch 36 Batch 30/173] avg loss 7.7245e-05, throughput 4.01102K wps
[Epoch 36 Batch 60/173] avg loss 7.06905e-05, throughput 3.88747K wps
[Epoch 36 Batch 90/173] avg loss 0.000102606, throughput 3.89696K wps
[Epoch 36 Batch 120/173] avg loss 7.96666e-05, throughput 3.91056K wps
[Epoch 36 Batch 150/173] avg loss 8.7445e-05, throughput 3.88219K wps
Begin Testing...
[Epoch 36] train avg loss 8.65917e-05, test acc 0.7979, test avg loss 0.829149, throughput 3.91608K wps
[Epoch 37 Batch 30/173] avg loss 8.90695e-05, throughput 3.98007K wps
[Epoch 37 Batch 60/173] avg loss 7.1294e-05, throughput 3.8839K wps
[Epoch 37 Batch 90/173] avg loss 7.37902e-05, throughput 3.89424K wps
[Epoch 37 Batch 120/173] avg loss 8.58727e-05, throughput 3.87918K wps
[Epoch 37 Batch 150/173] avg loss 9.64884e-05, throughput 3.87863K wps
Begin Testing...
[Epoch 37] train avg loss 8.32743e-05, test acc 0.7958, test avg loss 0.860609, throughput 3.90015K wps
[Epoch 38 Batch 30/173] avg loss 6.44066e-05, throughput 3.97642K wps
[Epoch 38 Batch 60/173] avg loss 6.49793e-05, throughput 3.87646K wps
[Epoch 38 Batch 90/173] avg loss 9.17856e-05, throughput 3.87893K wps
[Epoch 38 Batch 120/173] avg loss 6.40539e-05, throughput 3.88858K wps
[Epoch 38 Batch 150/173] avg loss 8.69101e-05, throughput 3.88158K wps
Begin Testing...
[Epoch 38] train avg loss 7.54632e-05, test acc 0.7927, test avg loss 0.871718, throughput 3.89706K wps
[Epoch 39 Batch 30/173] avg loss 7.03787e-05, throughput 3.98265K wps
[Epoch 39 Batch 60/173] avg loss 7.40764e-05, throughput 3.89198K wps
[Epoch 39 Batch 90/173] avg loss 6.46398e-05, throughput 3.89312K wps
[Epoch 39 Batch 120/173] avg loss 5.88485e-05, throughput 3.89382K wps
[Epoch 39 Batch 150/173] avg loss 5.70097e-05, throughput 3.88323K wps
Begin Testing...
[Epoch 39] train avg loss 6.63667e-05, test acc 0.7937, test avg loss 0.897728, throughput 3.90711K wps
[Epoch 40 Batch 30/173] avg loss 5.57973e-05, throughput 3.96858K wps
[Epoch 40 Batch 60/173] avg loss 5.25157e-05, throughput 3.87538K wps
[Epoch 40 Batch 90/173] avg loss 5.91391e-05, throughput 3.87887K wps
[Epoch 40 Batch 120/173] avg loss 6.24177e-05, throughput 3.87362K wps
[Epoch 40 Batch 150/173] avg loss 4.84972e-05, throughput 3.87329K wps
Begin Testing...
[Epoch 40] train avg loss 5.77215e-05, test acc 0.7917, test avg loss 0.89808, throughput 3.88969K wps
[Epoch 41 Batch 30/173] avg loss 4.11815e-05, throughput 3.97905K wps
[Epoch 41 Batch 60/173] avg loss 5.42301e-05, throughput 3.88552K wps
[Epoch 41 Batch 90/173] avg loss 5.21999e-05, throughput 3.87986K wps
[Epoch 41 Batch 120/173] avg loss 6.15313e-05, throughput 3.87393K wps
[Epoch 41 Batch 150/173] avg loss 8.48535e-05, throughput 3.87443K wps
Begin Testing...
[Epoch 41] train avg loss 5.74574e-05, test acc 0.7885, test avg loss 0.913422, throughput 3.89791K wps
[Epoch 42 Batch 30/173] avg loss 5.07567e-05, throughput 4.01742K wps
[Epoch 42 Batch 60/173] avg loss 4.45554e-05, throughput 3.88808K wps
[Epoch 42 Batch 90/173] avg loss 4.33579e-05, throughput 3.884K wps
[Epoch 42 Batch 120/173] avg loss 4.99778e-05, throughput 3.90634K wps
[Epoch 42 Batch 150/173] avg loss 4.23053e-05, throughput 3.89215K wps
Begin Testing...
[Epoch 42] train avg loss 4.74097e-05, test acc 0.7906, test avg loss 0.930324, throughput 3.91136K wps
[Epoch 43 Batch 30/173] avg loss 3.70739e-05, throughput 3.97037K wps
[Epoch 43 Batch 60/173] avg loss 3.4249e-05, throughput 3.87652K wps
[Epoch 43 Batch 90/173] avg loss 4.25696e-05, throughput 3.87072K wps
[Epoch 43 Batch 120/173] avg loss 5.14207e-05, throughput 3.88328K wps
[Epoch 43 Batch 150/173] avg loss 4.24883e-05, throughput 3.88852K wps
Begin Testing...
[Epoch 43] train avg loss 4.05439e-05, test acc 0.7917, test avg loss 0.953638, throughput 3.89348K wps
[Epoch 44 Batch 30/173] avg loss 3.02821e-05, throughput 3.97826K wps
[Epoch 44 Batch 60/173] avg loss 3.18433e-05, throughput 3.88163K wps
[Epoch 44 Batch 90/173] avg loss 3.90362e-05, throughput 3.8638K wps
[Epoch 44 Batch 120/173] avg loss 4.25757e-05, throughput 3.85447K wps
[Epoch 44 Batch 150/173] avg loss 3.72173e-05, throughput 3.86589K wps
Begin Testing...
[Epoch 44] train avg loss 3.60789e-05, test acc 0.7875, test avg loss 0.971544, throughput 3.88354K wps
[Epoch 45 Batch 30/173] avg loss 3.58916e-05, throughput 3.96681K wps
[Epoch 45 Batch 60/173] avg loss 2.99152e-05, throughput 3.89199K wps
[Epoch 45 Batch 90/173] avg loss 2.92245e-05, throughput 3.88916K wps
[Epoch 45 Batch 120/173] avg loss 3.47943e-05, throughput 3.89142K wps
[Epoch 45 Batch 150/173] avg loss 3.04305e-05, throughput 3.90629K wps
Begin Testing...
[Epoch 45] train avg loss 3.27692e-05, test acc 0.7896, test avg loss 0.982256, throughput 3.9059K wps
[Epoch 46 Batch 30/173] avg loss 2.27945e-05, throughput 3.9831K wps
[Epoch 46 Batch 60/173] avg loss 2.3002e-05, throughput 3.89357K wps
[Epoch 46 Batch 90/173] avg loss 3.03325e-05, throughput 3.88747K wps
[Epoch 46 Batch 120/173] avg loss 3.20791e-05, throughput 3.87703K wps
[Epoch 46 Batch 150/173] avg loss 5.28355e-05, throughput 3.88754K wps
Begin Testing...
[Epoch 46] train avg loss 3.31271e-05, test acc 0.7937, test avg loss 0.99553, throughput 3.90144K wps
[Epoch 47 Batch 30/173] avg loss 2.36473e-05, throughput 3.97542K wps
[Epoch 47 Batch 60/173] avg loss 3.42024e-05, throughput 3.87749K wps
[Epoch 47 Batch 90/173] avg loss 7.10405e-05, throughput 3.88729K wps
[Epoch 47 Batch 120/173] avg loss 5.07973e-05, throughput 3.87671K wps
[Epoch 47 Batch 150/173] avg loss 3.44926e-05, throughput 3.87892K wps
Begin Testing...
[Epoch 47] train avg loss 4.04523e-05, test acc 0.7875, test avg loss 1.01186, throughput 3.89875K wps
[Epoch 48 Batch 30/173] avg loss 2.70294e-05, throughput 3.95828K wps
[Epoch 48 Batch 60/173] avg loss 3.10505e-05, throughput 3.87996K wps
[Epoch 48 Batch 90/173] avg loss 2.89509e-05, throughput 3.89554K wps
[Epoch 48 Batch 120/173] avg loss 2.38566e-05, throughput 3.89035K wps
[Epoch 48 Batch 150/173] avg loss 2.70412e-05, throughput 3.89503K wps
Begin Testing...
[Epoch 48] train avg loss 3.09402e-05, test acc 0.7823, test avg loss 1.03472, throughput 3.90319K wps
[Epoch 49 Batch 30/173] avg loss 3.41906e-05, throughput 4.00241K wps
[Epoch 49 Batch 60/173] avg loss 3.74848e-05, throughput 3.88928K wps
[Epoch 49 Batch 90/173] avg loss 3.87876e-05, throughput 3.88043K wps
[Epoch 49 Batch 120/173] avg loss 2.04231e-05, throughput 3.90346K wps
[Epoch 49 Batch 150/173] avg loss 2.22341e-05, throughput 3.88015K wps
Begin Testing...
[Epoch 49] train avg loss 2.89998e-05, test acc 0.7844, test avg loss 1.05585, throughput 3.90575K wps
[Epoch 50 Batch 30/173] avg loss 2.08509e-05, throughput 3.98183K wps
[Epoch 50 Batch 60/173] avg loss 1.91694e-05, throughput 3.86579K wps
[Epoch 50 Batch 90/173] avg loss 2.19427e-05, throughput 3.8599K wps
[Epoch 50 Batch 120/173] avg loss 2.37646e-05, throughput 3.86998K wps
[Epoch 50 Batch 150/173] avg loss 1.64848e-05, throughput 3.86605K wps
Begin Testing...
[Epoch 50] train avg loss 2.02693e-05, test acc 0.7875, test avg loss 1.07896, throughput 3.88558K wps
[Epoch 51 Batch 30/173] avg loss 2.33928e-05, throughput 3.97227K wps
[Epoch 51 Batch 60/173] avg loss 1.58132e-05, throughput 3.87217K wps
[Epoch 51 Batch 90/173] avg loss 1.7034e-05, throughput 3.86872K wps
[Epoch 51 Batch 120/173] avg loss 1.8399e-05, throughput 3.86674K wps
[Epoch 51 Batch 150/173] avg loss 2.02257e-05, throughput 3.89242K wps
Begin Testing...
[Epoch 51] train avg loss 1.88057e-05, test acc 0.7885, test avg loss 1.08513, throughput 3.89369K wps
[Epoch 52 Batch 30/173] avg loss 1.62772e-05, throughput 3.98484K wps
[Epoch 52 Batch 60/173] avg loss 1.24714e-05, throughput 3.8616K wps
[Epoch 52 Batch 90/173] avg loss 1.61064e-05, throughput 3.88061K wps
[Epoch 52 Batch 120/173] avg loss 1.73634e-05, throughput 3.88002K wps
[Epoch 52 Batch 150/173] avg loss 1.43076e-05, throughput 3.88515K wps
Begin Testing...
[Epoch 52] train avg loss 1.59337e-05, test acc 0.7896, test avg loss 1.10455, throughput 3.89964K wps
[Epoch 53 Batch 30/173] avg loss 1.45425e-05, throughput 3.97815K wps
[Epoch 53 Batch 60/173] avg loss 1.50482e-05, throughput 3.88803K wps
[Epoch 53 Batch 90/173] avg loss 1.99059e-05, throughput 3.88794K wps
[Epoch 53 Batch 120/173] avg loss 1.56307e-05, throughput 3.88036K wps
[Epoch 53 Batch 150/173] avg loss 1.02563e-05, throughput 3.88855K wps
Begin Testing...
[Epoch 53] train avg loss 1.54896e-05, test acc 0.7875, test avg loss 1.11797, throughput 3.90383K wps
[Epoch 54 Batch 30/173] avg loss 1.25369e-05, throughput 3.98663K wps
[Epoch 54 Batch 60/173] avg loss 1.36016e-05, throughput 3.88258K wps
[Epoch 54 Batch 90/173] avg loss 1.76443e-05, throughput 3.86381K wps
[Epoch 54 Batch 120/173] avg loss 1.45894e-05, throughput 3.88917K wps
[Epoch 54 Batch 150/173] avg loss 1.59231e-05, throughput 3.87009K wps
Begin Testing...
[Epoch 54] train avg loss 1.50776e-05, test acc 0.7906, test avg loss 1.12905, throughput 3.90109K wps
[Epoch 55 Batch 30/173] avg loss 1.14384e-05, throughput 3.99261K wps
[Epoch 55 Batch 60/173] avg loss 1.02346e-05, throughput 3.87815K wps
[Epoch 55 Batch 90/173] avg loss 1.19817e-05, throughput 3.88652K wps
[Epoch 55 Batch 120/173] avg loss 1.54915e-05, throughput 3.87333K wps
[Epoch 55 Batch 150/173] avg loss 1.19774e-05, throughput 3.89348K wps
Begin Testing...
[Epoch 55] train avg loss 1.23368e-05, test acc 0.7885, test avg loss 1.14554, throughput 3.9017K wps
[Epoch 56 Batch 30/173] avg loss 8.09968e-06, throughput 3.97293K wps
[Epoch 56 Batch 60/173] avg loss 1.3239e-05, throughput 3.87747K wps
[Epoch 56 Batch 90/173] avg loss 1.19575e-05, throughput 3.87423K wps
[Epoch 56 Batch 120/173] avg loss 1.05802e-05, throughput 3.86861K wps
[Epoch 56 Batch 150/173] avg loss 1.18366e-05, throughput 3.87327K wps
Begin Testing...
[Epoch 56] train avg loss 1.20808e-05, test acc 0.7896, test avg loss 1.16268, throughput 3.89217K wps
[Epoch 57 Batch 30/173] avg loss 1.06087e-05, throughput 3.96785K wps
[Epoch 57 Batch 60/173] avg loss 9.84701e-06, throughput 3.87699K wps
[Epoch 57 Batch 90/173] avg loss 1.24752e-05, throughput 3.87437K wps
[Epoch 57 Batch 120/173] avg loss 1.47453e-05, throughput 3.88291K wps
[Epoch 57 Batch 150/173] avg loss 9.45894e-06, throughput 3.89206K wps
Begin Testing...
[Epoch 57] train avg loss 1.12076e-05, test acc 0.7927, test avg loss 1.17535, throughput 3.89556K wps
[Epoch 58 Batch 30/173] avg loss 1.14236e-05, throughput 3.99693K wps
[Epoch 58 Batch 60/173] avg loss 9.3526e-06, throughput 3.88238K wps
[Epoch 58 Batch 90/173] avg loss 9.41386e-06, throughput 3.8739K wps
[Epoch 58 Batch 120/173] avg loss 1.23959e-05, throughput 3.89796K wps
[Epoch 58 Batch 150/173] avg loss 1.11506e-05, throughput 3.89287K wps
Begin Testing...
[Epoch 58] train avg loss 1.1085e-05, test acc 0.7896, test avg loss 1.20502, throughput 3.90241K wps
[Epoch 59 Batch 30/173] avg loss 8.8837e-06, throughput 3.97706K wps
[Epoch 59 Batch 60/173] avg loss 1.03558e-05, throughput 3.86932K wps
[Epoch 59 Batch 90/173] avg loss 7.84946e-06, throughput 3.87685K wps
[Epoch 59 Batch 120/173] avg loss 8.13605e-06, throughput 3.86963K wps
[Epoch 59 Batch 150/173] avg loss 1.41969e-05, throughput 3.88155K wps
Begin Testing...
[Epoch 59] train avg loss 1.03796e-05, test acc 0.7854, test avg loss 1.21274, throughput 3.89421K wps
Test loss 0.439513, test acc 0.8049
Total time cost 554.57s
9596 1066
[Epoch 0 Batch 30/173] avg loss 0.0161264, throughput 3.65569K wps
[Epoch 0 Batch 60/173] avg loss 0.0149435, throughput 3.84488K wps
[Epoch 0 Batch 90/173] avg loss 0.0147136, throughput 3.86208K wps
[Epoch 0 Batch 120/173] avg loss 0.0141949, throughput 3.88983K wps
[Epoch 0 Batch 150/173] avg loss 0.0141473, throughput 3.87977K wps
Begin Testing...
[Epoch 0] train avg loss 0.0147197, test acc 0.6156, test avg loss 0.658216, throughput 3.8308K wps
Observed Improvement.
Begin Testing...
[Epoch 1 Batch 30/173] avg loss 0.0133945, throughput 3.98576K wps
[Epoch 1 Batch 60/173] avg loss 0.0133003, throughput 3.87524K wps
[Epoch 1 Batch 90/173] avg loss 0.0130908, throughput 3.87462K wps
[Epoch 1 Batch 120/173] avg loss 0.013111, throughput 3.86955K wps
[Epoch 1 Batch 150/173] avg loss 0.0130915, throughput 3.86702K wps
Begin Testing...
[Epoch 1] train avg loss 0.0131503, test acc 0.6667, test avg loss 0.632523, throughput 3.88842K wps
Observed Improvement.
Begin Testing...
[Epoch 2 Batch 30/173] avg loss 0.0123692, throughput 3.98153K wps
[Epoch 2 Batch 60/173] avg loss 0.0121804, throughput 3.87664K wps
[Epoch 2 Batch 90/173] avg loss 0.0119501, throughput 3.8745K wps
[Epoch 2 Batch 120/173] avg loss 0.0119224, throughput 3.87592K wps
[Epoch 2 Batch 150/173] avg loss 0.0121249, throughput 3.8658K wps
Begin Testing...
[Epoch 2] train avg loss 0.0120783, test acc 0.6958, test avg loss 0.600375, throughput 3.89376K wps
Observed Improvement.
Begin Testing...
[Epoch 3 Batch 30/173] avg loss 0.0114303, throughput 3.97053K wps
[Epoch 3 Batch 60/173] avg loss 0.0115226, throughput 3.86968K wps
[Epoch 3 Batch 90/173] avg loss 0.0111818, throughput 3.87269K wps
[Epoch 3 Batch 120/173] avg loss 0.0109741, throughput 3.89371K wps
[Epoch 3 Batch 150/173] avg loss 0.0108923, throughput 3.89887K wps
Begin Testing...
[Epoch 3] train avg loss 0.0111683, test acc 0.7198, test avg loss 0.569375, throughput 3.89585K wps
Observed Improvement.
Begin Testing...
[Epoch 4 Batch 30/173] avg loss 0.0103272, throughput 3.97527K wps
[Epoch 4 Batch 60/173] avg loss 0.0105045, throughput 3.88354K wps
[Epoch 4 Batch 90/173] avg loss 0.0102648, throughput 3.86629K wps
[Epoch 4 Batch 120/173] avg loss 0.0102311, throughput 3.88062K wps
[Epoch 4 Batch 150/173] avg loss 0.0101748, throughput 3.88004K wps
Begin Testing...
[Epoch 4] train avg loss 0.010234, test acc 0.7531, test avg loss 0.53602, throughput 3.89607K wps
Observed Improvement.
Begin Testing...
[Epoch 5 Batch 30/173] avg loss 0.00949476, throughput 3.96791K wps
[Epoch 5 Batch 60/173] avg loss 0.00931909, throughput 3.87424K wps
[Epoch 5 Batch 90/173] avg loss 0.0091308, throughput 3.87191K wps
[Epoch 5 Batch 120/173] avg loss 0.00892052, throughput 3.87987K wps
[Epoch 5 Batch 150/173] avg loss 0.00894038, throughput 3.90796K wps
Begin Testing...
[Epoch 5] train avg loss 0.00908389, test acc 0.7542, test avg loss 0.503363, throughput 3.89782K wps
Observed Improvement.
Begin Testing...
[Epoch 6 Batch 30/173] avg loss 0.00800384, throughput 3.99834K wps
[Epoch 6 Batch 60/173] avg loss 0.0081899, throughput 3.88995K wps
[Epoch 6 Batch 90/173] avg loss 0.00798711, throughput 3.88501K wps
[Epoch 6 Batch 120/173] avg loss 0.00837477, throughput 3.87191K wps
[Epoch 6 Batch 150/173] avg loss 0.00773959, throughput 3.8858K wps
Begin Testing...
[Epoch 6] train avg loss 0.00804294, test acc 0.7667, test avg loss 0.477141, throughput 3.90316K wps
Observed Improvement.
Begin Testing...
[Epoch 7 Batch 30/173] avg loss 0.00748529, throughput 3.95921K wps
[Epoch 7 Batch 60/173] avg loss 0.00704024, throughput 3.87197K wps
[Epoch 7 Batch 90/173] avg loss 0.0069848, throughput 3.87039K wps
[Epoch 7 Batch 120/173] avg loss 0.00699669, throughput 3.87803K wps
[Epoch 7 Batch 150/173] avg loss 0.00696755, throughput 3.8885K wps
Begin Testing...
[Epoch 7] train avg loss 0.00706973, test acc 0.7750, test avg loss 0.462402, throughput 3.89069K wps
Observed Improvement.
Begin Testing...
[Epoch 8 Batch 30/173] avg loss 0.00652737, throughput 3.94626K wps
[Epoch 8 Batch 60/173] avg loss 0.00624299, throughput 3.88099K wps
[Epoch 8 Batch 90/173] avg loss 0.0059398, throughput 3.8749K wps
[Epoch 8 Batch 120/173] avg loss 0.00649344, throughput 3.89936K wps
[Epoch 8 Batch 150/173] avg loss 0.0063673, throughput 3.88438K wps
Begin Testing...
[Epoch 8] train avg loss 0.00623653, test acc 0.7750, test avg loss 0.454074, throughput 3.89447K wps
Observed Improvement.
Begin Testing...
[Epoch 9 Batch 30/173] avg loss 0.00540839, throughput 3.99878K wps
[Epoch 9 Batch 60/173] avg loss 0.00551731, throughput 3.87248K wps
[Epoch 9 Batch 90/173] avg loss 0.00533344, throughput 3.87778K wps
[Epoch 9 Batch 120/173] avg loss 0.00513181, throughput 3.86292K wps
[Epoch 9 Batch 150/173] avg loss 0.00544198, throughput 3.87506K wps
Begin Testing...
[Epoch 9] train avg loss 0.00540523, test acc 0.7729, test avg loss 0.447098, throughput 3.89635K wps
[Epoch 10 Batch 30/173] avg loss 0.00467491, throughput 3.9711K wps
[Epoch 10 Batch 60/173] avg loss 0.00475561, throughput 3.88489K wps
[Epoch 10 Batch 90/173] avg loss 0.00482791, throughput 3.88559K wps
[Epoch 10 Batch 120/173] avg loss 0.00489775, throughput 3.88042K wps
[Epoch 10 Batch 150/173] avg loss 0.00454886, throughput 3.86982K wps
Begin Testing...
[Epoch 10] train avg loss 0.00471434, test acc 0.7812, test avg loss 0.447272, throughput 3.89326K wps
Observed Improvement.
Begin Testing...
[Epoch 11 Batch 30/173] avg loss 0.0041748, throughput 3.99355K wps
[Epoch 11 Batch 60/173] avg loss 0.00385691, throughput 3.88317K wps
[Epoch 11 Batch 90/173] avg loss 0.0037755, throughput 3.87915K wps
[Epoch 11 Batch 120/173] avg loss 0.00400323, throughput 3.90599K wps
[Epoch 11 Batch 150/173] avg loss 0.00433416, throughput 3.87834K wps
Begin Testing...
[Epoch 11] train avg loss 0.00402117, test acc 0.7708, test avg loss 0.456637, throughput 3.90225K wps
[Epoch 12 Batch 30/173] avg loss 0.00334765, throughput 3.95986K wps
[Epoch 12 Batch 60/173] avg loss 0.0035047, throughput 3.86098K wps
[Epoch 12 Batch 90/173] avg loss 0.0034853, throughput 3.86845K wps
[Epoch 12 Batch 120/173] avg loss 0.00342955, throughput 3.87381K wps
[Epoch 12 Batch 150/173] avg loss 0.00358041, throughput 3.87842K wps
Begin Testing...
[Epoch 12] train avg loss 0.00344192, test acc 0.7750, test avg loss 0.45799, throughput 3.88808K wps
[Epoch 13 Batch 30/173] avg loss 0.00299034, throughput 3.96575K wps
[Epoch 13 Batch 60/173] avg loss 0.00291132, throughput 3.86521K wps
[Epoch 13 Batch 90/173] avg loss 0.00275212, throughput 3.86509K wps
[Epoch 13 Batch 120/173] avg loss 0.00277729, throughput 3.89779K wps
[Epoch 13 Batch 150/173] avg loss 0.00304234, throughput 3.88955K wps
Begin Testing...
[Epoch 13] train avg loss 0.00288926, test acc 0.7729, test avg loss 0.472181, throughput 3.89306K wps
[Epoch 14 Batch 30/173] avg loss 0.00249426, throughput 3.98043K wps
[Epoch 14 Batch 60/173] avg loss 0.00253629, throughput 3.88338K wps
[Epoch 14 Batch 90/173] avg loss 0.00249953, throughput 3.87524K wps
[Epoch 14 Batch 120/173] avg loss 0.00260252, throughput 3.87781K wps
[Epoch 14 Batch 150/173] avg loss 0.00224171, throughput 3.88808K wps
Begin Testing...
[Epoch 14] train avg loss 0.00249092, test acc 0.7729, test avg loss 0.482806, throughput 3.89642K wps
[Epoch 15 Batch 30/173] avg loss 0.00208968, throughput 3.97822K wps
[Epoch 15 Batch 60/173] avg loss 0.00211713, throughput 3.88418K wps
[Epoch 15 Batch 90/173] avg loss 0.00210217, throughput 3.88203K wps
[Epoch 15 Batch 120/173] avg loss 0.00222657, throughput 3.87089K wps
[Epoch 15 Batch 150/173] avg loss 0.00206513, throughput 3.86881K wps
Begin Testing...
[Epoch 15] train avg loss 0.00209693, test acc 0.7771, test avg loss 0.499808, throughput 3.89725K wps
[Epoch 16 Batch 30/173] avg loss 0.0016877, throughput 3.97274K wps
[Epoch 16 Batch 60/173] avg loss 0.00169574, throughput 3.88805K wps
[Epoch 16 Batch 90/173] avg loss 0.00165878, throughput 3.89865K wps
[Epoch 16 Batch 120/173] avg loss 0.00182345, throughput 3.87837K wps
[Epoch 16 Batch 150/173] avg loss 0.00176618, throughput 3.88125K wps
Begin Testing...
[Epoch 16] train avg loss 0.00175549, test acc 0.7729, test avg loss 0.510402, throughput 3.90462K wps
[Epoch 17 Batch 30/173] avg loss 0.00144623, throughput 3.96629K wps
[Epoch 17 Batch 60/173] avg loss 0.00135263, throughput 3.87742K wps
[Epoch 17 Batch 90/173] avg loss 0.00135229, throughput 3.86507K wps
[Epoch 17 Batch 120/173] avg loss 0.00134735, throughput 3.88723K wps
[Epoch 17 Batch 150/173] avg loss 0.00163225, throughput 3.88009K wps
Begin Testing...
[Epoch 17] train avg loss 0.00141197, test acc 0.7760, test avg loss 0.539704, throughput 3.89177K wps
[Epoch 18 Batch 30/173] avg loss 0.0013808, throughput 3.97402K wps
[Epoch 18 Batch 60/173] avg loss 0.00114993, throughput 3.87147K wps
[Epoch 18 Batch 90/173] avg loss 0.00123133, throughput 3.86296K wps
[Epoch 18 Batch 120/173] avg loss 0.0013842, throughput 3.87557K wps
[Epoch 18 Batch 150/173] avg loss 0.00122985, throughput 3.88303K wps
Begin Testing...
[Epoch 18] train avg loss 0.0012495, test acc 0.7688, test avg loss 0.556162, throughput 3.89662K wps
[Epoch 19 Batch 30/173] avg loss 0.00115673, throughput 3.99916K wps
[Epoch 19 Batch 60/173] avg loss 0.00101992, throughput 3.87752K wps
[Epoch 19 Batch 90/173] avg loss 0.00112788, throughput 3.87888K wps
[Epoch 19 Batch 120/173] avg loss 0.00102605, throughput 3.89152K wps
[Epoch 19 Batch 150/173] avg loss 0.000941689, throughput 3.8888K wps
Begin Testing...
[Epoch 19] train avg loss 0.00107595, test acc 0.7667, test avg loss 0.579285, throughput 3.90236K wps
[Epoch 20 Batch 30/173] avg loss 0.000842979, throughput 3.97275K wps
[Epoch 20 Batch 60/173] avg loss 0.000900397, throughput 3.87627K wps
[Epoch 20 Batch 90/173] avg loss 0.000852386, throughput 3.87129K wps
[Epoch 20 Batch 120/173] avg loss 0.000888894, throughput 3.87668K wps
[Epoch 20 Batch 150/173] avg loss 0.000866485, throughput 3.87381K wps
Begin Testing...
[Epoch 20] train avg loss 0.000885295, test acc 0.7708, test avg loss 0.599748, throughput 3.89541K wps
[Epoch 21 Batch 30/173] avg loss 0.000679669, throughput 3.99448K wps
[Epoch 21 Batch 60/173] avg loss 0.000757232, throughput 3.87754K wps
[Epoch 21 Batch 90/173] avg loss 0.000777952, throughput 3.86258K wps
[Epoch 21 Batch 120/173] avg loss 0.000856816, throughput 3.85362K wps
[Epoch 21 Batch 150/173] avg loss 0.000829839, throughput 3.87019K wps
Begin Testing...
[Epoch 21] train avg loss 0.000769517, test acc 0.7646, test avg loss 0.618375, throughput 3.88938K wps
[Epoch 22 Batch 30/173] avg loss 0.000561705, throughput 3.98064K wps
[Epoch 22 Batch 60/173] avg loss 0.000589455, throughput 3.88242K wps
[Epoch 22 Batch 90/173] avg loss 0.000646363, throughput 3.91092K wps
[Epoch 22 Batch 120/173] avg loss 0.000727328, throughput 3.87438K wps
[Epoch 22 Batch 150/173] avg loss 0.000671902, throughput 3.87203K wps
Begin Testing...
[Epoch 22] train avg loss 0.000632, test acc 0.7646, test avg loss 0.646085, throughput 3.90009K wps
[Epoch 23 Batch 30/173] avg loss 0.000459736, throughput 3.95261K wps
[Epoch 23 Batch 60/173] avg loss 0.000513224, throughput 3.87173K wps
[Epoch 23 Batch 90/173] avg loss 0.000621266, throughput 3.87632K wps
[Epoch 23 Batch 120/173] avg loss 0.000587416, throughput 3.87067K wps
[Epoch 23 Batch 150/173] avg loss 0.000495416, throughput 3.89539K wps
Begin Testing...
[Epoch 23] train avg loss 0.000546783, test acc 0.7667, test avg loss 0.670006, throughput 3.89302K wps
[Epoch 24 Batch 30/173] avg loss 0.000451322, throughput 3.9628K wps
[Epoch 24 Batch 60/173] avg loss 0.000398195, throughput 3.88242K wps
[Epoch 24 Batch 90/173] avg loss 0.000420498, throughput 3.89234K wps
[Epoch 24 Batch 120/173] avg loss 0.000406603, throughput 3.88412K wps
[Epoch 24 Batch 150/173] avg loss 0.000481998, throughput 3.89285K wps
Begin Testing...
[Epoch 24] train avg loss 0.000447605, test acc 0.7615, test avg loss 0.692574, throughput 3.9035K wps
[Epoch 25 Batch 30/173] avg loss 0.000404664, throughput 3.99798K wps
[Epoch 25 Batch 60/173] avg loss 0.000390654, throughput 3.87875K wps
[Epoch 25 Batch 90/173] avg loss 0.000330289, throughput 3.87628K wps
[Epoch 25 Batch 120/173] avg loss 0.000397908, throughput 3.87891K wps
[Epoch 25 Batch 150/173] avg loss 0.000396552, throughput 3.87304K wps
Begin Testing...
[Epoch 25] train avg loss 0.000383173, test acc 0.7677, test avg loss 0.722635, throughput 3.89829K wps
[Epoch 26 Batch 30/173] avg loss 0.000307195, throughput 3.99272K wps
[Epoch 26 Batch 60/173] avg loss 0.000352623, throughput 3.88207K wps
[Epoch 26 Batch 90/173] avg loss 0.000364994, throughput 3.88005K wps
[Epoch 26 Batch 120/173] avg loss 0.000397305, throughput 3.87116K wps
[Epoch 26 Batch 150/173] avg loss 0.000336909, throughput 3.86613K wps
Begin Testing...
[Epoch 26] train avg loss 0.000350605, test acc 0.7656, test avg loss 0.743566, throughput 3.89264K wps
[Epoch 27 Batch 30/173] avg loss 0.000312398, throughput 3.9796K wps
[Epoch 27 Batch 60/173] avg loss 0.000287235, throughput 3.88625K wps
[Epoch 27 Batch 90/173] avg loss 0.000307342, throughput 3.89242K wps
[Epoch 27 Batch 120/173] avg loss 0.000283767, throughput 3.89921K wps
[Epoch 27 Batch 150/173] avg loss 0.000318135, throughput 3.88153K wps
Begin Testing...
[Epoch 27] train avg loss 0.000299386, test acc 0.7625, test avg loss 0.766954, throughput 3.90337K wps
[Epoch 28 Batch 30/173] avg loss 0.000289664, throughput 3.97062K wps
[Epoch 28 Batch 60/173] avg loss 0.000243545, throughput 3.87163K wps
[Epoch 28 Batch 90/173] avg loss 0.000233614, throughput 3.87956K wps
[Epoch 28 Batch 120/173] avg loss 0.000249995, throughput 3.89331K wps
[Epoch 28 Batch 150/173] avg loss 0.00022796, throughput 3.88618K wps
Begin Testing...
[Epoch 28] train avg loss 0.000251763, test acc 0.7677, test avg loss 0.783474, throughput 3.89561K wps
[Epoch 29 Batch 30/173] avg loss 0.000234513, throughput 3.97833K wps
[Epoch 29 Batch 60/173] avg loss 0.0002247, throughput 3.86713K wps
[Epoch 29 Batch 90/173] avg loss 0.000220776, throughput 3.87493K wps
[Epoch 29 Batch 120/173] avg loss 0.000215407, throughput 3.89401K wps
[Epoch 29 Batch 150/173] avg loss 0.000212168, throughput 3.92034K wps
Begin Testing...
[Epoch 29] train avg loss 0.000227982, test acc 0.7646, test avg loss 0.806296, throughput 3.90291K wps
[Epoch 30 Batch 30/173] avg loss 0.000159255, throughput 3.98035K wps
[Epoch 30 Batch 60/173] avg loss 0.000212165, throughput 3.89744K wps
[Epoch 30 Batch 90/173] avg loss 0.000173206, throughput 3.91539K wps
[Epoch 30 Batch 120/173] avg loss 0.000153714, throughput 3.87445K wps
[Epoch 30 Batch 150/173] avg loss 0.000148888, throughput 3.87398K wps
Begin Testing...
[Epoch 30] train avg loss 0.000176529, test acc 0.7594, test avg loss 0.841098, throughput 3.90593K wps
[Epoch 31 Batch 30/173] avg loss 0.000182014, throughput 3.96129K wps
[Epoch 31 Batch 60/173] avg loss 0.00015296, throughput 3.88009K wps
[Epoch 31 Batch 90/173] avg loss 0.000176843, throughput 3.89505K wps
[Epoch 31 Batch 120/173] avg loss 0.000199294, throughput 3.8907K wps
[Epoch 31 Batch 150/173] avg loss 0.000177491, throughput 3.87977K wps
Begin Testing...
[Epoch 31] train avg loss 0.000180899, test acc 0.7667, test avg loss 0.843506, throughput 3.89665K wps
[Epoch 32 Batch 30/173] avg loss 0.000138725, throughput 3.97048K wps
[Epoch 32 Batch 60/173] avg loss 0.000146547, throughput 3.8651K wps
[Epoch 32 Batch 90/173] avg loss 0.000151687, throughput 3.88498K wps
[Epoch 32 Batch 120/173] avg loss 0.00013262, throughput 3.88769K wps
[Epoch 32 Batch 150/173] avg loss 0.00016517, throughput 3.90606K wps
Begin Testing...
[Epoch 32] train avg loss 0.000142665, test acc 0.7625, test avg loss 0.871483, throughput 3.90144K wps
[Epoch 33 Batch 30/173] avg loss 9.43894e-05, throughput 3.97871K wps
[Epoch 33 Batch 60/173] avg loss 0.000119358, throughput 3.87054K wps
[Epoch 33 Batch 90/173] avg loss 0.000151585, throughput 3.86716K wps
[Epoch 33 Batch 120/173] avg loss 0.000119525, throughput 3.86787K wps
[Epoch 33 Batch 150/173] avg loss 0.000132513, throughput 3.87967K wps
Begin Testing...
[Epoch 33] train avg loss 0.000130255, test acc 0.7688, test avg loss 0.897717, throughput 3.89302K wps
[Epoch 34 Batch 30/173] avg loss 0.000115579, throughput 3.97242K wps
[Epoch 34 Batch 60/173] avg loss 0.000119464, throughput 3.89379K wps
[Epoch 34 Batch 90/173] avg loss 0.000122235, throughput 3.87469K wps
[Epoch 34 Batch 120/173] avg loss 0.000123368, throughput 3.86779K wps
[Epoch 34 Batch 150/173] avg loss 0.000123586, throughput 3.86902K wps
Begin Testing...
[Epoch 34] train avg loss 0.000120339, test acc 0.7615, test avg loss 0.919556, throughput 3.89279K wps
[Epoch 35 Batch 30/173] avg loss 7.7057e-05, throughput 3.98681K wps
[Epoch 35 Batch 60/173] avg loss 9.27559e-05, throughput 3.90288K wps
[Epoch 35 Batch 90/173] avg loss 0.000106429, throughput 3.90972K wps
[Epoch 35 Batch 120/173] avg loss 0.000122894, throughput 3.88897K wps
[Epoch 35 Batch 150/173] avg loss 0.000119169, throughput 3.89976K wps
Begin Testing...
[Epoch 35] train avg loss 0.000105141, test acc 0.7656, test avg loss 0.940307, throughput 3.91332K wps
[Epoch 36 Batch 30/173] avg loss 7.9375e-05, throughput 3.99199K wps
[Epoch 36 Batch 60/173] avg loss 8.62549e-05, throughput 3.87882K wps
[Epoch 36 Batch 90/173] avg loss 8.33672e-05, throughput 3.87097K wps
[Epoch 36 Batch 120/173] avg loss 7.99067e-05, throughput 3.87854K wps
[Epoch 36 Batch 150/173] avg loss 7.20503e-05, throughput 3.87308K wps
Begin Testing...
[Epoch 36] train avg loss 8.21342e-05, test acc 0.7625, test avg loss 0.968476, throughput 3.89949K wps
[Epoch 37 Batch 30/173] avg loss 6.87712e-05, throughput 3.97385K wps
[Epoch 37 Batch 60/173] avg loss 8.90578e-05, throughput 3.88519K wps
[Epoch 37 Batch 90/173] avg loss 9.4073e-05, throughput 3.86626K wps
[Epoch 37 Batch 120/173] avg loss 9.17144e-05, throughput 3.87575K wps
[Epoch 37 Batch 150/173] avg loss 7.69833e-05, throughput 3.88782K wps
Begin Testing...
[Epoch 37] train avg loss 8.61775e-05, test acc 0.7656, test avg loss 0.992869, throughput 3.90008K wps
[Epoch 38 Batch 30/173] avg loss 6.49795e-05, throughput 4.00662K wps
[Epoch 38 Batch 60/173] avg loss 6.83686e-05, throughput 3.88947K wps
[Epoch 38 Batch 90/173] avg loss 6.30117e-05, throughput 3.88221K wps
[Epoch 38 Batch 120/173] avg loss 6.2631e-05, throughput 3.87748K wps
[Epoch 38 Batch 150/173] avg loss 8.38584e-05, throughput 3.88129K wps
Begin Testing...
[Epoch 38] train avg loss 6.8333e-05, test acc 0.7562, test avg loss 1.01148, throughput 3.90337K wps
[Epoch 39 Batch 30/173] avg loss 7.03718e-05, throughput 3.94057K wps
[Epoch 39 Batch 60/173] avg loss 5.06818e-05, throughput 3.87073K wps
[Epoch 39 Batch 90/173] avg loss 6.67349e-05, throughput 3.86859K wps
[Epoch 39 Batch 120/173] avg loss 7.21289e-05, throughput 3.89032K wps
[Epoch 39 Batch 150/173] avg loss 7.92328e-05, throughput 3.88946K wps
Begin Testing...
[Epoch 39] train avg loss 6.67368e-05, test acc 0.7531, test avg loss 1.03397, throughput 3.88973K wps
[Epoch 40 Batch 30/173] avg loss 7.3364e-05, throughput 3.97052K wps
[Epoch 40 Batch 60/173] avg loss 7.27549e-05, throughput 3.86827K wps
[Epoch 40 Batch 90/173] avg loss 4.87883e-05, throughput 3.87388K wps
[Epoch 40 Batch 120/173] avg loss 6.35304e-05, throughput 3.89481K wps
[Epoch 40 Batch 150/173] avg loss 8.62451e-05, throughput 3.89567K wps
Begin Testing...
[Epoch 40] train avg loss 6.73214e-05, test acc 0.7458, test avg loss 1.0734, throughput 3.89875K wps
[Epoch 41 Batch 30/173] avg loss 5.15987e-05, throughput 3.98139K wps
[Epoch 41 Batch 60/173] avg loss 4.6988e-05, throughput 3.89147K wps
[Epoch 41 Batch 90/173] avg loss 4.77711e-05, throughput 3.88479K wps
[Epoch 41 Batch 120/173] avg loss 6.62195e-05, throughput 3.88753K wps
[Epoch 41 Batch 150/173] avg loss 6.51973e-05, throughput 3.87486K wps
Begin Testing...
[Epoch 41] train avg loss 5.46261e-05, test acc 0.7531, test avg loss 1.07727, throughput 3.89891K wps
[Epoch 42 Batch 30/173] avg loss 5.43616e-05, throughput 3.98009K wps
[Epoch 42 Batch 60/173] avg loss 6.48919e-05, throughput 3.88307K wps
[Epoch 42 Batch 90/173] avg loss 4.15845e-05, throughput 3.86951K wps
[Epoch 42 Batch 120/173] avg loss 4.37753e-05, throughput 3.87082K wps
[Epoch 42 Batch 150/173] avg loss 6.3279e-05, throughput 3.87731K wps
Begin Testing...
[Epoch 42] train avg loss 5.30844e-05, test acc 0.7500, test avg loss 1.10976, throughput 3.89343K wps
[Epoch 43 Batch 30/173] avg loss 4.11709e-05, throughput 3.97068K wps
[Epoch 43 Batch 60/173] avg loss 5.35006e-05, throughput 3.91419K wps
[Epoch 43 Batch 90/173] avg loss 3.48924e-05, throughput 3.88314K wps
[Epoch 43 Batch 120/173] avg loss 3.36291e-05, throughput 3.87981K wps
[Epoch 43 Batch 150/173] avg loss 4.56015e-05, throughput 3.88376K wps
Begin Testing...
[Epoch 43] train avg loss 4.41672e-05, test acc 0.7448, test avg loss 1.14463, throughput 3.90673K wps
[Epoch 44 Batch 30/173] avg loss 3.8232e-05, throughput 3.99302K wps
[Epoch 44 Batch 60/173] avg loss 4.80046e-05, throughput 3.87736K wps
[Epoch 44 Batch 90/173] avg loss 5.44009e-05, throughput 3.87806K wps
[Epoch 44 Batch 120/173] avg loss 3.96923e-05, throughput 3.86971K wps
[Epoch 44 Batch 150/173] avg loss 5.10419e-05, throughput 3.88075K wps
Begin Testing...
[Epoch 44] train avg loss 4.62104e-05, test acc 0.7500, test avg loss 1.15671, throughput 3.89775K wps
[Epoch 45 Batch 30/173] avg loss 3.26481e-05, throughput 3.97009K wps
[Epoch 45 Batch 60/173] avg loss 2.94799e-05, throughput 3.89733K wps
[Epoch 45 Batch 90/173] avg loss 3.67459e-05, throughput 3.8791K wps
[Epoch 45 Batch 120/173] avg loss 4.42709e-05, throughput 3.87141K wps
[Epoch 45 Batch 150/173] avg loss 3.04081e-05, throughput 3.87772K wps
Begin Testing...
[Epoch 45] train avg loss 3.68673e-05, test acc 0.7583, test avg loss 1.18288, throughput 3.89601K wps
[Epoch 46 Batch 30/173] avg loss 4.47247e-05, throughput 3.96864K wps
[Epoch 46 Batch 60/173] avg loss 3.47584e-05, throughput 3.89276K wps
[Epoch 46 Batch 90/173] avg loss 3.16124e-05, throughput 3.89754K wps
[Epoch 46 Batch 120/173] avg loss 2.86825e-05, throughput 3.90402K wps
[Epoch 46 Batch 150/173] avg loss 3.26379e-05, throughput 3.88921K wps
Begin Testing...
[Epoch 46] train avg loss 3.32517e-05, test acc 0.7521, test avg loss 1.20571, throughput 3.90854K wps
[Epoch 47 Batch 30/173] avg loss 3.1093e-05, throughput 3.96948K wps
[Epoch 47 Batch 60/173] avg loss 3.32868e-05, throughput 3.87734K wps
[Epoch 47 Batch 90/173] avg loss 3.44902e-05, throughput 3.87745K wps
[Epoch 47 Batch 120/173] avg loss 3.15715e-05, throughput 3.89156K wps
[Epoch 47 Batch 150/173] avg loss 2.60713e-05, throughput 3.87497K wps
Begin Testing...
[Epoch 47] train avg loss 3.38757e-05, test acc 0.7490, test avg loss 1.2309, throughput 3.8946K wps
[Epoch 48 Batch 30/173] avg loss 2.88307e-05, throughput 3.97493K wps
[Epoch 48 Batch 60/173] avg loss 2.9483e-05, throughput 3.87307K wps
[Epoch 48 Batch 90/173] avg loss 2.88247e-05, throughput 3.86614K wps
[Epoch 48 Batch 120/173] avg loss 2.51264e-05, throughput 3.85136K wps
[Epoch 48 Batch 150/173] avg loss 2.01466e-05, throughput 3.87307K wps
Begin Testing...
[Epoch 48] train avg loss 2.73905e-05, test acc 0.7552, test avg loss 1.25207, throughput 3.89137K wps
[Epoch 49 Batch 30/173] avg loss 2.8113e-05, throughput 4.00368K wps
[Epoch 49 Batch 60/173] avg loss 1.9338e-05, throughput 3.88401K wps
[Epoch 49 Batch 90/173] avg loss 2.53441e-05, throughput 3.89299K wps
[Epoch 49 Batch 120/173] avg loss 2.958e-05, throughput 3.89692K wps
[Epoch 49 Batch 150/173] avg loss 2.00944e-05, throughput 3.88597K wps
Begin Testing...
[Epoch 49] train avg loss 2.38014e-05, test acc 0.7552, test avg loss 1.27445, throughput 3.9051K wps
[Epoch 50 Batch 30/173] avg loss 1.68048e-05, throughput 3.97259K wps
[Epoch 50 Batch 60/173] avg loss 3.49689e-05, throughput 3.87637K wps
[Epoch 50 Batch 90/173] avg loss 2.01705e-05, throughput 3.87971K wps
[Epoch 50 Batch 120/173] avg loss 2.23694e-05, throughput 3.8695K wps
[Epoch 50 Batch 150/173] avg loss 1.74169e-05, throughput 3.87457K wps
Begin Testing...
[Epoch 50] train avg loss 2.49538e-05, test acc 0.7469, test avg loss 1.31466, throughput 3.89542K wps
[Epoch 51 Batch 30/173] avg loss 2.35824e-05, throughput 3.96331K wps
[Epoch 51 Batch 60/173] avg loss 2.42188e-05, throughput 3.87906K wps
[Epoch 51 Batch 90/173] avg loss 2.28955e-05, throughput 3.89272K wps
[Epoch 51 Batch 120/173] avg loss 2.12252e-05, throughput 3.89012K wps
[Epoch 51 Batch 150/173] avg loss 2.02077e-05, throughput 3.89016K wps
Begin Testing...
[Epoch 51] train avg loss 2.16331e-05, test acc 0.7531, test avg loss 1.29601, throughput 3.90382K wps
[Epoch 52 Batch 30/173] avg loss 1.89613e-05, throughput 3.9988K wps
[Epoch 52 Batch 60/173] avg loss 1.42837e-05, throughput 3.88251K wps
[Epoch 52 Batch 90/173] avg loss 1.64383e-05, throughput 3.87807K wps
[Epoch 52 Batch 120/173] avg loss 1.98967e-05, throughput 3.87754K wps
[Epoch 52 Batch 150/173] avg loss 1.64949e-05, throughput 3.88082K wps
Begin Testing...
[Epoch 52] train avg loss 1.72661e-05, test acc 0.7521, test avg loss 1.32315, throughput 3.90216K wps
[Epoch 53 Batch 30/173] avg loss 1.99189e-05, throughput 3.99006K wps
[Epoch 53 Batch 60/173] avg loss 1.78405e-05, throughput 3.87239K wps
[Epoch 53 Batch 90/173] avg loss 3.26488e-05, throughput 3.87573K wps
[Epoch 53 Batch 120/173] avg loss 1.46192e-05, throughput 3.87269K wps
[Epoch 53 Batch 150/173] avg loss 2.33214e-05, throughput 3.85954K wps
Begin Testing...
[Epoch 53] train avg loss 2.05258e-05, test acc 0.7490, test avg loss 1.34783, throughput 3.89187K wps
[Epoch 54 Batch 30/173] avg loss 1.93557e-05, throughput 3.98205K wps
[Epoch 54 Batch 60/173] avg loss 1.1222e-05, throughput 3.90535K wps
[Epoch 54 Batch 90/173] avg loss 1.19426e-05, throughput 3.88916K wps
[Epoch 54 Batch 120/173] avg loss 1.35858e-05, throughput 3.88979K wps
[Epoch 54 Batch 150/173] avg loss 1.5693e-05, throughput 3.89809K wps
Begin Testing...
[Epoch 54] train avg loss 1.41798e-05, test acc 0.7438, test avg loss 1.36283, throughput 3.91002K wps
[Epoch 55 Batch 30/173] avg loss 1.11616e-05, throughput 3.96726K wps
[Epoch 55 Batch 60/173] avg loss 1.61887e-05, throughput 3.88144K wps
[Epoch 55 Batch 90/173] avg loss 1.19049e-05, throughput 3.85926K wps
[Epoch 55 Batch 120/173] avg loss 1.25205e-05, throughput 3.87454K wps
[Epoch 55 Batch 150/173] avg loss 1.16578e-05, throughput 3.87292K wps
Begin Testing...
[Epoch 55] train avg loss 1.26384e-05, test acc 0.7469, test avg loss 1.38303, throughput 3.88796K wps
[Epoch 56 Batch 30/173] avg loss 1.50372e-05, throughput 3.95813K wps
[Epoch 56 Batch 60/173] avg loss 9.54449e-06, throughput 3.86608K wps
[Epoch 56 Batch 90/173] avg loss 1.42499e-05, throughput 3.86447K wps
[Epoch 56 Batch 120/173] avg loss 1.41638e-05, throughput 3.87004K wps
[Epoch 56 Batch 150/173] avg loss 1.28506e-05, throughput 3.87775K wps
Begin Testing...
[Epoch 56] train avg loss 1.28559e-05, test acc 0.7438, test avg loss 1.39927, throughput 3.89028K wps
[Epoch 57 Batch 30/173] avg loss 8.87758e-06, throughput 3.99861K wps
[Epoch 57 Batch 60/173] avg loss 1.48458e-05, throughput 3.88308K wps
[Epoch 57 Batch 90/173] avg loss 9.04871e-06, throughput 3.87279K wps
[Epoch 57 Batch 120/173] avg loss 1.06903e-05, throughput 3.88399K wps
[Epoch 57 Batch 150/173] avg loss 8.94411e-06, throughput 3.88821K wps
Begin Testing...
[Epoch 57] train avg loss 1.0277e-05, test acc 0.7417, test avg loss 1.436, throughput 3.90136K wps
[Epoch 58 Batch 30/173] avg loss 8.62444e-06, throughput 3.98563K wps
[Epoch 58 Batch 60/173] avg loss 6.99537e-06, throughput 3.88113K wps
[Epoch 58 Batch 90/173] avg loss 8.31078e-06, throughput 3.87315K wps
[Epoch 58 Batch 120/173] avg loss 1.35574e-05, throughput 3.86947K wps
[Epoch 58 Batch 150/173] avg loss 1.65709e-05, throughput 3.87791K wps
Begin Testing...
[Epoch 58] train avg loss 1.05865e-05, test acc 0.7469, test avg loss 1.44833, throughput 3.89604K wps
[Epoch 59 Batch 30/173] avg loss 8.87733e-06, throughput 3.95991K wps
[Epoch 59 Batch 60/173] avg loss 8.48651e-06, throughput 3.87302K wps
[Epoch 59 Batch 90/173] avg loss 8.03479e-06, throughput 3.8999K wps
[Epoch 59 Batch 120/173] avg loss 9.82652e-06, throughput 3.89392K wps
[Epoch 59 Batch 150/173] avg loss 1.20099e-05, throughput 3.87508K wps
Begin Testing...
[Epoch 59] train avg loss 9.43386e-06, test acc 0.7510, test avg loss 1.4588, throughput 3.89956K wps
Test loss 0.409517, test acc 0.8011
Total time cost 554.83s
9596 1066
[Epoch 0 Batch 30/173] avg loss 0.0155154, throughput 3.69766K wps
[Epoch 0 Batch 60/173] avg loss 0.0148297, throughput 3.88679K wps
[Epoch 0 Batch 90/173] avg loss 0.0147948, throughput 3.87016K wps
[Epoch 0 Batch 120/173] avg loss 0.0142729, throughput 3.8647K wps
[Epoch 0 Batch 150/173] avg loss 0.0140915, throughput 3.85988K wps
Begin Testing...
[Epoch 0] train avg loss 0.0145876, test acc 0.6208, test avg loss 0.644792, throughput 3.83892K wps
Observed Improvement.
Begin Testing...
[Epoch 1 Batch 30/173] avg loss 0.0131339, throughput 3.97939K wps
[Epoch 1 Batch 60/173] avg loss 0.0131523, throughput 3.87482K wps
[Epoch 1 Batch 90/173] avg loss 0.0132151, throughput 3.867K wps
[Epoch 1 Batch 120/173] avg loss 0.0128988, throughput 3.8631K wps
[Epoch 1 Batch 150/173] avg loss 0.0128633, throughput 3.85754K wps
Begin Testing...
[Epoch 1] train avg loss 0.0130035, test acc 0.6896, test avg loss 0.615105, throughput 3.88531K wps
Observed Improvement.
Begin Testing...
[Epoch 2 Batch 30/173] avg loss 0.0120454, throughput 3.98711K wps
[Epoch 2 Batch 60/173] avg loss 0.0121362, throughput 3.88203K wps
[Epoch 2 Batch 90/173] avg loss 0.0118337, throughput 3.87935K wps
[Epoch 2 Batch 120/173] avg loss 0.011801, throughput 3.88836K wps
[Epoch 2 Batch 150/173] avg loss 0.0117562, throughput 3.89726K wps
Begin Testing...
[Epoch 2] train avg loss 0.0119204, test acc 0.7219, test avg loss 0.585513, throughput 3.901K wps
Observed Improvement.
Begin Testing...
[Epoch 3 Batch 30/173] avg loss 0.0111852, throughput 3.95535K wps
[Epoch 3 Batch 60/173] avg loss 0.0108203, throughput 3.85682K wps
[Epoch 3 Batch 90/173] avg loss 0.0107206, throughput 3.86692K wps
[Epoch 3 Batch 120/173] avg loss 0.0108854, throughput 3.8699K wps
[Epoch 3 Batch 150/173] avg loss 0.0108266, throughput 3.8689K wps
Begin Testing...
[Epoch 3] train avg loss 0.0108831, test acc 0.7260, test avg loss 0.557314, throughput 3.88408K wps
Observed Improvement.
Begin Testing...
[Epoch 4 Batch 30/173] avg loss 0.0101162, throughput 3.96775K wps
[Epoch 4 Batch 60/173] avg loss 0.0100614, throughput 3.87395K wps
[Epoch 4 Batch 90/173] avg loss 0.00984933, throughput 3.87829K wps
[Epoch 4 Batch 120/173] avg loss 0.0098037, throughput 3.90001K wps
[Epoch 4 Batch 150/173] avg loss 0.00989224, throughput 3.8911K wps
Begin Testing...
[Epoch 4] train avg loss 0.00991136, test acc 0.7646, test avg loss 0.509011, throughput 3.90298K wps
Observed Improvement.
Begin Testing...
[Epoch 5 Batch 30/173] avg loss 0.00906616, throughput 3.97986K wps
[Epoch 5 Batch 60/173] avg loss 0.00888494, throughput 3.9001K wps
[Epoch 5 Batch 90/173] avg loss 0.00884521, throughput 3.88393K wps
[Epoch 5 Batch 120/173] avg loss 0.00884501, throughput 3.87094K wps
[Epoch 5 Batch 150/173] avg loss 0.00855419, throughput 3.88313K wps
Begin Testing...
[Epoch 5] train avg loss 0.00882403, test acc 0.7875, test avg loss 0.474285, throughput 3.89933K wps
Observed Improvement.
Begin Testing...
[Epoch 6 Batch 30/173] avg loss 0.00802997, throughput 3.97111K wps
[Epoch 6 Batch 60/173] avg loss 0.00783278, throughput 3.88581K wps
[Epoch 6 Batch 90/173] avg loss 0.00782113, throughput 3.87991K wps
[Epoch 6 Batch 120/173] avg loss 0.00783574, throughput 3.8711K wps
[Epoch 6 Batch 150/173] avg loss 0.0079415, throughput 3.86576K wps
Begin Testing...
[Epoch 6] train avg loss 0.00783707, test acc 0.7969, test avg loss 0.449579, throughput 3.88875K wps
Observed Improvement.
Begin Testing...
[Epoch 7 Batch 30/173] avg loss 0.00724109, throughput 3.97576K wps
[Epoch 7 Batch 60/173] avg loss 0.00681322, throughput 3.8982K wps
[Epoch 7 Batch 90/173] avg loss 0.00701176, throughput 3.87909K wps
[Epoch 7 Batch 120/173] avg loss 0.00674281, throughput 3.87988K wps
[Epoch 7 Batch 150/173] avg loss 0.00711537, throughput 3.90068K wps
Begin Testing...
[Epoch 7] train avg loss 0.00694817, test acc 0.7969, test avg loss 0.428819, throughput 3.90361K wps
Observed Improvement.
Begin Testing...
[Epoch 8 Batch 30/173] avg loss 0.00633803, throughput 3.96491K wps
[Epoch 8 Batch 60/173] avg loss 0.00610755, throughput 3.87264K wps
[Epoch 8 Batch 90/173] avg loss 0.00592885, throughput 3.87901K wps
[Epoch 8 Batch 120/173] avg loss 0.00613811, throughput 3.85902K wps
[Epoch 8 Batch 150/173] avg loss 0.00594204, throughput 3.85918K wps
Begin Testing...
[Epoch 8] train avg loss 0.00608515, test acc 0.8063, test avg loss 0.421402, throughput 3.88267K wps
Observed Improvement.
Begin Testing...
[Epoch 9 Batch 30/173] avg loss 0.00532503, throughput 3.98381K wps
[Epoch 9 Batch 60/173] avg loss 0.00524222, throughput 3.8914K wps
[Epoch 9 Batch 90/173] avg loss 0.0051273, throughput 3.87252K wps
[Epoch 9 Batch 120/173] avg loss 0.00538238, throughput 3.87514K wps
[Epoch 9 Batch 150/173] avg loss 0.00519205, throughput 3.86487K wps
Begin Testing...
[Epoch 9] train avg loss 0.00526863, test acc 0.7979, test avg loss 0.410869, throughput 3.89572K wps
[Epoch 10 Batch 30/173] avg loss 0.00462763, throughput 3.97695K wps
[Epoch 10 Batch 60/173] avg loss 0.00458153, throughput 3.89541K wps
[Epoch 10 Batch 90/173] avg loss 0.00471638, throughput 3.90709K wps
[Epoch 10 Batch 120/173] avg loss 0.00483439, throughput 3.88334K wps
[Epoch 10 Batch 150/173] avg loss 0.00447651, throughput 3.88695K wps
Begin Testing...
[Epoch 10] train avg loss 0.00467115, test acc 0.8073, test avg loss 0.411364, throughput 3.91038K wps
Observed Improvement.
Begin Testing...
[Epoch 11 Batch 30/173] avg loss 0.00383454, throughput 3.97166K wps
[Epoch 11 Batch 60/173] avg loss 0.00398748, throughput 3.88299K wps
[Epoch 11 Batch 90/173] avg loss 0.00403564, throughput 3.88061K wps
[Epoch 11 Batch 120/173] avg loss 0.00393517, throughput 3.88357K wps
[Epoch 11 Batch 150/173] avg loss 0.00397362, throughput 3.87277K wps
Begin Testing...
[Epoch 11] train avg loss 0.00399142, test acc 0.8042, test avg loss 0.407563, throughput 3.8938K wps
[Epoch 12 Batch 30/173] avg loss 0.00331972, throughput 3.96864K wps
[Epoch 12 Batch 60/173] avg loss 0.00317841, throughput 3.86708K wps
[Epoch 12 Batch 90/173] avg loss 0.00338757, throughput 3.85949K wps
[Epoch 12 Batch 120/173] avg loss 0.00354451, throughput 3.86869K wps
[Epoch 12 Batch 150/173] avg loss 0.00370108, throughput 3.88399K wps
Begin Testing...
[Epoch 12] train avg loss 0.00343275, test acc 0.8094, test avg loss 0.40551, throughput 3.89348K wps
Observed Improvement.
Begin Testing...
[Epoch 13 Batch 30/173] avg loss 0.00305368, throughput 3.96531K wps
[Epoch 13 Batch 60/173] avg loss 0.00303868, throughput 3.90578K wps
[Epoch 13 Batch 90/173] avg loss 0.0028036, throughput 3.87627K wps
[Epoch 13 Batch 120/173] avg loss 0.00307295, throughput 3.8709K wps
[Epoch 13 Batch 150/173] avg loss 0.00283231, throughput 3.86502K wps
Begin Testing...
[Epoch 13] train avg loss 0.00296814, test acc 0.8094, test avg loss 0.412399, throughput 3.8921K wps
Observed Improvement.
Begin Testing...
[Epoch 14 Batch 30/173] avg loss 0.00257497, throughput 3.96526K wps
[Epoch 14 Batch 60/173] avg loss 0.00248392, throughput 3.88523K wps
[Epoch 14 Batch 90/173] avg loss 0.00246357, throughput 3.87324K wps
[Epoch 14 Batch 120/173] avg loss 0.00259688, throughput 3.86809K wps
[Epoch 14 Batch 150/173] avg loss 0.00249913, throughput 3.86376K wps
Begin Testing...
[Epoch 14] train avg loss 0.00254834, test acc 0.8115, test avg loss 0.419124, throughput 3.88688K wps
Observed Improvement.
Begin Testing...
[Epoch 15 Batch 30/173] avg loss 0.00217891, throughput 3.96685K wps
[Epoch 15 Batch 60/173] avg loss 0.002125, throughput 3.90235K wps
[Epoch 15 Batch 90/173] avg loss 0.00218974, throughput 3.88787K wps
[Epoch 15 Batch 120/173] avg loss 0.00223538, throughput 3.8903K wps
[Epoch 15 Batch 150/173] avg loss 0.00216046, throughput 3.88741K wps
Begin Testing...
[Epoch 15] train avg loss 0.00216485, test acc 0.7990, test avg loss 0.433817, throughput 3.90663K wps
[Epoch 16 Batch 30/173] avg loss 0.00171628, throughput 3.98959K wps
[Epoch 16 Batch 60/173] avg loss 0.00166839, throughput 3.88324K wps
[Epoch 16 Batch 90/173] avg loss 0.00186804, throughput 3.85582K wps
[Epoch 16 Batch 120/173] avg loss 0.00193457, throughput 3.87212K wps
[Epoch 16 Batch 150/173] avg loss 0.00193394, throughput 3.87825K wps
Begin Testing...
[Epoch 16] train avg loss 0.00182067, test acc 0.7969, test avg loss 0.446083, throughput 3.89455K wps
[Epoch 17 Batch 30/173] avg loss 0.00148126, throughput 3.96006K wps
[Epoch 17 Batch 60/173] avg loss 0.00129753, throughput 3.87967K wps
[Epoch 17 Batch 90/173] avg loss 0.00167183, throughput 3.86424K wps
[Epoch 17 Batch 120/173] avg loss 0.00156078, throughput 3.86556K wps
[Epoch 17 Batch 150/173] avg loss 0.00145313, throughput 3.87776K wps
Begin Testing...
[Epoch 17] train avg loss 0.00152664, test acc 0.7958, test avg loss 0.460177, throughput 3.88951K wps
[Epoch 18 Batch 30/173] avg loss 0.00142537, throughput 3.97979K wps
[Epoch 18 Batch 60/173] avg loss 0.00114218, throughput 3.88369K wps
[Epoch 18 Batch 90/173] avg loss 0.00124724, throughput 3.88452K wps
[Epoch 18 Batch 120/173] avg loss 0.00130518, throughput 3.88868K wps
[Epoch 18 Batch 150/173] avg loss 0.00129896, throughput 3.89117K wps
Begin Testing...
[Epoch 18] train avg loss 0.00129914, test acc 0.7990, test avg loss 0.466858, throughput 3.90178K wps
[Epoch 19 Batch 30/173] avg loss 0.00102731, throughput 3.98106K wps
[Epoch 19 Batch 60/173] avg loss 0.00115295, throughput 3.87091K wps
[Epoch 19 Batch 90/173] avg loss 0.00111652, throughput 3.86829K wps
[Epoch 19 Batch 120/173] avg loss 0.00101793, throughput 3.87464K wps
[Epoch 19 Batch 150/173] avg loss 0.00113699, throughput 3.86783K wps
Begin Testing...
[Epoch 19] train avg loss 0.00109274, test acc 0.7990, test avg loss 0.479851, throughput 3.89095K wps
[Epoch 20 Batch 30/173] avg loss 0.000988456, throughput 3.95799K wps
[Epoch 20 Batch 60/173] avg loss 0.00092189, throughput 3.85793K wps
[Epoch 20 Batch 90/173] avg loss 0.000993188, throughput 3.88205K wps
[Epoch 20 Batch 120/173] avg loss 0.000867005, throughput 3.91274K wps
[Epoch 20 Batch 150/173] avg loss 0.000895414, throughput 3.89K wps
Begin Testing...
[Epoch 20] train avg loss 0.00093499, test acc 0.7948, test avg loss 0.500286, throughput 3.90036K wps
[Epoch 21 Batch 30/173] avg loss 0.000756057, throughput 3.97934K wps
[Epoch 21 Batch 60/173] avg loss 0.000653019, throughput 3.89691K wps
[Epoch 21 Batch 90/173] avg loss 0.000872391, throughput 3.88128K wps
[Epoch 21 Batch 120/173] avg loss 0.000856111, throughput 3.8753K wps
[Epoch 21 Batch 150/173] avg loss 0.000814861, throughput 3.87674K wps
Begin Testing...
[Epoch 21] train avg loss 0.000801229, test acc 0.7917, test avg loss 0.516545, throughput 3.89713K wps
[Epoch 22 Batch 30/173] avg loss 0.000689473, throughput 3.97861K wps
[Epoch 22 Batch 60/173] avg loss 0.000657918, throughput 3.87526K wps
[Epoch 22 Batch 90/173] avg loss 0.000661073, throughput 3.87702K wps
[Epoch 22 Batch 120/173] avg loss 0.000715688, throughput 3.89415K wps
[Epoch 22 Batch 150/173] avg loss 0.000644509, throughput 3.89027K wps
Begin Testing...
[Epoch 22] train avg loss 0.000668764, test acc 0.7948, test avg loss 0.526053, throughput 3.89813K wps
[Epoch 23 Batch 30/173] avg loss 0.000577207, throughput 3.98694K wps
[Epoch 23 Batch 60/173] avg loss 0.000555505, throughput 3.89299K wps
[Epoch 23 Batch 90/173] avg loss 0.000591268, throughput 3.88949K wps
[Epoch 23 Batch 120/173] avg loss 0.000565045, throughput 3.91777K wps
[Epoch 23 Batch 150/173] avg loss 0.000645612, throughput 3.88314K wps
Begin Testing...
[Epoch 23] train avg loss 0.00059129, test acc 0.7948, test avg loss 0.541645, throughput 3.90887K wps
[Epoch 24 Batch 30/173] avg loss 0.00043725, throughput 3.96396K wps
[Epoch 24 Batch 60/173] avg loss 0.00048394, throughput 3.87836K wps
[Epoch 24 Batch 90/173] avg loss 0.000499295, throughput 3.87195K wps
[Epoch 24 Batch 120/173] avg loss 0.00040524, throughput 3.87325K wps
[Epoch 24 Batch 150/173] avg loss 0.000536737, throughput 3.88115K wps
Begin Testing...
[Epoch 24] train avg loss 0.000488442, test acc 0.7865, test avg loss 0.567022, throughput 3.89309K wps
[Epoch 25 Batch 30/173] avg loss 0.000426813, throughput 3.96489K wps
[Epoch 25 Batch 60/173] avg loss 0.000383324, throughput 3.8659K wps
[Epoch 25 Batch 90/173] avg loss 0.00042928, throughput 3.86848K wps
[Epoch 25 Batch 120/173] avg loss 0.000380793, throughput 3.87632K wps
[Epoch 25 Batch 150/173] avg loss 0.000487292, throughput 3.87973K wps
Begin Testing...
[Epoch 25] train avg loss 0.000419297, test acc 0.7979, test avg loss 0.570636, throughput 3.89006K wps
[Epoch 26 Batch 30/173] avg loss 0.000400992, throughput 3.9752K wps
[Epoch 26 Batch 60/173] avg loss 0.000367596, throughput 3.87464K wps
[Epoch 26 Batch 90/173] avg loss 0.000380092, throughput 3.88937K wps
[Epoch 26 Batch 120/173] avg loss 0.000344904, throughput 3.89505K wps
[Epoch 26 Batch 150/173] avg loss 0.000346903, throughput 3.86778K wps
Begin Testing...
[Epoch 26] train avg loss 0.000372417, test acc 0.7906, test avg loss 0.589648, throughput 3.89656K wps
[Epoch 27 Batch 30/173] avg loss 0.000283326, throughput 3.96426K wps
[Epoch 27 Batch 60/173] avg loss 0.000304899, throughput 3.86778K wps
[Epoch 27 Batch 90/173] avg loss 0.000302422, throughput 3.87809K wps
[Epoch 27 Batch 120/173] avg loss 0.000335175, throughput 3.87275K wps
[Epoch 27 Batch 150/173] avg loss 0.000336645, throughput 3.87427K wps
Begin Testing...
[Epoch 27] train avg loss 0.000315786, test acc 0.7906, test avg loss 0.611071, throughput 3.89215K wps
[Epoch 28 Batch 30/173] avg loss 0.000283345, throughput 3.96672K wps
[Epoch 28 Batch 60/173] avg loss 0.000247841, throughput 3.87894K wps
[Epoch 28 Batch 90/173] avg loss 0.000288229, throughput 3.88424K wps
[Epoch 28 Batch 120/173] avg loss 0.000309653, throughput 3.88531K wps
[Epoch 28 Batch 150/173] avg loss 0.000281924, throughput 3.88642K wps
Begin Testing...
[Epoch 28] train avg loss 0.000280981, test acc 0.7865, test avg loss 0.628194, throughput 3.90262K wps
[Epoch 29 Batch 30/173] avg loss 0.000247311, throughput 4.00015K wps
[Epoch 29 Batch 60/173] avg loss 0.000214413, throughput 3.89101K wps
[Epoch 29 Batch 90/173] avg loss 0.000274204, throughput 3.88052K wps
[Epoch 29 Batch 120/173] avg loss 0.000255725, throughput 3.8694K wps
[Epoch 29 Batch 150/173] avg loss 0.0002428, throughput 3.88609K wps
Begin Testing...
[Epoch 29] train avg loss 0.000250361, test acc 0.7885, test avg loss 0.644462, throughput 3.90284K wps
[Epoch 30 Batch 30/173] avg loss 0.000222027, throughput 3.97044K wps
[Epoch 30 Batch 60/173] avg loss 0.000219926, throughput 3.88074K wps
[Epoch 30 Batch 90/173] avg loss 0.000175693, throughput 3.88684K wps
[Epoch 30 Batch 120/173] avg loss 0.000193374, throughput 3.88396K wps
[Epoch 30 Batch 150/173] avg loss 0.000197376, throughput 3.86638K wps
Begin Testing...
[Epoch 30] train avg loss 0.000201768, test acc 0.7885, test avg loss 0.65706, throughput 3.89284K wps
[Epoch 31 Batch 30/173] avg loss 0.000182517, throughput 3.97104K wps
[Epoch 31 Batch 60/173] avg loss 0.000166599, throughput 3.89104K wps
[Epoch 31 Batch 90/173] avg loss 0.0001721, throughput 3.88729K wps
[Epoch 31 Batch 120/173] avg loss 0.00018636, throughput 3.90726K wps
[Epoch 31 Batch 150/173] avg loss 0.00018003, throughput 3.8898K wps
Begin Testing...
[Epoch 31] train avg loss 0.000181794, test acc 0.7885, test avg loss 0.670693, throughput 3.90442K wps
[Epoch 32 Batch 30/173] avg loss 0.000159499, throughput 3.98283K wps
[Epoch 32 Batch 60/173] avg loss 0.000169991, throughput 3.89051K wps
[Epoch 32 Batch 90/173] avg loss 0.00015906, throughput 3.88127K wps
[Epoch 32 Batch 120/173] avg loss 0.000177449, throughput 3.88119K wps
[Epoch 32 Batch 150/173] avg loss 0.000153302, throughput 3.88195K wps
Begin Testing...
[Epoch 32] train avg loss 0.000162234, test acc 0.7906, test avg loss 0.690676, throughput 3.89934K wps
[Epoch 33 Batch 30/173] avg loss 0.000167145, throughput 3.98381K wps
[Epoch 33 Batch 60/173] avg loss 0.000143053, throughput 3.87328K wps
[Epoch 33 Batch 90/173] avg loss 0.000131489, throughput 3.89997K wps
[Epoch 33 Batch 120/173] avg loss 0.000149087, throughput 3.89719K wps
[Epoch 33 Batch 150/173] avg loss 0.000145129, throughput 3.88343K wps
Begin Testing...
[Epoch 33] train avg loss 0.000148822, test acc 0.7844, test avg loss 0.713358, throughput 3.90327K wps
[Epoch 34 Batch 30/173] avg loss 0.000124411, throughput 3.97574K wps
[Epoch 34 Batch 60/173] avg loss 0.00013383, throughput 3.89315K wps
[Epoch 34 Batch 90/173] avg loss 0.00011659, throughput 3.90534K wps
[Epoch 34 Batch 120/173] avg loss 0.000108933, throughput 3.87342K wps
[Epoch 34 Batch 150/173] avg loss 0.000118636, throughput 3.88242K wps
Begin Testing...
[Epoch 34] train avg loss 0.000122688, test acc 0.7812, test avg loss 0.736599, throughput 3.90731K wps
[Epoch 35 Batch 30/173] avg loss 0.00011108, throughput 3.98435K wps
[Epoch 35 Batch 60/173] avg loss 9.44763e-05, throughput 3.88951K wps
[Epoch 35 Batch 90/173] avg loss 0.000117617, throughput 3.88837K wps
[Epoch 35 Batch 120/173] avg loss 0.000112068, throughput 3.89417K wps
[Epoch 35 Batch 150/173] avg loss 0.000115525, throughput 3.90456K wps
Begin Testing...
[Epoch 35] train avg loss 0.000114213, test acc 0.7865, test avg loss 0.740635, throughput 3.90913K wps
[Epoch 36 Batch 30/173] avg loss 9.41659e-05, throughput 3.97191K wps
[Epoch 36 Batch 60/173] avg loss 8.17407e-05, throughput 3.8822K wps
[Epoch 36 Batch 90/173] avg loss 0.000103287, throughput 3.88644K wps
[Epoch 36 Batch 120/173] avg loss 8.01273e-05, throughput 3.87549K wps
[Epoch 36 Batch 150/173] avg loss 8.58116e-05, throughput 3.90564K wps
Begin Testing...
[Epoch 36] train avg loss 9.01145e-05, test acc 0.7906, test avg loss 0.758668, throughput 3.89985K wps
[Epoch 37 Batch 30/173] avg loss 8.50956e-05, throughput 3.9816K wps
[Epoch 37 Batch 60/173] avg loss 7.49953e-05, throughput 3.86454K wps
[Epoch 37 Batch 90/173] avg loss 9.90406e-05, throughput 3.87688K wps
[Epoch 37 Batch 120/173] avg loss 8.75521e-05, throughput 3.89051K wps
[Epoch 37 Batch 150/173] avg loss 8.0548e-05, throughput 3.91245K wps
Begin Testing...
[Epoch 37] train avg loss 8.49554e-05, test acc 0.7875, test avg loss 0.771303, throughput 3.90239K wps
[Epoch 38 Batch 30/173] avg loss 7.08365e-05, throughput 3.95881K wps
[Epoch 38 Batch 60/173] avg loss 9.27405e-05, throughput 3.87073K wps
[Epoch 38 Batch 90/173] avg loss 8.43741e-05, throughput 3.86329K wps
[Epoch 38 Batch 120/173] avg loss 7.4882e-05, throughput 3.86652K wps
[Epoch 38 Batch 150/173] avg loss 7.58617e-05, throughput 3.88548K wps
Begin Testing...
[Epoch 38] train avg loss 8.2691e-05, test acc 0.7833, test avg loss 0.803583, throughput 3.89126K wps
[Epoch 39 Batch 30/173] avg loss 9.15336e-05, throughput 3.99495K wps
[Epoch 39 Batch 60/173] avg loss 6.62011e-05, throughput 3.88638K wps
[Epoch 39 Batch 90/173] avg loss 7.11751e-05, throughput 3.87084K wps
[Epoch 39 Batch 120/173] avg loss 6.6275e-05, throughput 3.87596K wps
[Epoch 39 Batch 150/173] avg loss 6.01585e-05, throughput 3.85513K wps
Begin Testing...
[Epoch 39] train avg loss 7.31018e-05, test acc 0.7865, test avg loss 0.801491, throughput 3.89619K wps
[Epoch 40 Batch 30/173] avg loss 6.67782e-05, throughput 3.98019K wps
[Epoch 40 Batch 60/173] avg loss 6.87714e-05, throughput 3.90334K wps
[Epoch 40 Batch 90/173] avg loss 5.35052e-05, throughput 3.88886K wps
[Epoch 40 Batch 120/173] avg loss 5.93476e-05, throughput 3.87053K wps
[Epoch 40 Batch 150/173] avg loss 5.78271e-05, throughput 3.87263K wps
Begin Testing...
[Epoch 40] train avg loss 6.22007e-05, test acc 0.7865, test avg loss 0.824859, throughput 3.8984K wps
[Epoch 41 Batch 30/173] avg loss 5.287e-05, throughput 3.9856K wps
[Epoch 41 Batch 60/173] avg loss 4.69228e-05, throughput 3.88272K wps
[Epoch 41 Batch 90/173] avg loss 4.95147e-05, throughput 3.92043K wps
[Epoch 41 Batch 120/173] avg loss 5.69215e-05, throughput 3.88219K wps
[Epoch 41 Batch 150/173] avg loss 6.3794e-05, throughput 3.87932K wps
Begin Testing...
[Epoch 41] train avg loss 5.49139e-05, test acc 0.7833, test avg loss 0.830188, throughput 3.90748K wps
[Epoch 42 Batch 30/173] avg loss 7.45299e-05, throughput 3.9623K wps
[Epoch 42 Batch 60/173] avg loss 6.00014e-05, throughput 3.90998K wps
[Epoch 42 Batch 90/173] avg loss 4.8854e-05, throughput 3.89605K wps
[Epoch 42 Batch 120/173] avg loss 5.52391e-05, throughput 3.86893K wps
[Epoch 42 Batch 150/173] avg loss 5.49415e-05, throughput 3.88668K wps
Begin Testing...
[Epoch 42] train avg loss 6.0275e-05, test acc 0.7896, test avg loss 0.847045, throughput 3.90381K wps
[Epoch 43 Batch 30/173] avg loss 3.90417e-05, throughput 3.97177K wps
[Epoch 43 Batch 60/173] avg loss 5.53199e-05, throughput 3.90884K wps
[Epoch 43 Batch 90/173] avg loss 3.60206e-05, throughput 3.89259K wps
[Epoch 43 Batch 120/173] avg loss 3.89478e-05, throughput 3.8868K wps
[Epoch 43 Batch 150/173] avg loss 5.61933e-05, throughput 3.89761K wps
Begin Testing...
[Epoch 43] train avg loss 4.38523e-05, test acc 0.7844, test avg loss 0.870418, throughput 3.90866K wps
[Epoch 44 Batch 30/173] avg loss 3.33908e-05, throughput 3.966K wps
[Epoch 44 Batch 60/173] avg loss 3.44092e-05, throughput 3.88268K wps
[Epoch 44 Batch 90/173] avg loss 4.11544e-05, throughput 3.87179K wps
[Epoch 44 Batch 120/173] avg loss 4.17744e-05, throughput 3.87313K wps
[Epoch 44 Batch 150/173] avg loss 4.48751e-05, throughput 3.8892K wps
Begin Testing...
[Epoch 44] train avg loss 4.00393e-05, test acc 0.7875, test avg loss 0.879967, throughput 3.89798K wps
[Epoch 45 Batch 30/173] avg loss 3.51204e-05, throughput 3.96611K wps
[Epoch 45 Batch 60/173] avg loss 3.8195e-05, throughput 3.87765K wps
[Epoch 45 Batch 90/173] avg loss 3.21215e-05, throughput 3.86937K wps
[Epoch 45 Batch 120/173] avg loss 4.28473e-05, throughput 3.88559K wps
[Epoch 45 Batch 150/173] avg loss 3.4883e-05, throughput 3.85435K wps
Begin Testing...
[Epoch 45] train avg loss 3.76385e-05, test acc 0.7865, test avg loss 0.895417, throughput 3.88643K wps
[Epoch 46 Batch 30/173] avg loss 3.77068e-05, throughput 3.98539K wps
[Epoch 46 Batch 60/173] avg loss 2.99953e-05, throughput 3.88787K wps
[Epoch 46 Batch 90/173] avg loss 2.56105e-05, throughput 3.89075K wps
[Epoch 46 Batch 120/173] avg loss 3.13789e-05, throughput 3.91176K wps
[Epoch 46 Batch 150/173] avg loss 3.82291e-05, throughput 3.86867K wps
Begin Testing...
[Epoch 46] train avg loss 3.54278e-05, test acc 0.7885, test avg loss 0.906933, throughput 3.90337K wps
[Epoch 47 Batch 30/173] avg loss 3.03594e-05, throughput 3.95988K wps
[Epoch 47 Batch 60/173] avg loss 3.50058e-05, throughput 3.87879K wps
[Epoch 47 Batch 90/173] avg loss 2.96941e-05, throughput 3.85785K wps
[Epoch 47 Batch 120/173] avg loss 3.56157e-05, throughput 3.88413K wps
[Epoch 47 Batch 150/173] avg loss 3.01668e-05, throughput 3.90021K wps
Begin Testing...
[Epoch 47] train avg loss 3.44189e-05, test acc 0.7833, test avg loss 0.948268, throughput 3.89646K wps
[Epoch 48 Batch 30/173] avg loss 3.53915e-05, throughput 3.98938K wps
[Epoch 48 Batch 60/173] avg loss 3.14191e-05, throughput 3.88061K wps
[Epoch 48 Batch 90/173] avg loss 2.02191e-05, throughput 3.886K wps
[Epoch 48 Batch 120/173] avg loss 2.19516e-05, throughput 3.89489K wps
[Epoch 48 Batch 150/173] avg loss 2.81687e-05, throughput 3.88679K wps
Begin Testing...
[Epoch 48] train avg loss 2.69779e-05, test acc 0.7833, test avg loss 0.93947, throughput 3.903K wps
[Epoch 49 Batch 30/173] avg loss 1.99931e-05, throughput 3.97877K wps
[Epoch 49 Batch 60/173] avg loss 2.90933e-05, throughput 3.87553K wps
[Epoch 49 Batch 90/173] avg loss 1.92278e-05, throughput 3.86785K wps
[Epoch 49 Batch 120/173] avg loss 2.29519e-05, throughput 3.8962K wps
[Epoch 49 Batch 150/173] avg loss 2.27465e-05, throughput 3.90572K wps
Begin Testing...
[Epoch 49] train avg loss 2.30013e-05, test acc 0.7854, test avg loss 0.953409, throughput 3.90133K wps
[Epoch 50 Batch 30/173] avg loss 1.9676e-05, throughput 3.98285K wps
[Epoch 50 Batch 60/173] avg loss 2.14944e-05, throughput 3.88327K wps
[Epoch 50 Batch 90/173] avg loss 2.13205e-05, throughput 3.88417K wps
[Epoch 50 Batch 120/173] avg loss 2.5248e-05, throughput 3.88132K wps
[Epoch 50 Batch 150/173] avg loss 1.72525e-05, throughput 3.87035K wps
Begin Testing...
[Epoch 50] train avg loss 2.03782e-05, test acc 0.7875, test avg loss 0.972837, throughput 3.89594K wps
[Epoch 51 Batch 30/173] avg loss 1.89471e-05, throughput 3.98323K wps
[Epoch 51 Batch 60/173] avg loss 2.30638e-05, throughput 3.86984K wps
[Epoch 51 Batch 90/173] avg loss 4.41154e-05, throughput 3.87098K wps
[Epoch 51 Batch 120/173] avg loss 3.48487e-05, throughput 3.86239K wps
[Epoch 51 Batch 150/173] avg loss 4.44915e-05, throughput 3.86652K wps
Begin Testing...
[Epoch 51] train avg loss 3.25135e-05, test acc 0.7854, test avg loss 0.984829, throughput 3.89225K wps
[Epoch 52 Batch 30/173] avg loss 3.25259e-05, throughput 3.97647K wps
[Epoch 52 Batch 60/173] avg loss 2.35451e-05, throughput 3.91018K wps
[Epoch 52 Batch 90/173] avg loss 2.21358e-05, throughput 3.88813K wps
[Epoch 52 Batch 120/173] avg loss 2.24686e-05, throughput 3.88325K wps
[Epoch 52 Batch 150/173] avg loss 3.72894e-05, throughput 3.88494K wps
Begin Testing...
[Epoch 52] train avg loss 2.6474e-05, test acc 0.7823, test avg loss 1.00402, throughput 3.90619K wps
[Epoch 53 Batch 30/173] avg loss 1.80867e-05, throughput 3.98387K wps
[Epoch 53 Batch 60/173] avg loss 2.20094e-05, throughput 3.87407K wps
[Epoch 53 Batch 90/173] avg loss 1.71961e-05, throughput 3.88016K wps
[Epoch 53 Batch 120/173] avg loss 1.60087e-05, throughput 3.86417K wps
[Epoch 53 Batch 150/173] avg loss 2.31384e-05, throughput 3.87114K wps
Begin Testing...
[Epoch 53] train avg loss 1.86217e-05, test acc 0.7865, test avg loss 1.02614, throughput 3.89411K wps
[Epoch 54 Batch 30/173] avg loss 1.67961e-05, throughput 3.98001K wps
[Epoch 54 Batch 60/173] avg loss 1.94191e-05, throughput 3.90226K wps
[Epoch 54 Batch 90/173] avg loss 1.48718e-05, throughput 3.8856K wps
[Epoch 54 Batch 120/173] avg loss 1.55233e-05, throughput 3.86902K wps
[Epoch 54 Batch 150/173] avg loss 1.27956e-05, throughput 3.87233K wps
Begin Testing...
[Epoch 54] train avg loss 1.58943e-05, test acc 0.7875, test avg loss 1.03992, throughput 3.89512K wps
[Epoch 55 Batch 30/173] avg loss 1.33522e-05, throughput 3.97406K wps
[Epoch 55 Batch 60/173] avg loss 1.30724e-05, throughput 3.87378K wps
[Epoch 55 Batch 90/173] avg loss 1.3639e-05, throughput 3.86962K wps
[Epoch 55 Batch 120/173] avg loss 1.05312e-05, throughput 3.87218K wps
[Epoch 55 Batch 150/173] avg loss 1.04708e-05, throughput 3.8547K wps
Begin Testing...
[Epoch 55] train avg loss 1.26912e-05, test acc 0.7875, test avg loss 1.05392, throughput 3.88744K wps
[Epoch 56 Batch 30/173] avg loss 1.35449e-05, throughput 3.96939K wps
[Epoch 56 Batch 60/173] avg loss 1.04234e-05, throughput 3.89028K wps
[Epoch 56 Batch 90/173] avg loss 1.59989e-05, throughput 3.91179K wps
[Epoch 56 Batch 120/173] avg loss 9.95354e-06, throughput 3.89127K wps
[Epoch 56 Batch 150/173] avg loss 1.51705e-05, throughput 3.88445K wps
Begin Testing...
[Epoch 56] train avg loss 1.29372e-05, test acc 0.7875, test avg loss 1.06843, throughput 3.90349K wps
[Epoch 57 Batch 30/173] avg loss 1.30978e-05, throughput 3.9719K wps
[Epoch 57 Batch 60/173] avg loss 1.78952e-05, throughput 3.87734K wps
[Epoch 57 Batch 90/173] avg loss 1.25624e-05, throughput 3.87574K wps
[Epoch 57 Batch 120/173] avg loss 1.45463e-05, throughput 3.89399K wps
[Epoch 57 Batch 150/173] avg loss 9.67533e-06, throughput 3.89339K wps
Begin Testing...
[Epoch 57] train avg loss 1.30819e-05, test acc 0.7823, test avg loss 1.09767, throughput 3.90115K wps
[Epoch 58 Batch 30/173] avg loss 2.00985e-05, throughput 3.9759K wps
[Epoch 58 Batch 60/173] avg loss 9.34679e-06, throughput 3.88314K wps
[Epoch 58 Batch 90/173] avg loss 1.07447e-05, throughput 3.9069K wps
[Epoch 58 Batch 120/173] avg loss 1.65186e-05, throughput 3.87591K wps
[Epoch 58 Batch 150/173] avg loss 9.79297e-06, throughput 3.87282K wps
Begin Testing...
[Epoch 58] train avg loss 1.27078e-05, test acc 0.7844, test avg loss 1.09152, throughput 3.90207K wps
[Epoch 59 Batch 30/173] avg loss 9.61479e-06, throughput 3.98236K wps
[Epoch 59 Batch 60/173] avg loss 1.3151e-05, throughput 3.90636K wps
[Epoch 59 Batch 90/173] avg loss 1.02309e-05, throughput 3.88796K wps
[Epoch 59 Batch 120/173] avg loss 7.96248e-06, throughput 3.87331K wps
[Epoch 59 Batch 150/173] avg loss 1.13054e-05, throughput 3.86878K wps
Begin Testing...
[Epoch 59] train avg loss 1.19782e-05, test acc 0.7792, test avg loss 1.10977, throughput 3.90118K wps
Test loss 0.453364, test acc 0.7974
Total time cost 555.72s
9596 1066
[Epoch 0 Batch 30/173] avg loss 0.0151875, throughput 3.66771K wps
[Epoch 0 Batch 60/173] avg loss 0.0149634, throughput 3.88887K wps
[Epoch 0 Batch 90/173] avg loss 0.0143521, throughput 3.90402K wps
[Epoch 0 Batch 120/173] avg loss 0.0141206, throughput 3.88506K wps
[Epoch 0 Batch 150/173] avg loss 0.0141325, throughput 3.87174K wps
Begin Testing...
[Epoch 0] train avg loss 0.0144289, test acc 0.6125, test avg loss 0.661647, throughput 3.84595K wps
Observed Improvement.
Begin Testing...
[Epoch 1 Batch 30/173] avg loss 0.0132986, throughput 3.95332K wps
[Epoch 1 Batch 60/173] avg loss 0.0132822, throughput 3.88121K wps
[Epoch 1 Batch 90/173] avg loss 0.0130659, throughput 3.88803K wps
[Epoch 1 Batch 120/173] avg loss 0.0131227, throughput 3.88941K wps
[Epoch 1 Batch 150/173] avg loss 0.0129303, throughput 3.9111K wps
Begin Testing...
[Epoch 1] train avg loss 0.0131274, test acc 0.6448, test avg loss 0.630661, throughput 3.89974K wps
Observed Improvement.
Begin Testing...
[Epoch 2 Batch 30/173] avg loss 0.0124491, throughput 3.96588K wps
[Epoch 2 Batch 60/173] avg loss 0.0123832, throughput 3.90569K wps
[Epoch 2 Batch 90/173] avg loss 0.0122514, throughput 3.88871K wps
[Epoch 2 Batch 120/173] avg loss 0.0119512, throughput 3.8824K wps
[Epoch 2 Batch 150/173] avg loss 0.0119672, throughput 3.86795K wps
Begin Testing...
[Epoch 2] train avg loss 0.0122034, test acc 0.6885, test avg loss 0.598003, throughput 3.90107K wps
Observed Improvement.
Begin Testing...
[Epoch 3 Batch 30/173] avg loss 0.0113876, throughput 3.95535K wps
[Epoch 3 Batch 60/173] avg loss 0.0115691, throughput 3.88548K wps
[Epoch 3 Batch 90/173] avg loss 0.0114486, throughput 3.89923K wps
[Epoch 3 Batch 120/173] avg loss 0.0112501, throughput 3.9025K wps
[Epoch 3 Batch 150/173] avg loss 0.0112924, throughput 3.87992K wps
Begin Testing...
[Epoch 3] train avg loss 0.0113058, test acc 0.7302, test avg loss 0.560705, throughput 3.89959K wps
Observed Improvement.
Begin Testing...
[Epoch 4 Batch 30/173] avg loss 0.0103669, throughput 3.99512K wps
[Epoch 4 Batch 60/173] avg loss 0.0102138, throughput 3.86521K wps
[Epoch 4 Batch 90/173] avg loss 0.0102794, throughput 3.88404K wps
[Epoch 4 Batch 120/173] avg loss 0.00994669, throughput 3.89035K wps
[Epoch 4 Batch 150/173] avg loss 0.0101716, throughput 3.90644K wps
Begin Testing...
[Epoch 4] train avg loss 0.0101689, test acc 0.7281, test avg loss 0.534076, throughput 3.90524K wps
[Epoch 5 Batch 30/173] avg loss 0.00948689, throughput 3.97905K wps
[Epoch 5 Batch 60/173] avg loss 0.00929547, throughput 3.8788K wps
[Epoch 5 Batch 90/173] avg loss 0.00911989, throughput 3.8722K wps
[Epoch 5 Batch 120/173] avg loss 0.0090993, throughput 3.87182K wps
[Epoch 5 Batch 150/173] avg loss 0.00894249, throughput 3.87969K wps
Begin Testing...
[Epoch 5] train avg loss 0.00915015, test acc 0.7885, test avg loss 0.48048, throughput 3.89495K wps
Observed Improvement.
Begin Testing...
[Epoch 6 Batch 30/173] avg loss 0.00825305, throughput 3.98035K wps
[Epoch 6 Batch 60/173] avg loss 0.00814128, throughput 3.89623K wps
[Epoch 6 Batch 90/173] avg loss 0.0080898, throughput 3.88001K wps
[Epoch 6 Batch 120/173] avg loss 0.00795616, throughput 3.87271K wps
[Epoch 6 Batch 150/173] avg loss 0.00801976, throughput 3.87554K wps
Begin Testing...
[Epoch 6] train avg loss 0.00803466, test acc 0.7979, test avg loss 0.451147, throughput 3.89856K wps
Observed Improvement.
Begin Testing...
[Epoch 7 Batch 30/173] avg loss 0.00730116, throughput 3.97006K wps
[Epoch 7 Batch 60/173] avg loss 0.00750444, throughput 3.87891K wps
[Epoch 7 Batch 90/173] avg loss 0.00719897, throughput 3.88678K wps
[Epoch 7 Batch 120/173] avg loss 0.00713951, throughput 3.92234K wps
[Epoch 7 Batch 150/173] avg loss 0.00711078, throughput 3.88549K wps
Begin Testing...
[Epoch 7] train avg loss 0.00719272, test acc 0.8042, test avg loss 0.430027, throughput 3.90389K wps
Observed Improvement.
Begin Testing...
[Epoch 8 Batch 30/173] avg loss 0.00608352, throughput 3.97453K wps
[Epoch 8 Batch 60/173] avg loss 0.0066472, throughput 3.87436K wps
[Epoch 8 Batch 90/173] avg loss 0.00606418, throughput 3.86807K wps
[Epoch 8 Batch 120/173] avg loss 0.00622558, throughput 3.88042K wps
[Epoch 8 Batch 150/173] avg loss 0.00619904, throughput 3.89215K wps
Begin Testing...
[Epoch 8] train avg loss 0.00624795, test acc 0.8208, test avg loss 0.41098, throughput 3.89543K wps
Observed Improvement.
Begin Testing...
[Epoch 9 Batch 30/173] avg loss 0.00534491, throughput 3.97663K wps
[Epoch 9 Batch 60/173] avg loss 0.00524431, throughput 3.89256K wps
[Epoch 9 Batch 90/173] avg loss 0.00529566, throughput 3.91457K wps
[Epoch 9 Batch 120/173] avg loss 0.00566576, throughput 3.88643K wps
[Epoch 9 Batch 150/173] avg loss 0.00550505, throughput 3.89364K wps
Begin Testing...
[Epoch 9] train avg loss 0.00540957, test acc 0.8146, test avg loss 0.406859, throughput 3.91301K wps
[Epoch 10 Batch 30/173] avg loss 0.00457661, throughput 3.97667K wps
[Epoch 10 Batch 60/173] avg loss 0.00495288, throughput 3.87368K wps
[Epoch 10 Batch 90/173] avg loss 0.00492979, throughput 3.8717K wps
[Epoch 10 Batch 120/173] avg loss 0.00461859, throughput 3.87557K wps
[Epoch 10 Batch 150/173] avg loss 0.00445378, throughput 3.87256K wps
Begin Testing...
[Epoch 10] train avg loss 0.00471624, test acc 0.8208, test avg loss 0.405675, throughput 3.89101K wps
Observed Improvement.
Begin Testing...
[Epoch 11 Batch 30/173] avg loss 0.00418061, throughput 3.97121K wps
[Epoch 11 Batch 60/173] avg loss 0.0041039, throughput 3.87876K wps
[Epoch 11 Batch 90/173] avg loss 0.00383058, throughput 3.88513K wps
[Epoch 11 Batch 120/173] avg loss 0.00419672, throughput 3.8811K wps
[Epoch 11 Batch 150/173] avg loss 0.00400432, throughput 3.91131K wps
Begin Testing...
[Epoch 11] train avg loss 0.00408727, test acc 0.8156, test avg loss 0.401816, throughput 3.90141K wps
[Epoch 12 Batch 30/173] avg loss 0.00362837, throughput 3.97995K wps
[Epoch 12 Batch 60/173] avg loss 0.00367074, throughput 3.86904K wps
[Epoch 12 Batch 90/173] avg loss 0.00361756, throughput 3.86792K wps
[Epoch 12 Batch 120/173] avg loss 0.00331144, throughput 3.88141K wps
[Epoch 12 Batch 150/173] avg loss 0.00325867, throughput 3.88003K wps
Begin Testing...
[Epoch 12] train avg loss 0.00344779, test acc 0.8104, test avg loss 0.408191, throughput 3.89107K wps
[Epoch 13 Batch 30/173] avg loss 0.00285552, throughput 3.96561K wps
[Epoch 13 Batch 60/173] avg loss 0.00295952, throughput 3.87975K wps
[Epoch 13 Batch 90/173] avg loss 0.00301832, throughput 3.89434K wps
[Epoch 13 Batch 120/173] avg loss 0.0027807, throughput 3.92382K wps
[Epoch 13 Batch 150/173] avg loss 0.00283666, throughput 3.88708K wps
Begin Testing...
[Epoch 13] train avg loss 0.00286964, test acc 0.8094, test avg loss 0.419803, throughput 3.90884K wps
[Epoch 14 Batch 30/173] avg loss 0.00236663, throughput 3.9657K wps
[Epoch 14 Batch 60/173] avg loss 0.00227388, throughput 3.88036K wps
[Epoch 14 Batch 90/173] avg loss 0.00236139, throughput 3.87062K wps
[Epoch 14 Batch 120/173] avg loss 0.00226953, throughput 3.89397K wps
[Epoch 14 Batch 150/173] avg loss 0.00235289, throughput 3.87921K wps
Begin Testing...
[Epoch 14] train avg loss 0.00238236, test acc 0.8135, test avg loss 0.428497, throughput 3.8948K wps
[Epoch 15 Batch 30/173] avg loss 0.00202838, throughput 3.97119K wps
[Epoch 15 Batch 60/173] avg loss 0.0019924, throughput 3.86949K wps
[Epoch 15 Batch 90/173] avg loss 0.00210565, throughput 3.88159K wps
[Epoch 15 Batch 120/173] avg loss 0.00202722, throughput 3.89166K wps
[Epoch 15 Batch 150/173] avg loss 0.00215161, throughput 3.90704K wps
Begin Testing...
[Epoch 15] train avg loss 0.00205076, test acc 0.8052, test avg loss 0.442637, throughput 3.90134K wps
[Epoch 16 Batch 30/173] avg loss 0.00179794, throughput 3.96287K wps
[Epoch 16 Batch 60/173] avg loss 0.00165675, throughput 3.88914K wps
[Epoch 16 Batch 90/173] avg loss 0.00184066, throughput 3.88696K wps
[Epoch 16 Batch 120/173] avg loss 0.00176502, throughput 3.89306K wps
[Epoch 16 Batch 150/173] avg loss 0.00164848, throughput 3.87917K wps
Begin Testing...
[Epoch 16] train avg loss 0.00174179, test acc 0.8021, test avg loss 0.4564, throughput 3.89817K wps
[Epoch 17 Batch 30/173] avg loss 0.00134547, throughput 3.96603K wps
[Epoch 17 Batch 60/173] avg loss 0.00138563, throughput 3.88641K wps
[Epoch 17 Batch 90/173] avg loss 0.00126436, throughput 3.88172K wps
[Epoch 17 Batch 120/173] avg loss 0.00151675, throughput 3.91121K wps
[Epoch 17 Batch 150/173] avg loss 0.00160615, throughput 3.88507K wps
Begin Testing...
[Epoch 17] train avg loss 0.00143511, test acc 0.8010, test avg loss 0.474759, throughput 3.90093K wps
[Epoch 18 Batch 30/173] avg loss 0.00112887, throughput 3.97065K wps
[Epoch 18 Batch 60/173] avg loss 0.0012001, throughput 3.87427K wps
[Epoch 18 Batch 90/173] avg loss 0.00126198, throughput 3.87158K wps
[Epoch 18 Batch 120/173] avg loss 0.00111249, throughput 3.88643K wps
[Epoch 18 Batch 150/173] avg loss 0.00124771, throughput 3.91365K wps
Begin Testing...
[Epoch 18] train avg loss 0.00118323, test acc 0.8000, test avg loss 0.489976, throughput 3.89948K wps
[Epoch 19 Batch 30/173] avg loss 0.000953328, throughput 3.97507K wps
[Epoch 19 Batch 60/173] avg loss 0.00102288, throughput 3.86961K wps
[Epoch 19 Batch 90/173] avg loss 0.000998511, throughput 3.86718K wps
[Epoch 19 Batch 120/173] avg loss 0.00118161, throughput 3.891K wps
[Epoch 19 Batch 150/173] avg loss 0.000985493, throughput 3.91086K wps
Begin Testing...
[Epoch 19] train avg loss 0.0010174, test acc 0.8010, test avg loss 0.514948, throughput 3.90087K wps
[Epoch 20 Batch 30/173] avg loss 0.00078323, throughput 3.98865K wps
[Epoch 20 Batch 60/173] avg loss 0.000807275, throughput 3.87649K wps
[Epoch 20 Batch 90/173] avg loss 0.000797112, throughput 3.89268K wps
[Epoch 20 Batch 120/173] avg loss 0.000914266, throughput 3.89034K wps
[Epoch 20 Batch 150/173] avg loss 0.000884047, throughput 3.89125K wps
Begin Testing...
[Epoch 20] train avg loss 0.000844375, test acc 0.8021, test avg loss 0.536773, throughput 3.90628K wps
[Epoch 21 Batch 30/173] avg loss 0.000732806, throughput 3.98363K wps
[Epoch 21 Batch 60/173] avg loss 0.000693998, throughput 3.89796K wps
[Epoch 21 Batch 90/173] avg loss 0.000692956, throughput 3.89397K wps
[Epoch 21 Batch 120/173] avg loss 0.000736127, throughput 3.87298K wps
[Epoch 21 Batch 150/173] avg loss 0.000727648, throughput 3.89447K wps
Begin Testing...
[Epoch 21] train avg loss 0.000728787, test acc 0.7906, test avg loss 0.557397, throughput 3.91044K wps
[Epoch 22 Batch 30/173] avg loss 0.000633715, throughput 3.98062K wps
[Epoch 22 Batch 60/173] avg loss 0.000577873, throughput 3.88606K wps
[Epoch 22 Batch 90/173] avg loss 0.000560142, throughput 3.87821K wps
[Epoch 22 Batch 120/173] avg loss 0.000569576, throughput 3.89552K wps
[Epoch 22 Batch 150/173] avg loss 0.000567701, throughput 3.92166K wps
Begin Testing...
[Epoch 22] train avg loss 0.000592319, test acc 0.7969, test avg loss 0.581351, throughput 3.90746K wps
[Epoch 23 Batch 30/173] avg loss 0.000463547, throughput 3.98839K wps
[Epoch 23 Batch 60/173] avg loss 0.000539622, throughput 3.87613K wps
[Epoch 23 Batch 90/173] avg loss 0.000518207, throughput 3.88242K wps
[Epoch 23 Batch 120/173] avg loss 0.000537275, throughput 3.88325K wps
[Epoch 23 Batch 150/173] avg loss 0.000552029, throughput 3.88121K wps
Begin Testing...
[Epoch 23] train avg loss 0.000517992, test acc 0.7896, test avg loss 0.602917, throughput 3.89992K wps
[Epoch 24 Batch 30/173] avg loss 0.000409709, throughput 3.99634K wps
[Epoch 24 Batch 60/173] avg loss 0.00051213, throughput 3.88947K wps
[Epoch 24 Batch 90/173] avg loss 0.000352606, throughput 3.89601K wps
[Epoch 24 Batch 120/173] avg loss 0.000440318, throughput 3.87885K wps
[Epoch 24 Batch 150/173] avg loss 0.000485831, throughput 3.88127K wps
Begin Testing...
[Epoch 24] train avg loss 0.000441663, test acc 0.7948, test avg loss 0.620109, throughput 3.90278K wps
[Epoch 25 Batch 30/173] avg loss 0.000366798, throughput 3.97532K wps
[Epoch 25 Batch 60/173] avg loss 0.000352607, throughput 3.87717K wps
[Epoch 25 Batch 90/173] avg loss 0.000424788, throughput 3.89173K wps
[Epoch 25 Batch 120/173] avg loss 0.000407969, throughput 3.91352K wps
[Epoch 25 Batch 150/173] avg loss 0.000336764, throughput 3.88692K wps
Begin Testing...
[Epoch 25] train avg loss 0.000378577, test acc 0.7885, test avg loss 0.644242, throughput 3.90523K wps
[Epoch 26 Batch 30/173] avg loss 0.000284353, throughput 3.96718K wps
[Epoch 26 Batch 60/173] avg loss 0.000304998, throughput 3.87774K wps
[Epoch 26 Batch 90/173] avg loss 0.000329496, throughput 3.87754K wps
[Epoch 26 Batch 120/173] avg loss 0.000299774, throughput 3.88386K wps
[Epoch 26 Batch 150/173] avg loss 0.000370209, throughput 3.90251K wps
Begin Testing...
[Epoch 26] train avg loss 0.000325725, test acc 0.7875, test avg loss 0.666606, throughput 3.89824K wps
[Epoch 27 Batch 30/173] avg loss 0.000282502, throughput 3.97681K wps
[Epoch 27 Batch 60/173] avg loss 0.000321816, throughput 3.88661K wps
[Epoch 27 Batch 90/173] avg loss 0.000266528, throughput 3.89034K wps
[Epoch 27 Batch 120/173] avg loss 0.000260186, throughput 3.89435K wps
[Epoch 27 Batch 150/173] avg loss 0.000297989, throughput 3.87815K wps
Begin Testing...
[Epoch 27] train avg loss 0.000290123, test acc 0.7875, test avg loss 0.682519, throughput 3.90152K wps
[Epoch 28 Batch 30/173] avg loss 0.000260436, throughput 3.97471K wps
[Epoch 28 Batch 60/173] avg loss 0.000227222, throughput 3.86985K wps
[Epoch 28 Batch 90/173] avg loss 0.00023979, throughput 3.87663K wps
[Epoch 28 Batch 120/173] avg loss 0.000271486, throughput 3.89479K wps
[Epoch 28 Batch 150/173] avg loss 0.000234114, throughput 3.9033K wps
Begin Testing...
[Epoch 28] train avg loss 0.000252166, test acc 0.7823, test avg loss 0.70513, throughput 3.90105K wps
[Epoch 29 Batch 30/173] avg loss 0.000200929, throughput 3.97597K wps
[Epoch 29 Batch 60/173] avg loss 0.000184228, throughput 3.87933K wps
[Epoch 29 Batch 90/173] avg loss 0.000221556, throughput 3.87636K wps
[Epoch 29 Batch 120/173] avg loss 0.000231814, throughput 3.89436K wps
[Epoch 29 Batch 150/173] avg loss 0.000216766, throughput 3.88424K wps
Begin Testing...
[Epoch 29] train avg loss 0.000209182, test acc 0.7854, test avg loss 0.724747, throughput 3.89749K wps
[Epoch 30 Batch 30/173] avg loss 0.00018874, throughput 3.98042K wps
[Epoch 30 Batch 60/173] avg loss 0.00016458, throughput 3.8721K wps
[Epoch 30 Batch 90/173] avg loss 0.000140591, throughput 3.87309K wps
[Epoch 30 Batch 120/173] avg loss 0.000194605, throughput 3.86382K wps
[Epoch 30 Batch 150/173] avg loss 0.000171254, throughput 3.87175K wps
Begin Testing...
[Epoch 30] train avg loss 0.000169871, test acc 0.7844, test avg loss 0.746248, throughput 3.89437K wps
[Epoch 31 Batch 30/173] avg loss 0.000141022, throughput 4.00375K wps
[Epoch 31 Batch 60/173] avg loss 0.000173511, throughput 3.89469K wps
[Epoch 31 Batch 90/173] avg loss 0.000146119, throughput 3.88974K wps
[Epoch 31 Batch 120/173] avg loss 0.000176843, throughput 3.8993K wps
[Epoch 31 Batch 150/173] avg loss 0.000178812, throughput 3.87866K wps
Begin Testing...
[Epoch 31] train avg loss 0.000167403, test acc 0.7812, test avg loss 0.767954, throughput 3.90718K wps
[Epoch 32 Batch 30/173] avg loss 0.000111227, throughput 3.98244K wps
[Epoch 32 Batch 60/173] avg loss 0.000140802, throughput 3.87789K wps
[Epoch 32 Batch 90/173] avg loss 0.000140869, throughput 3.8741K wps
[Epoch 32 Batch 120/173] avg loss 0.000138826, throughput 3.87479K wps
[Epoch 32 Batch 150/173] avg loss 0.000159369, throughput 3.91153K wps
Begin Testing...
[Epoch 32] train avg loss 0.000141424, test acc 0.7844, test avg loss 0.788787, throughput 3.90111K wps
[Epoch 33 Batch 30/173] avg loss 0.000118187, throughput 3.98552K wps
[Epoch 33 Batch 60/173] avg loss 0.000130852, throughput 3.89628K wps
[Epoch 33 Batch 90/173] avg loss 0.0001153, throughput 3.9071K wps
[Epoch 33 Batch 120/173] avg loss 0.000114849, throughput 3.8826K wps
[Epoch 33 Batch 150/173] avg loss 0.000120869, throughput 3.8764K wps
Begin Testing...
[Epoch 33] train avg loss 0.000121604, test acc 0.7875, test avg loss 0.810817, throughput 3.9065K wps
[Epoch 34 Batch 30/173] avg loss 0.000115373, throughput 3.97228K wps
[Epoch 34 Batch 60/173] avg loss 0.000120721, throughput 3.87126K wps
[Epoch 34 Batch 90/173] avg loss 0.000106836, throughput 3.88669K wps
[Epoch 34 Batch 120/173] avg loss 0.000111128, throughput 3.86135K wps
[Epoch 34 Batch 150/173] avg loss 0.000130692, throughput 3.88082K wps
Begin Testing...
[Epoch 34] train avg loss 0.000116059, test acc 0.7854, test avg loss 0.838174, throughput 3.89347K wps
[Epoch 35 Batch 30/173] avg loss 9.01472e-05, throughput 3.97421K wps
[Epoch 35 Batch 60/173] avg loss 0.000105616, throughput 3.88207K wps
[Epoch 35 Batch 90/173] avg loss 8.00912e-05, throughput 3.88385K wps
[Epoch 35 Batch 120/173] avg loss 0.000112039, throughput 3.9081K wps
[Epoch 35 Batch 150/173] avg loss 9.37354e-05, throughput 3.8711K wps
Begin Testing...
[Epoch 35] train avg loss 9.79216e-05, test acc 0.7771, test avg loss 0.857753, throughput 3.9006K wps
[Epoch 36 Batch 30/173] avg loss 7.26568e-05, throughput 3.9909K wps
[Epoch 36 Batch 60/173] avg loss 8.69132e-05, throughput 3.87921K wps
[Epoch 36 Batch 90/173] avg loss 0.000113029, throughput 3.86927K wps
[Epoch 36 Batch 120/173] avg loss 0.000108594, throughput 3.86899K wps
[Epoch 36 Batch 150/173] avg loss 8.40718e-05, throughput 3.90962K wps
Begin Testing...
[Epoch 36] train avg loss 9.12814e-05, test acc 0.7823, test avg loss 0.868041, throughput 3.90113K wps
[Epoch 37 Batch 30/173] avg loss 7.02868e-05, throughput 3.99018K wps
[Epoch 37 Batch 60/173] avg loss 7.36232e-05, throughput 3.87528K wps
[Epoch 37 Batch 90/173] avg loss 8.40262e-05, throughput 3.88009K wps
[Epoch 37 Batch 120/173] avg loss 8.18814e-05, throughput 3.88781K wps
[Epoch 37 Batch 150/173] avg loss 6.25273e-05, throughput 3.87489K wps
Begin Testing...
[Epoch 37] train avg loss 7.57583e-05, test acc 0.7802, test avg loss 0.889554, throughput 3.89846K wps
[Epoch 38 Batch 30/173] avg loss 7.09995e-05, throughput 3.9798K wps
[Epoch 38 Batch 60/173] avg loss 6.28389e-05, throughput 3.87782K wps
[Epoch 38 Batch 90/173] avg loss 6.3176e-05, throughput 3.88466K wps
[Epoch 38 Batch 120/173] avg loss 6.24668e-05, throughput 3.88461K wps
[Epoch 38 Batch 150/173] avg loss 7.5743e-05, throughput 3.89557K wps
Begin Testing...
[Epoch 38] train avg loss 6.57345e-05, test acc 0.7802, test avg loss 0.911938, throughput 3.89874K wps
[Epoch 39 Batch 30/173] avg loss 4.93374e-05, throughput 3.99134K wps
[Epoch 39 Batch 60/173] avg loss 6.09223e-05, throughput 3.8929K wps
[Epoch 39 Batch 90/173] avg loss 5.56564e-05, throughput 3.9027K wps
[Epoch 39 Batch 120/173] avg loss 5.69234e-05, throughput 3.89885K wps
[Epoch 39 Batch 150/173] avg loss 5.20937e-05, throughput 3.87835K wps
Begin Testing...
[Epoch 39] train avg loss 5.53952e-05, test acc 0.7760, test avg loss 0.924712, throughput 3.90733K wps
[Epoch 40 Batch 30/173] avg loss 4.66178e-05, throughput 3.95687K wps
[Epoch 40 Batch 60/173] avg loss 5.66321e-05, throughput 3.87631K wps
[Epoch 40 Batch 90/173] avg loss 6.53149e-05, throughput 3.88707K wps
[Epoch 40 Batch 120/173] avg loss 4.16306e-05, throughput 3.90795K wps
[Epoch 40 Batch 150/173] avg loss 6.01415e-05, throughput 3.87962K wps
Begin Testing...
[Epoch 40] train avg loss 5.5769e-05, test acc 0.7833, test avg loss 0.9386, throughput 3.89724K wps
[Epoch 41 Batch 30/173] avg loss 5.00493e-05, throughput 3.97196K wps
[Epoch 41 Batch 60/173] avg loss 4.93188e-05, throughput 3.89044K wps
[Epoch 41 Batch 90/173] avg loss 5.85449e-05, throughput 3.88827K wps
[Epoch 41 Batch 120/173] avg loss 5.43949e-05, throughput 3.90782K wps
[Epoch 41 Batch 150/173] avg loss 5.91514e-05, throughput 3.88015K wps
Begin Testing...
[Epoch 41] train avg loss 5.25506e-05, test acc 0.7802, test avg loss 0.956927, throughput 3.90084K wps
[Epoch 42 Batch 30/173] avg loss 5.09035e-05, throughput 3.97215K wps
[Epoch 42 Batch 60/173] avg loss 3.93654e-05, throughput 3.88383K wps
[Epoch 42 Batch 90/173] avg loss 3.78993e-05, throughput 3.8882K wps
[Epoch 42 Batch 120/173] avg loss 4.35394e-05, throughput 3.90693K wps
[Epoch 42 Batch 150/173] avg loss 3.89554e-05, throughput 3.89321K wps
Begin Testing...
[Epoch 42] train avg loss 4.36441e-05, test acc 0.7792, test avg loss 0.976064, throughput 3.90266K wps
[Epoch 43 Batch 30/173] avg loss 3.37942e-05, throughput 3.96934K wps
[Epoch 43 Batch 60/173] avg loss 3.7348e-05, throughput 3.87375K wps
[Epoch 43 Batch 90/173] avg loss 3.79856e-05, throughput 3.87611K wps
[Epoch 43 Batch 120/173] avg loss 3.11068e-05, throughput 3.8934K wps
[Epoch 43 Batch 150/173] avg loss 4.03911e-05, throughput 3.90906K wps
Begin Testing...
[Epoch 43] train avg loss 3.49962e-05, test acc 0.7844, test avg loss 0.995405, throughput 3.90026K wps
[Epoch 44 Batch 30/173] avg loss 2.83979e-05, throughput 3.97011K wps
[Epoch 44 Batch 60/173] avg loss 3.19031e-05, throughput 3.87599K wps
[Epoch 44 Batch 90/173] avg loss 3.38948e-05, throughput 3.88305K wps
[Epoch 44 Batch 120/173] avg loss 2.89506e-05, throughput 3.87502K wps
[Epoch 44 Batch 150/173] avg loss 3.18507e-05, throughput 3.89012K wps
Begin Testing...
[Epoch 44] train avg loss 3.15174e-05, test acc 0.7771, test avg loss 1.01762, throughput 3.89771K wps
[Epoch 45 Batch 30/173] avg loss 3.2485e-05, throughput 3.99376K wps
[Epoch 45 Batch 60/173] avg loss 3.95947e-05, throughput 3.89179K wps
[Epoch 45 Batch 90/173] avg loss 2.83036e-05, throughput 3.89355K wps
[Epoch 45 Batch 120/173] avg loss 3.03139e-05, throughput 3.87129K wps
[Epoch 45 Batch 150/173] avg loss 2.83586e-05, throughput 3.89394K wps
Begin Testing...
[Epoch 45] train avg loss 3.15961e-05, test acc 0.7781, test avg loss 1.0364, throughput 3.90368K wps
[Epoch 46 Batch 30/173] avg loss 3.17206e-05, throughput 3.97116K wps
[Epoch 46 Batch 60/173] avg loss 2.52883e-05, throughput 3.86944K wps
[Epoch 46 Batch 90/173] avg loss 3.3056e-05, throughput 3.88718K wps
[Epoch 46 Batch 120/173] avg loss 2.31142e-05, throughput 3.88286K wps
[Epoch 46 Batch 150/173] avg loss 2.87068e-05, throughput 3.91095K wps
Begin Testing...
[Epoch 46] train avg loss 2.93542e-05, test acc 0.7688, test avg loss 1.05474, throughput 3.90134K wps
[Epoch 47 Batch 30/173] avg loss 2.14521e-05, throughput 3.96932K wps
[Epoch 47 Batch 60/173] avg loss 2.99008e-05, throughput 3.86848K wps
[Epoch 47 Batch 90/173] avg loss 3.67896e-05, throughput 3.8767K wps
[Epoch 47 Batch 120/173] avg loss 3.1279e-05, throughput 3.87874K wps
[Epoch 47 Batch 150/173] avg loss 2.5442e-05, throughput 3.88899K wps
Begin Testing...
[Epoch 47] train avg loss 3.42277e-05, test acc 0.7708, test avg loss 1.05679, throughput 3.89552K wps
[Epoch 48 Batch 30/173] avg loss 3.64313e-05, throughput 3.95202K wps
[Epoch 48 Batch 60/173] avg loss 2.73214e-05, throughput 3.8813K wps
[Epoch 48 Batch 90/173] avg loss 3.34175e-05, throughput 3.897K wps
[Epoch 48 Batch 120/173] avg loss 2.59555e-05, throughput 3.88564K wps
[Epoch 48 Batch 150/173] avg loss 2.99829e-05, throughput 3.88793K wps
Begin Testing...
[Epoch 48] train avg loss 3.13855e-05, test acc 0.7698, test avg loss 1.08781, throughput 3.90217K wps
[Epoch 49 Batch 30/173] avg loss 4.74579e-05, throughput 3.97036K wps
[Epoch 49 Batch 60/173] avg loss 2.62251e-05, throughput 3.88527K wps
[Epoch 49 Batch 90/173] avg loss 4.46037e-05, throughput 3.85985K wps
[Epoch 49 Batch 120/173] avg loss 3.049e-05, throughput 3.88776K wps
[Epoch 49 Batch 150/173] avg loss 3.38956e-05, throughput 3.88185K wps
Begin Testing...
[Epoch 49] train avg loss 3.52101e-05, test acc 0.7729, test avg loss 1.10972, throughput 3.89251K wps
[Epoch 50 Batch 30/173] avg loss 2.36617e-05, throughput 3.96984K wps
[Epoch 50 Batch 60/173] avg loss 2.17325e-05, throughput 3.86948K wps
[Epoch 50 Batch 90/173] avg loss 1.89172e-05, throughput 3.87858K wps
[Epoch 50 Batch 120/173] avg loss 3.29707e-05, throughput 3.90195K wps
[Epoch 50 Batch 150/173] avg loss 2.58693e-05, throughput 3.90053K wps
Begin Testing...
[Epoch 50] train avg loss 2.40827e-05, test acc 0.7708, test avg loss 1.13633, throughput 3.90119K wps
[Epoch 51 Batch 30/173] avg loss 2.20971e-05, throughput 3.96492K wps
[Epoch 51 Batch 60/173] avg loss 2.12609e-05, throughput 3.87257K wps
[Epoch 51 Batch 90/173] avg loss 1.75647e-05, throughput 3.86505K wps
[Epoch 51 Batch 120/173] avg loss 1.71649e-05, throughput 3.87516K wps
[Epoch 51 Batch 150/173] avg loss 3.18293e-05, throughput 3.87457K wps
Begin Testing...
[Epoch 51] train avg loss 2.17017e-05, test acc 0.7760, test avg loss 1.16088, throughput 3.89002K wps
[Epoch 52 Batch 30/173] avg loss 1.05863e-05, throughput 3.97033K wps
[Epoch 52 Batch 60/173] avg loss 1.72045e-05, throughput 3.88917K wps
[Epoch 52 Batch 90/173] avg loss 1.68308e-05, throughput 3.90382K wps
[Epoch 52 Batch 120/173] avg loss 1.86308e-05, throughput 3.88385K wps
[Epoch 52 Batch 150/173] avg loss 1.80171e-05, throughput 3.88746K wps
Begin Testing...
[Epoch 52] train avg loss 1.65429e-05, test acc 0.7698, test avg loss 1.17425, throughput 3.9071K wps
[Epoch 53 Batch 30/173] avg loss 1.61271e-05, throughput 3.99784K wps
[Epoch 53 Batch 60/173] avg loss 1.95679e-05, throughput 3.87634K wps
[Epoch 53 Batch 90/173] avg loss 1.46658e-05, throughput 3.87273K wps
[Epoch 53 Batch 120/173] avg loss 1.97684e-05, throughput 3.88456K wps
[Epoch 53 Batch 150/173] avg loss 2.10054e-05, throughput 3.89786K wps
Begin Testing...
[Epoch 53] train avg loss 1.78827e-05, test acc 0.7750, test avg loss 1.18582, throughput 3.90274K wps
[Epoch 54 Batch 30/173] avg loss 1.24231e-05, throughput 3.96533K wps
[Epoch 54 Batch 60/173] avg loss 9.66968e-06, throughput 3.88236K wps
[Epoch 54 Batch 90/173] avg loss 1.33575e-05, throughput 3.88601K wps
[Epoch 54 Batch 120/173] avg loss 1.32544e-05, throughput 3.89008K wps
[Epoch 54 Batch 150/173] avg loss 1.30334e-05, throughput 3.90835K wps
Begin Testing...
[Epoch 54] train avg loss 1.25876e-05, test acc 0.7750, test avg loss 1.19413, throughput 3.90435K wps
[Epoch 55 Batch 30/173] avg loss 1.82953e-05, throughput 3.98453K wps
[Epoch 55 Batch 60/173] avg loss 1.23792e-05, throughput 3.89058K wps
[Epoch 55 Batch 90/173] avg loss 1.27002e-05, throughput 3.90067K wps
[Epoch 55 Batch 120/173] avg loss 1.044e-05, throughput 3.9036K wps
[Epoch 55 Batch 150/173] avg loss 1.08453e-05, throughput 3.89717K wps
Begin Testing...
[Epoch 55] train avg loss 1.33213e-05, test acc 0.7677, test avg loss 1.23889, throughput 3.91616K wps
[Epoch 56 Batch 30/173] avg loss 2.61591e-05, throughput 4.00553K wps
[Epoch 56 Batch 60/173] avg loss 1.82393e-05, throughput 3.88832K wps
[Epoch 56 Batch 90/173] avg loss 1.0378e-05, throughput 3.89326K wps
[Epoch 56 Batch 120/173] avg loss 1.69582e-05, throughput 3.90544K wps
[Epoch 56 Batch 150/173] avg loss 1.54086e-05, throughput 3.88681K wps
Begin Testing...
[Epoch 56] train avg loss 1.74595e-05, test acc 0.7625, test avg loss 1.26503, throughput 3.91103K wps
[Epoch 57 Batch 30/173] avg loss 1.41649e-05, throughput 3.98023K wps
[Epoch 57 Batch 60/173] avg loss 1.10372e-05, throughput 3.88946K wps
[Epoch 57 Batch 90/173] avg loss 1.34612e-05, throughput 3.91234K wps
[Epoch 57 Batch 120/173] avg loss 1.14486e-05, throughput 3.88889K wps
[Epoch 57 Batch 150/173] avg loss 1.34307e-05, throughput 3.88788K wps
Begin Testing...
[Epoch 57] train avg loss 1.27892e-05, test acc 0.7677, test avg loss 1.26745, throughput 3.91158K wps
[Epoch 58 Batch 30/173] avg loss 9.08037e-06, throughput 3.99988K wps
[Epoch 58 Batch 60/173] avg loss 8.1598e-06, throughput 3.88573K wps
[Epoch 58 Batch 90/173] avg loss 1.02959e-05, throughput 3.87903K wps
[Epoch 58 Batch 120/173] avg loss 8.66436e-06, throughput 3.89438K wps
[Epoch 58 Batch 150/173] avg loss 1.05221e-05, throughput 3.90385K wps
Begin Testing...
[Epoch 58] train avg loss 9.69866e-06, test acc 0.7667, test avg loss 1.30406, throughput 3.90852K wps
[Epoch 59 Batch 30/173] avg loss 9.85143e-06, throughput 3.98466K wps
[Epoch 59 Batch 60/173] avg loss 7.26707e-06, throughput 3.88902K wps
[Epoch 59 Batch 90/173] avg loss 9.69114e-06, throughput 3.91893K wps
[Epoch 59 Batch 120/173] avg loss 7.71787e-06, throughput 3.89704K wps
[Epoch 59 Batch 150/173] avg loss 8.46126e-06, throughput 3.90144K wps
Begin Testing...
[Epoch 59] train avg loss 8.39676e-06, test acc 0.7656, test avg loss 1.30167, throughput 3.91741K wps
Test loss 0.414864, test acc 0.8086
Total time cost 553.88s
9596 1066
[Epoch 0 Batch 30/173] avg loss 0.0155794, throughput 3.68564K wps
[Epoch 0 Batch 60/173] avg loss 0.0151298, throughput 3.91063K wps
[Epoch 0 Batch 90/173] avg loss 0.0142535, throughput 3.89758K wps
[Epoch 0 Batch 120/173] avg loss 0.0144468, throughput 3.88751K wps
[Epoch 0 Batch 150/173] avg loss 0.0141655, throughput 3.89861K wps
Begin Testing...
[Epoch 0] train avg loss 0.0145968, test acc 0.6302, test avg loss 0.657421, throughput 3.86018K wps
Observed Improvement.
Begin Testing...
[Epoch 1 Batch 30/173] avg loss 0.0131345, throughput 3.99197K wps
[Epoch 1 Batch 60/173] avg loss 0.013153, throughput 3.89855K wps
[Epoch 1 Batch 90/173] avg loss 0.0131008, throughput 3.88406K wps
[Epoch 1 Batch 120/173] avg loss 0.0130032, throughput 3.89363K wps
[Epoch 1 Batch 150/173] avg loss 0.012799, throughput 3.91448K wps
Begin Testing...
[Epoch 1] train avg loss 0.013021, test acc 0.6750, test avg loss 0.624194, throughput 3.91274K wps
Observed Improvement.
Begin Testing...
[Epoch 2 Batch 30/173] avg loss 0.0119394, throughput 3.9931K wps
[Epoch 2 Batch 60/173] avg loss 0.0120509, throughput 3.88969K wps
[Epoch 2 Batch 90/173] avg loss 0.0120588, throughput 3.89489K wps
[Epoch 2 Batch 120/173] avg loss 0.0117995, throughput 3.90313K wps
[Epoch 2 Batch 150/173] avg loss 0.0118741, throughput 3.88557K wps
Begin Testing...
[Epoch 2] train avg loss 0.0119397, test acc 0.7073, test avg loss 0.59401, throughput 3.91045K wps
Observed Improvement.
Begin Testing...
[Epoch 3 Batch 30/173] avg loss 0.0109509, throughput 3.98476K wps
[Epoch 3 Batch 60/173] avg loss 0.0111092, throughput 3.8957K wps
[Epoch 3 Batch 90/173] avg loss 0.0110705, throughput 3.91695K wps
[Epoch 3 Batch 120/173] avg loss 0.0109329, throughput 3.89195K wps
[Epoch 3 Batch 150/173] avg loss 0.0108372, throughput 3.88816K wps
Begin Testing...
[Epoch 3] train avg loss 0.010928, test acc 0.7594, test avg loss 0.557845, throughput 3.91499K wps
Observed Improvement.
Begin Testing...
[Epoch 4 Batch 30/173] avg loss 0.0100819, throughput 3.99242K wps
[Epoch 4 Batch 60/173] avg loss 0.00985398, throughput 3.88634K wps
[Epoch 4 Batch 90/173] avg loss 0.00981938, throughput 3.8935K wps
[Epoch 4 Batch 120/173] avg loss 0.00994965, throughput 3.90115K wps
[Epoch 4 Batch 150/173] avg loss 0.00985282, throughput 3.90711K wps
Begin Testing...
[Epoch 4] train avg loss 0.0098984, test acc 0.7792, test avg loss 0.51983, throughput 3.91118K wps
Observed Improvement.
Begin Testing...
[Epoch 5 Batch 30/173] avg loss 0.00878041, throughput 3.99781K wps
[Epoch 5 Batch 60/173] avg loss 0.0088594, throughput 3.89418K wps
[Epoch 5 Batch 90/173] avg loss 0.00897107, throughput 3.91929K wps
[Epoch 5 Batch 120/173] avg loss 0.00891175, throughput 3.88713K wps
[Epoch 5 Batch 150/173] avg loss 0.00872471, throughput 3.89825K wps
Begin Testing...
[Epoch 5] train avg loss 0.00883168, test acc 0.7854, test avg loss 0.488826, throughput 3.91947K wps
Observed Improvement.
Begin Testing...
[Epoch 6 Batch 30/173] avg loss 0.00813827, throughput 3.98481K wps
[Epoch 6 Batch 60/173] avg loss 0.00780325, throughput 3.90375K wps
[Epoch 6 Batch 90/173] avg loss 0.00770624, throughput 3.88287K wps
[Epoch 6 Batch 120/173] avg loss 0.00766924, throughput 3.90367K wps
[Epoch 6 Batch 150/173] avg loss 0.00753482, throughput 3.90207K wps
Begin Testing...
[Epoch 6] train avg loss 0.00775445, test acc 0.7948, test avg loss 0.465644, throughput 3.91131K wps
Observed Improvement.
Begin Testing...
[Epoch 7 Batch 30/173] avg loss 0.00711864, throughput 4.00222K wps
[Epoch 7 Batch 60/173] avg loss 0.00691532, throughput 3.88892K wps
[Epoch 7 Batch 90/173] avg loss 0.00704048, throughput 3.8909K wps
[Epoch 7 Batch 120/173] avg loss 0.00665718, throughput 3.91006K wps
[Epoch 7 Batch 150/173] avg loss 0.00684155, throughput 3.8869K wps
Begin Testing...
[Epoch 7] train avg loss 0.00689039, test acc 0.8052, test avg loss 0.446045, throughput 3.91355K wps
Observed Improvement.
Begin Testing...
[Epoch 8 Batch 30/173] avg loss 0.00621079, throughput 3.96807K wps
[Epoch 8 Batch 60/173] avg loss 0.00607095, throughput 3.89013K wps
[Epoch 8 Batch 90/173] avg loss 0.00594526, throughput 3.90298K wps
[Epoch 8 Batch 120/173] avg loss 0.00584666, throughput 3.89861K wps
[Epoch 8 Batch 150/173] avg loss 0.00585878, throughput 3.88523K wps
Begin Testing...
[Epoch 8] train avg loss 0.00599308, test acc 0.8052, test avg loss 0.435367, throughput 3.90996K wps
Observed Improvement.
Begin Testing...
[Epoch 9 Batch 30/173] avg loss 0.00516349, throughput 3.97993K wps
[Epoch 9 Batch 60/173] avg loss 0.00517158, throughput 3.89538K wps
[Epoch 9 Batch 90/173] avg loss 0.00529348, throughput 3.90075K wps
[Epoch 9 Batch 120/173] avg loss 0.00535947, throughput 3.87821K wps
[Epoch 9 Batch 150/173] avg loss 0.00534634, throughput 3.89202K wps
Begin Testing...
[Epoch 9] train avg loss 0.00525098, test acc 0.8083, test avg loss 0.426687, throughput 3.91069K wps
Observed Improvement.
Begin Testing...
[Epoch 10 Batch 30/173] avg loss 0.0047962, throughput 3.99903K wps
[Epoch 10 Batch 60/173] avg loss 0.00462766, throughput 3.90106K wps
[Epoch 10 Batch 90/173] avg loss 0.00468113, throughput 3.89016K wps
[Epoch 10 Batch 120/173] avg loss 0.00447571, throughput 3.90112K wps
[Epoch 10 Batch 150/173] avg loss 0.00445101, throughput 3.90464K wps
Begin Testing...
[Epoch 10] train avg loss 0.00459022, test acc 0.8073, test avg loss 0.432321, throughput 3.91353K wps
[Epoch 11 Batch 30/173] avg loss 0.00399142, throughput 3.98869K wps
[Epoch 11 Batch 60/173] avg loss 0.00397918, throughput 3.89093K wps
[Epoch 11 Batch 90/173] avg loss 0.00378537, throughput 3.92625K wps
[Epoch 11 Batch 120/173] avg loss 0.00385852, throughput 3.89177K wps
[Epoch 11 Batch 150/173] avg loss 0.00388274, throughput 3.89528K wps
Begin Testing...
[Epoch 11] train avg loss 0.0039186, test acc 0.8115, test avg loss 0.428953, throughput 3.91808K wps
Observed Improvement.
Begin Testing...
[Epoch 12 Batch 30/173] avg loss 0.00338849, throughput 4.00824K wps
[Epoch 12 Batch 60/173] avg loss 0.00328058, throughput 3.88684K wps
[Epoch 12 Batch 90/173] avg loss 0.00338308, throughput 3.89405K wps
[Epoch 12 Batch 120/173] avg loss 0.00320174, throughput 3.9208K wps
[Epoch 12 Batch 150/173] avg loss 0.00338956, throughput 3.88667K wps
Begin Testing...
[Epoch 12] train avg loss 0.00333932, test acc 0.8073, test avg loss 0.430934, throughput 3.91748K wps
[Epoch 13 Batch 30/173] avg loss 0.00267384, throughput 3.97394K wps
[Epoch 13 Batch 60/173] avg loss 0.00299378, throughput 3.90156K wps
[Epoch 13 Batch 90/173] avg loss 0.00290767, throughput 3.90121K wps
[Epoch 13 Batch 120/173] avg loss 0.00271098, throughput 3.88357K wps
[Epoch 13 Batch 150/173] avg loss 0.00270902, throughput 3.88946K wps
Begin Testing...
[Epoch 13] train avg loss 0.00282816, test acc 0.8083, test avg loss 0.445342, throughput 3.90882K wps
[Epoch 14 Batch 30/173] avg loss 0.00231421, throughput 3.99186K wps
[Epoch 14 Batch 60/173] avg loss 0.00243848, throughput 3.87839K wps
[Epoch 14 Batch 90/173] avg loss 0.0023987, throughput 3.88382K wps
[Epoch 14 Batch 120/173] avg loss 0.00237951, throughput 3.87861K wps
[Epoch 14 Batch 150/173] avg loss 0.00228337, throughput 3.89615K wps
Begin Testing...
[Epoch 14] train avg loss 0.00237426, test acc 0.8031, test avg loss 0.454278, throughput 3.90206K wps
[Epoch 15 Batch 30/173] avg loss 0.00195891, throughput 3.98739K wps
[Epoch 15 Batch 60/173] avg loss 0.0020779, throughput 3.87323K wps
[Epoch 15 Batch 90/173] avg loss 0.00189202, throughput 3.88235K wps
[Epoch 15 Batch 120/173] avg loss 0.00192094, throughput 3.89058K wps
[Epoch 15 Batch 150/173] avg loss 0.0018839, throughput 3.89687K wps
Begin Testing...
[Epoch 15] train avg loss 0.00200181, test acc 0.7969, test avg loss 0.467948, throughput 3.90209K wps
[Epoch 16 Batch 30/173] avg loss 0.00189293, throughput 3.99557K wps
[Epoch 16 Batch 60/173] avg loss 0.00164092, throughput 3.87142K wps
[Epoch 16 Batch 90/173] avg loss 0.00173329, throughput 3.88062K wps
[Epoch 16 Batch 120/173] avg loss 0.00162667, throughput 3.89868K wps
[Epoch 16 Batch 150/173] avg loss 0.00165529, throughput 3.90633K wps
Begin Testing...
[Epoch 16] train avg loss 0.00174278, test acc 0.7990, test avg loss 0.482265, throughput 3.90617K wps
[Epoch 17 Batch 30/173] avg loss 0.00157685, throughput 3.96461K wps
[Epoch 17 Batch 60/173] avg loss 0.00136474, throughput 3.87554K wps
[Epoch 17 Batch 90/173] avg loss 0.00147994, throughput 3.8714K wps
[Epoch 17 Batch 120/173] avg loss 0.00139766, throughput 3.8851K wps
[Epoch 17 Batch 150/173] avg loss 0.00127222, throughput 3.91008K wps
Begin Testing...
[Epoch 17] train avg loss 0.00142405, test acc 0.7906, test avg loss 0.495366, throughput 3.89951K wps
[Epoch 18 Batch 30/173] avg loss 0.00118413, throughput 3.97166K wps
[Epoch 18 Batch 60/173] avg loss 0.00132967, throughput 3.87621K wps
[Epoch 18 Batch 90/173] avg loss 0.00121786, throughput 3.88044K wps
[Epoch 18 Batch 120/173] avg loss 0.00120733, throughput 3.87928K wps
[Epoch 18 Batch 150/173] avg loss 0.00124917, throughput 3.89544K wps
Begin Testing...
[Epoch 18] train avg loss 0.00122464, test acc 0.7958, test avg loss 0.512028, throughput 3.89872K wps
[Epoch 19 Batch 30/173] avg loss 0.0010658, throughput 3.97841K wps
[Epoch 19 Batch 60/173] avg loss 0.000936076, throughput 3.88587K wps
[Epoch 19 Batch 90/173] avg loss 0.000976805, throughput 3.88553K wps
[Epoch 19 Batch 120/173] avg loss 0.00102804, throughput 3.87804K wps
[Epoch 19 Batch 150/173] avg loss 0.00107349, throughput 3.88766K wps
Begin Testing...
[Epoch 19] train avg loss 0.00101362, test acc 0.7906, test avg loss 0.542877, throughput 3.89763K wps
[Epoch 20 Batch 30/173] avg loss 0.000840797, throughput 3.98481K wps
[Epoch 20 Batch 60/173] avg loss 0.000883923, throughput 3.88352K wps
[Epoch 20 Batch 90/173] avg loss 0.000824348, throughput 3.88032K wps
[Epoch 20 Batch 120/173] avg loss 0.000884344, throughput 3.90079K wps
[Epoch 20 Batch 150/173] avg loss 0.000858885, throughput 3.88082K wps
Begin Testing...
[Epoch 20] train avg loss 0.000862623, test acc 0.7865, test avg loss 0.554006, throughput 3.90091K wps
[Epoch 21 Batch 30/173] avg loss 0.000752231, throughput 3.97905K wps
[Epoch 21 Batch 60/173] avg loss 0.000741169, throughput 3.872K wps
[Epoch 21 Batch 90/173] avg loss 0.000692621, throughput 3.87868K wps
[Epoch 21 Batch 120/173] avg loss 0.000790656, throughput 3.89026K wps
[Epoch 21 Batch 150/173] avg loss 0.000637565, throughput 3.90832K wps
Begin Testing...
[Epoch 21] train avg loss 0.000728005, test acc 0.7875, test avg loss 0.569987, throughput 3.90159K wps
[Epoch 22 Batch 30/173] avg loss 0.000624355, throughput 3.98732K wps
[Epoch 22 Batch 60/173] avg loss 0.000598661, throughput 3.88006K wps
[Epoch 22 Batch 90/173] avg loss 0.000632161, throughput 3.87202K wps
[Epoch 22 Batch 120/173] avg loss 0.000655299, throughput 3.8846K wps
[Epoch 22 Batch 150/173] avg loss 0.000617229, throughput 3.90535K wps
Begin Testing...
[Epoch 22] train avg loss 0.000614431, test acc 0.7833, test avg loss 0.5958, throughput 3.90452K wps
[Epoch 23 Batch 30/173] avg loss 0.000539403, throughput 3.98246K wps
[Epoch 23 Batch 60/173] avg loss 0.000608026, throughput 3.88275K wps
[Epoch 23 Batch 90/173] avg loss 0.000583507, throughput 3.86358K wps
[Epoch 23 Batch 120/173] avg loss 0.000612035, throughput 3.88157K wps
[Epoch 23 Batch 150/173] avg loss 0.000538577, throughput 3.87268K wps
Begin Testing...
[Epoch 23] train avg loss 0.000576145, test acc 0.7865, test avg loss 0.611084, throughput 3.89611K wps
[Epoch 24 Batch 30/173] avg loss 0.000469386, throughput 3.99588K wps
[Epoch 24 Batch 60/173] avg loss 0.000440391, throughput 3.88832K wps
[Epoch 24 Batch 90/173] avg loss 0.000474677, throughput 3.89205K wps
[Epoch 24 Batch 120/173] avg loss 0.000460684, throughput 3.87967K wps
[Epoch 24 Batch 150/173] avg loss 0.00042435, throughput 3.89025K wps
Begin Testing...
[Epoch 24] train avg loss 0.00045687, test acc 0.7812, test avg loss 0.629122, throughput 3.90228K wps
[Epoch 25 Batch 30/173] avg loss 0.000423295, throughput 3.97576K wps
[Epoch 25 Batch 60/173] avg loss 0.000381084, throughput 3.8886K wps
[Epoch 25 Batch 90/173] avg loss 0.000485241, throughput 3.88471K wps
[Epoch 25 Batch 120/173] avg loss 0.000401088, throughput 3.91158K wps
[Epoch 25 Batch 150/173] avg loss 0.000368277, throughput 3.89815K wps
Begin Testing...
[Epoch 25] train avg loss 0.000414663, test acc 0.7802, test avg loss 0.648977, throughput 3.90546K wps
[Epoch 26 Batch 30/173] avg loss 0.000333839, throughput 3.97379K wps
[Epoch 26 Batch 60/173] avg loss 0.000335124, throughput 3.88213K wps
[Epoch 26 Batch 90/173] avg loss 0.000328436, throughput 3.86621K wps
[Epoch 26 Batch 120/173] avg loss 0.00032894, throughput 3.88914K wps
[Epoch 26 Batch 150/173] avg loss 0.000297478, throughput 3.91285K wps
Begin Testing...
[Epoch 26] train avg loss 0.000330936, test acc 0.7781, test avg loss 0.672189, throughput 3.90231K wps
[Epoch 27 Batch 30/173] avg loss 0.00024841, throughput 3.98113K wps
[Epoch 27 Batch 60/173] avg loss 0.000348022, throughput 3.88879K wps
[Epoch 27 Batch 90/173] avg loss 0.000264552, throughput 3.89591K wps
[Epoch 27 Batch 120/173] avg loss 0.000252869, throughput 3.88592K wps
[Epoch 27 Batch 150/173] avg loss 0.000321374, throughput 3.86615K wps
Begin Testing...
[Epoch 27] train avg loss 0.000285109, test acc 0.7812, test avg loss 0.688306, throughput 3.8957K wps
[Epoch 28 Batch 30/173] avg loss 0.000221421, throughput 3.99268K wps
[Epoch 28 Batch 60/173] avg loss 0.000279914, throughput 3.88461K wps
[Epoch 28 Batch 90/173] avg loss 0.000246418, throughput 3.88324K wps
[Epoch 28 Batch 120/173] avg loss 0.000228162, throughput 3.89026K wps
[Epoch 28 Batch 150/173] avg loss 0.000226995, throughput 3.89368K wps
Begin Testing...
[Epoch 28] train avg loss 0.000245477, test acc 0.7792, test avg loss 0.712694, throughput 3.90305K wps
[Epoch 29 Batch 30/173] avg loss 0.00021, throughput 3.97745K wps
[Epoch 29 Batch 60/173] avg loss 0.000207191, throughput 3.87844K wps
[Epoch 29 Batch 90/173] avg loss 0.00024084, throughput 3.86603K wps
[Epoch 29 Batch 120/173] avg loss 0.000256449, throughput 3.87908K wps
[Epoch 29 Batch 150/173] avg loss 0.000255638, throughput 3.88118K wps
Begin Testing...
[Epoch 29] train avg loss 0.000229849, test acc 0.7802, test avg loss 0.72167, throughput 3.89741K wps
[Epoch 30 Batch 30/173] avg loss 0.00013846, throughput 3.98585K wps
[Epoch 30 Batch 60/173] avg loss 0.00019764, throughput 3.88548K wps
[Epoch 30 Batch 90/173] avg loss 0.000179813, throughput 3.87742K wps
[Epoch 30 Batch 120/173] avg loss 0.000209559, throughput 3.87505K wps
[Epoch 30 Batch 150/173] avg loss 0.000189668, throughput 3.8588K wps
Begin Testing...
[Epoch 30] train avg loss 0.000184487, test acc 0.7802, test avg loss 0.749217, throughput 3.89574K wps
[Epoch 31 Batch 30/173] avg loss 0.000166581, throughput 3.97999K wps
[Epoch 31 Batch 60/173] avg loss 0.000190139, throughput 3.90747K wps
[Epoch 31 Batch 90/173] avg loss 0.000181222, throughput 3.89151K wps
[Epoch 31 Batch 120/173] avg loss 0.000203513, throughput 3.87571K wps
[Epoch 31 Batch 150/173] avg loss 0.000171739, throughput 3.87829K wps
Begin Testing...
[Epoch 31] train avg loss 0.000178609, test acc 0.7760, test avg loss 0.778708, throughput 3.90616K wps
[Epoch 32 Batch 30/173] avg loss 0.000146338, throughput 3.97307K wps
[Epoch 32 Batch 60/173] avg loss 0.00016247, throughput 3.87824K wps
[Epoch 32 Batch 90/173] avg loss 0.000126226, throughput 3.90454K wps
[Epoch 32 Batch 120/173] avg loss 0.000160582, throughput 3.88426K wps
[Epoch 32 Batch 150/173] avg loss 0.000172255, throughput 3.90466K wps
Begin Testing...
[Epoch 32] train avg loss 0.000150427, test acc 0.7708, test avg loss 0.801566, throughput 3.90661K wps
[Epoch 33 Batch 30/173] avg loss 0.000130047, throughput 3.9843K wps
[Epoch 33 Batch 60/173] avg loss 0.000125826, throughput 3.87421K wps
[Epoch 33 Batch 90/173] avg loss 0.000137518, throughput 3.87319K wps
[Epoch 33 Batch 120/173] avg loss 0.000128283, throughput 3.86931K wps
[Epoch 33 Batch 150/173] avg loss 0.000150925, throughput 3.86582K wps
Begin Testing...
[Epoch 33] train avg loss 0.0001408, test acc 0.7646, test avg loss 0.806665, throughput 3.89567K wps
[Epoch 34 Batch 30/173] avg loss 0.000107653, throughput 3.98791K wps
[Epoch 34 Batch 60/173] avg loss 0.00012944, throughput 3.886K wps
[Epoch 34 Batch 90/173] avg loss 0.000136986, throughput 3.88886K wps
[Epoch 34 Batch 120/173] avg loss 9.05187e-05, throughput 3.87566K wps
[Epoch 34 Batch 150/173] avg loss 0.000112092, throughput 3.86777K wps
Begin Testing...
[Epoch 34] train avg loss 0.000116353, test acc 0.7729, test avg loss 0.840252, throughput 3.90087K wps
[Epoch 35 Batch 30/173] avg loss 9.96964e-05, throughput 3.974K wps
[Epoch 35 Batch 60/173] avg loss 8.59554e-05, throughput 3.87372K wps
[Epoch 35 Batch 90/173] avg loss 0.000136581, throughput 3.89393K wps
[Epoch 35 Batch 120/173] avg loss 8.54198e-05, throughput 3.88712K wps
[Epoch 35 Batch 150/173] avg loss 9.14899e-05, throughput 3.88923K wps
Begin Testing...
[Epoch 35] train avg loss 0.000100009, test acc 0.7677, test avg loss 0.844439, throughput 3.9048K wps
[Epoch 36 Batch 30/173] avg loss 7.77578e-05, throughput 3.98432K wps
[Epoch 36 Batch 60/173] avg loss 7.87439e-05, throughput 3.87691K wps
[Epoch 36 Batch 90/173] avg loss 9.24453e-05, throughput 3.87669K wps
[Epoch 36 Batch 120/173] avg loss 0.000102454, throughput 3.8725K wps
[Epoch 36 Batch 150/173] avg loss 8.54549e-05, throughput 3.87023K wps
Begin Testing...
[Epoch 36] train avg loss 8.79337e-05, test acc 0.7740, test avg loss 0.870042, throughput 3.89844K wps
[Epoch 37 Batch 30/173] avg loss 9.13451e-05, throughput 4.00684K wps
[Epoch 37 Batch 60/173] avg loss 7.75718e-05, throughput 3.87991K wps
[Epoch 37 Batch 90/173] avg loss 7.51147e-05, throughput 3.88658K wps
[Epoch 37 Batch 120/173] avg loss 7.86551e-05, throughput 3.90264K wps
[Epoch 37 Batch 150/173] avg loss 8.76174e-05, throughput 3.87692K wps
Begin Testing...
[Epoch 37] train avg loss 8.37516e-05, test acc 0.7729, test avg loss 0.878405, throughput 3.90456K wps
[Epoch 38 Batch 30/173] avg loss 8.37268e-05, throughput 3.97392K wps
[Epoch 38 Batch 60/173] avg loss 6.32626e-05, throughput 3.88172K wps
[Epoch 38 Batch 90/173] avg loss 6.65613e-05, throughput 3.88172K wps
[Epoch 38 Batch 120/173] avg loss 8.39608e-05, throughput 3.91068K wps
[Epoch 38 Batch 150/173] avg loss 7.39731e-05, throughput 3.89775K wps
Begin Testing...
[Epoch 38] train avg loss 7.39954e-05, test acc 0.7719, test avg loss 0.903337, throughput 3.90606K wps
[Epoch 39 Batch 30/173] avg loss 5.35375e-05, throughput 3.96845K wps
[Epoch 39 Batch 60/173] avg loss 7.96752e-05, throughput 3.86712K wps
[Epoch 39 Batch 90/173] avg loss 5.55166e-05, throughput 3.87049K wps
[Epoch 39 Batch 120/173] avg loss 5.22369e-05, throughput 3.88224K wps
[Epoch 39 Batch 150/173] avg loss 6.84118e-05, throughput 3.89661K wps
Begin Testing...
[Epoch 39] train avg loss 6.35967e-05, test acc 0.7625, test avg loss 0.957907, throughput 3.89417K wps
[Epoch 40 Batch 30/173] avg loss 5.9801e-05, throughput 3.96668K wps
[Epoch 40 Batch 60/173] avg loss 4.57284e-05, throughput 3.9155K wps
[Epoch 40 Batch 90/173] avg loss 6.64016e-05, throughput 3.89522K wps
[Epoch 40 Batch 120/173] avg loss 6.88119e-05, throughput 3.91699K wps
[Epoch 40 Batch 150/173] avg loss 5.99853e-05, throughput 3.88815K wps
Begin Testing...
[Epoch 40] train avg loss 6.06533e-05, test acc 0.7708, test avg loss 0.939335, throughput 3.91171K wps
[Epoch 41 Batch 30/173] avg loss 4.20016e-05, throughput 3.9797K wps
[Epoch 41 Batch 60/173] avg loss 5.46439e-05, throughput 3.88275K wps
[Epoch 41 Batch 90/173] avg loss 3.9563e-05, throughput 3.8808K wps
[Epoch 41 Batch 120/173] avg loss 4.76231e-05, throughput 3.87535K wps
[Epoch 41 Batch 150/173] avg loss 4.87896e-05, throughput 3.9041K wps
Begin Testing...
[Epoch 41] train avg loss 4.79354e-05, test acc 0.7688, test avg loss 0.961738, throughput 3.90174K wps
[Epoch 42 Batch 30/173] avg loss 4.00686e-05, throughput 3.97775K wps
[Epoch 42 Batch 60/173] avg loss 4.56991e-05, throughput 3.88857K wps
[Epoch 42 Batch 90/173] avg loss 5.21591e-05, throughput 3.89425K wps
[Epoch 42 Batch 120/173] avg loss 4.78527e-05, throughput 3.87574K wps
[Epoch 42 Batch 150/173] avg loss 4.83328e-05, throughput 3.8821K wps
Begin Testing...
[Epoch 42] train avg loss 4.61993e-05, test acc 0.7708, test avg loss 0.978434, throughput 3.89928K wps
[Epoch 43 Batch 30/173] avg loss 3.23541e-05, throughput 3.97088K wps
[Epoch 43 Batch 60/173] avg loss 4.12956e-05, throughput 3.8687K wps
[Epoch 43 Batch 90/173] avg loss 4.64673e-05, throughput 3.88534K wps
[Epoch 43 Batch 120/173] avg loss 4.33405e-05, throughput 3.90704K wps
[Epoch 43 Batch 150/173] avg loss 3.92708e-05, throughput 3.89454K wps
Begin Testing...
[Epoch 43] train avg loss 4.11904e-05, test acc 0.7677, test avg loss 1.00136, throughput 3.9016K wps
[Epoch 44 Batch 30/173] avg loss 4.63811e-05, throughput 3.96434K wps
[Epoch 44 Batch 60/173] avg loss 4.40922e-05, throughput 3.87975K wps
[Epoch 44 Batch 90/173] avg loss 5.09463e-05, throughput 3.87704K wps
[Epoch 44 Batch 120/173] avg loss 3.69478e-05, throughput 3.88302K wps
[Epoch 44 Batch 150/173] avg loss 5.50102e-05, throughput 3.88526K wps
Begin Testing...
[Epoch 44] train avg loss 4.693e-05, test acc 0.7656, test avg loss 1.01142, throughput 3.89324K wps
[Epoch 45 Batch 30/173] avg loss 3.02322e-05, throughput 4.00693K wps
[Epoch 45 Batch 60/173] avg loss 4.45416e-05, throughput 3.89052K wps
[Epoch 45 Batch 90/173] avg loss 4.17812e-05, throughput 3.88803K wps
[Epoch 45 Batch 120/173] avg loss 3.57116e-05, throughput 3.90855K wps
[Epoch 45 Batch 150/173] avg loss 3.41126e-05, throughput 3.88798K wps
Begin Testing...
[Epoch 45] train avg loss 3.68617e-05, test acc 0.7667, test avg loss 1.03447, throughput 3.91111K wps
[Epoch 46 Batch 30/173] avg loss 3.1568e-05, throughput 3.97018K wps
[Epoch 46 Batch 60/173] avg loss 2.82903e-05, throughput 3.87308K wps
[Epoch 46 Batch 90/173] avg loss 2.61989e-05, throughput 3.8797K wps
[Epoch 46 Batch 120/173] avg loss 3.43859e-05, throughput 3.8813K wps
[Epoch 46 Batch 150/173] avg loss 2.75064e-05, throughput 3.90368K wps
Begin Testing...
[Epoch 46] train avg loss 2.9445e-05, test acc 0.7688, test avg loss 1.04504, throughput 3.89851K wps
[Epoch 47 Batch 30/173] avg loss 2.49607e-05, throughput 3.98935K wps
[Epoch 47 Batch 60/173] avg loss 2.77221e-05, throughput 3.88259K wps
[Epoch 47 Batch 90/173] avg loss 2.82233e-05, throughput 3.88986K wps
[Epoch 47 Batch 120/173] avg loss 2.25191e-05, throughput 3.89653K wps
[Epoch 47 Batch 150/173] avg loss 2.69678e-05, throughput 3.87917K wps
Begin Testing...
[Epoch 47] train avg loss 2.54937e-05, test acc 0.7667, test avg loss 1.06935, throughput 3.90562K wps
[Epoch 48 Batch 30/173] avg loss 2.28915e-05, throughput 3.98457K wps
[Epoch 48 Batch 60/173] avg loss 3.05693e-05, throughput 3.90632K wps
[Epoch 48 Batch 90/173] avg loss 2.38572e-05, throughput 3.88221K wps
[Epoch 48 Batch 120/173] avg loss 2.60924e-05, throughput 3.8756K wps
[Epoch 48 Batch 150/173] avg loss 2.22807e-05, throughput 3.88239K wps
Begin Testing...
[Epoch 48] train avg loss 2.46144e-05, test acc 0.7677, test avg loss 1.08584, throughput 3.90254K wps
[Epoch 49 Batch 30/173] avg loss 1.92589e-05, throughput 3.98613K wps
[Epoch 49 Batch 60/173] avg loss 2.97103e-05, throughput 3.91248K wps
[Epoch 49 Batch 90/173] avg loss 2.18229e-05, throughput 3.88552K wps
[Epoch 49 Batch 120/173] avg loss 2.9079e-05, throughput 3.87612K wps
[Epoch 49 Batch 150/173] avg loss 2.06924e-05, throughput 3.86947K wps
Begin Testing...
[Epoch 49] train avg loss 2.34922e-05, test acc 0.7583, test avg loss 1.11619, throughput 3.90508K wps
[Epoch 50 Batch 30/173] avg loss 2.08905e-05, throughput 3.9612K wps
[Epoch 50 Batch 60/173] avg loss 4.12034e-05, throughput 3.90266K wps
[Epoch 50 Batch 90/173] avg loss 3.21435e-05, throughput 3.89809K wps
[Epoch 50 Batch 120/173] avg loss 2.30933e-05, throughput 3.89192K wps
[Epoch 50 Batch 150/173] avg loss 2.21697e-05, throughput 3.88052K wps
Begin Testing...
[Epoch 50] train avg loss 2.76019e-05, test acc 0.7635, test avg loss 1.11248, throughput 3.90655K wps
[Epoch 51 Batch 30/173] avg loss 1.9019e-05, throughput 3.96808K wps
[Epoch 51 Batch 60/173] avg loss 2.14228e-05, throughput 3.89116K wps
[Epoch 51 Batch 90/173] avg loss 1.81505e-05, throughput 3.89096K wps
[Epoch 51 Batch 120/173] avg loss 1.96969e-05, throughput 3.88625K wps
[Epoch 51 Batch 150/173] avg loss 1.84662e-05, throughput 3.89902K wps
Begin Testing...
[Epoch 51] train avg loss 1.95111e-05, test acc 0.7573, test avg loss 1.12834, throughput 3.90523K wps
[Epoch 52 Batch 30/173] avg loss 1.64828e-05, throughput 3.99607K wps
[Epoch 52 Batch 60/173] avg loss 1.9498e-05, throughput 3.88288K wps
[Epoch 52 Batch 90/173] avg loss 1.76043e-05, throughput 3.87202K wps
[Epoch 52 Batch 120/173] avg loss 2.67775e-05, throughput 3.89468K wps
[Epoch 52 Batch 150/173] avg loss 2.37484e-05, throughput 3.90273K wps
Begin Testing...
[Epoch 52] train avg loss 2.104e-05, test acc 0.7562, test avg loss 1.13925, throughput 3.90721K wps
[Epoch 53 Batch 30/173] avg loss 1.48248e-05, throughput 3.99579K wps
[Epoch 53 Batch 60/173] avg loss 2.50563e-05, throughput 3.87211K wps
[Epoch 53 Batch 90/173] avg loss 1.85525e-05, throughput 3.88254K wps
[Epoch 53 Batch 120/173] avg loss 2.09272e-05, throughput 3.87704K wps
[Epoch 53 Batch 150/173] avg loss 2.14152e-05, throughput 3.86389K wps
Begin Testing...
[Epoch 53] train avg loss 2.10734e-05, test acc 0.7625, test avg loss 1.16779, throughput 3.8982K wps
[Epoch 54 Batch 30/173] avg loss 1.6787e-05, throughput 4.00129K wps
[Epoch 54 Batch 60/173] avg loss 1.582e-05, throughput 3.89722K wps
[Epoch 54 Batch 90/173] avg loss 2.0524e-05, throughput 3.89135K wps
[Epoch 54 Batch 120/173] avg loss 1.1548e-05, throughput 3.8923K wps
[Epoch 54 Batch 150/173] avg loss 2.05387e-05, throughput 3.87502K wps
Begin Testing...
[Epoch 54] train avg loss 1.77232e-05, test acc 0.7604, test avg loss 1.18279, throughput 3.90519K wps
[Epoch 55 Batch 30/173] avg loss 2.99228e-05, throughput 3.97761K wps
[Epoch 55 Batch 60/173] avg loss 1.12034e-05, throughput 3.89025K wps
[Epoch 55 Batch 90/173] avg loss 2.79722e-05, throughput 3.88954K wps
[Epoch 55 Batch 120/173] avg loss 2.54763e-05, throughput 3.8902K wps
[Epoch 55 Batch 150/173] avg loss 1.9483e-05, throughput 3.89367K wps
Begin Testing...
[Epoch 55] train avg loss 2.25468e-05, test acc 0.7604, test avg loss 1.20604, throughput 3.90307K wps
[Epoch 56 Batch 30/173] avg loss 1.86962e-05, throughput 3.98422K wps
[Epoch 56 Batch 60/173] avg loss 1.34833e-05, throughput 3.87104K wps
[Epoch 56 Batch 90/173] avg loss 1.44636e-05, throughput 3.87926K wps
[Epoch 56 Batch 120/173] avg loss 1.33757e-05, throughput 3.90382K wps
[Epoch 56 Batch 150/173] avg loss 1.74747e-05, throughput 3.90593K wps
Begin Testing...
[Epoch 56] train avg loss 1.59004e-05, test acc 0.7604, test avg loss 1.22039, throughput 3.90573K wps
[Epoch 57 Batch 30/173] avg loss 9.4012e-06, throughput 3.9725K wps
[Epoch 57 Batch 60/173] avg loss 1.65523e-05, throughput 3.87736K wps
[Epoch 57 Batch 90/173] avg loss 1.08408e-05, throughput 3.89156K wps
[Epoch 57 Batch 120/173] avg loss 1.00544e-05, throughput 3.89392K wps
[Epoch 57 Batch 150/173] avg loss 8.63828e-06, throughput 3.88337K wps
Begin Testing...
[Epoch 57] train avg loss 1.12439e-05, test acc 0.7604, test avg loss 1.2484, throughput 3.90286K wps
[Epoch 58 Batch 30/173] avg loss 6.80945e-06, throughput 3.97954K wps
[Epoch 58 Batch 60/173] avg loss 9.43512e-06, throughput 3.90045K wps
[Epoch 58 Batch 90/173] avg loss 1.36013e-05, throughput 3.89944K wps
[Epoch 58 Batch 120/173] avg loss 8.66531e-06, throughput 3.88363K wps
[Epoch 58 Batch 150/173] avg loss 1.25638e-05, throughput 3.88077K wps
Begin Testing...
[Epoch 58] train avg loss 1.0622e-05, test acc 0.7656, test avg loss 1.25876, throughput 3.90804K wps
[Epoch 59 Batch 30/173] avg loss 1.04671e-05, throughput 3.96722K wps
[Epoch 59 Batch 60/173] avg loss 1.33414e-05, throughput 3.90709K wps
[Epoch 59 Batch 90/173] avg loss 1.58837e-05, throughput 3.87642K wps
[Epoch 59 Batch 120/173] avg loss 8.82605e-06, throughput 3.89163K wps
[Epoch 59 Batch 150/173] avg loss 1.26599e-05, throughput 3.89036K wps
Begin Testing...
[Epoch 59] train avg loss 1.33632e-05, test acc 0.7562, test avg loss 1.27836, throughput 3.90596K wps
Test loss 0.438021, test acc 0.7974
Total time cost 554.02s
9596 1066
[Epoch 0 Batch 30/173] avg loss 0.0155134, throughput 3.67627K wps
[Epoch 0 Batch 60/173] avg loss 0.0147817, throughput 3.87368K wps
[Epoch 0 Batch 90/173] avg loss 0.0150212, throughput 3.87404K wps
[Epoch 0 Batch 120/173] avg loss 0.0142751, throughput 3.86877K wps
[Epoch 0 Batch 150/173] avg loss 0.014094, throughput 3.89619K wps
Begin Testing...
[Epoch 0] train avg loss 0.0146065, test acc 0.6000, test avg loss 0.65487, throughput 3.84178K wps
Observed Improvement.
Begin Testing...
[Epoch 1 Batch 30/173] avg loss 0.0133359, throughput 3.98129K wps
[Epoch 1 Batch 60/173] avg loss 0.0130978, throughput 3.9034K wps
[Epoch 1 Batch 90/173] avg loss 0.0130544, throughput 3.88169K wps
[Epoch 1 Batch 120/173] avg loss 0.0129425, throughput 3.86397K wps
[Epoch 1 Batch 150/173] avg loss 0.0124174, throughput 3.88491K wps
Begin Testing...
[Epoch 1] train avg loss 0.0129431, test acc 0.6740, test avg loss 0.621142, throughput 3.90161K wps
Observed Improvement.
Begin Testing...
[Epoch 2 Batch 30/173] avg loss 0.0121822, throughput 3.97865K wps
[Epoch 2 Batch 60/173] avg loss 0.0120134, throughput 3.88805K wps
[Epoch 2 Batch 90/173] avg loss 0.0121212, throughput 3.88593K wps
[Epoch 2 Batch 120/173] avg loss 0.0122253, throughput 3.91329K wps
[Epoch 2 Batch 150/173] avg loss 0.0120123, throughput 3.89032K wps
Begin Testing...
[Epoch 2] train avg loss 0.012071, test acc 0.6885, test avg loss 0.596724, throughput 3.90544K wps
Observed Improvement.
Begin Testing...
[Epoch 3 Batch 30/173] avg loss 0.0113668, throughput 3.98904K wps
[Epoch 3 Batch 60/173] avg loss 0.0110902, throughput 3.87326K wps
[Epoch 3 Batch 90/173] avg loss 0.0113349, throughput 3.87044K wps
[Epoch 3 Batch 120/173] avg loss 0.010974, throughput 3.88629K wps
[Epoch 3 Batch 150/173] avg loss 0.0110425, throughput 3.91622K wps
Begin Testing...
[Epoch 3] train avg loss 0.0111763, test acc 0.7167, test avg loss 0.56542, throughput 3.90328K wps
Observed Improvement.
Begin Testing...
[Epoch 4 Batch 30/173] avg loss 0.0104056, throughput 3.99045K wps
[Epoch 4 Batch 60/173] avg loss 0.0100325, throughput 3.88152K wps
[Epoch 4 Batch 90/173] avg loss 0.0100761, throughput 3.87005K wps
[Epoch 4 Batch 120/173] avg loss 0.0102187, throughput 3.86224K wps
[Epoch 4 Batch 150/173] avg loss 0.0101455, throughput 3.86685K wps
Begin Testing...
[Epoch 4] train avg loss 0.0101325, test acc 0.7625, test avg loss 0.520497, throughput 3.89317K wps
Observed Improvement.
Begin Testing...
[Epoch 5 Batch 30/173] avg loss 0.00943613, throughput 3.98056K wps
[Epoch 5 Batch 60/173] avg loss 0.00904824, throughput 3.88469K wps
[Epoch 5 Batch 90/173] avg loss 0.00924876, throughput 3.90421K wps
[Epoch 5 Batch 120/173] avg loss 0.0089553, throughput 3.87865K wps
[Epoch 5 Batch 150/173] avg loss 0.00862165, throughput 3.87063K wps
Begin Testing...
[Epoch 5] train avg loss 0.00905175, test acc 0.7781, test avg loss 0.49088, throughput 3.89917K wps
Observed Improvement.
Begin Testing...
[Epoch 6 Batch 30/173] avg loss 0.00840391, throughput 3.97561K wps
[Epoch 6 Batch 60/173] avg loss 0.00785643, throughput 3.86425K wps
[Epoch 6 Batch 90/173] avg loss 0.00789745, throughput 3.88468K wps
[Epoch 6 Batch 120/173] avg loss 0.00822979, throughput 3.88679K wps
[Epoch 6 Batch 150/173] avg loss 0.00769824, throughput 3.91354K wps
Begin Testing...
[Epoch 6] train avg loss 0.00801901, test acc 0.7792, test avg loss 0.464793, throughput 3.9015K wps
Observed Improvement.
Begin Testing...
[Epoch 7 Batch 30/173] avg loss 0.00697853, throughput 3.99572K wps
[Epoch 7 Batch 60/173] avg loss 0.00738756, throughput 3.86967K wps
[Epoch 7 Batch 90/173] avg loss 0.00712462, throughput 3.87376K wps
[Epoch 7 Batch 120/173] avg loss 0.00700949, throughput 3.87756K wps
[Epoch 7 Batch 150/173] avg loss 0.0070252, throughput 3.86548K wps
Begin Testing...
[Epoch 7] train avg loss 0.00709586, test acc 0.7958, test avg loss 0.447366, throughput 3.89605K wps
Observed Improvement.
Begin Testing...
[Epoch 8 Batch 30/173] avg loss 0.00617178, throughput 3.98098K wps
[Epoch 8 Batch 60/173] avg loss 0.00648375, throughput 3.89859K wps
[Epoch 8 Batch 90/173] avg loss 0.00598955, throughput 3.90033K wps
[Epoch 8 Batch 120/173] avg loss 0.00649522, throughput 3.87876K wps
[Epoch 8 Batch 150/173] avg loss 0.00618435, throughput 3.88051K wps
Begin Testing...
[Epoch 8] train avg loss 0.00624415, test acc 0.8063, test avg loss 0.436377, throughput 3.90647K wps
Observed Improvement.
Begin Testing...
[Epoch 9 Batch 30/173] avg loss 0.00551228, throughput 3.97106K wps
[Epoch 9 Batch 60/173] avg loss 0.00558511, throughput 3.8731K wps
[Epoch 9 Batch 90/173] avg loss 0.00561362, throughput 3.88055K wps
[Epoch 9 Batch 120/173] avg loss 0.00538621, throughput 3.9077K wps
[Epoch 9 Batch 150/173] avg loss 0.0051747, throughput 3.88977K wps
Begin Testing...
[Epoch 9] train avg loss 0.00543002, test acc 0.8052, test avg loss 0.426271, throughput 3.90019K wps
[Epoch 10 Batch 30/173] avg loss 0.00463374, throughput 3.97203K wps
[Epoch 10 Batch 60/173] avg loss 0.0046193, throughput 3.87492K wps
[Epoch 10 Batch 90/173] avg loss 0.00477147, throughput 3.87639K wps
[Epoch 10 Batch 120/173] avg loss 0.00518452, throughput 3.88837K wps
[Epoch 10 Batch 150/173] avg loss 0.0045618, throughput 3.86358K wps
Begin Testing...
[Epoch 10] train avg loss 0.00472282, test acc 0.8146, test avg loss 0.426103, throughput 3.89177K wps
Observed Improvement.
Begin Testing...
[Epoch 11 Batch 30/173] avg loss 0.00397196, throughput 3.99346K wps
[Epoch 11 Batch 60/173] avg loss 0.00391925, throughput 3.90676K wps
[Epoch 11 Batch 90/173] avg loss 0.00401951, throughput 3.88779K wps
[Epoch 11 Batch 120/173] avg loss 0.00396469, throughput 3.88743K wps
[Epoch 11 Batch 150/173] avg loss 0.00403233, throughput 3.88311K wps
Begin Testing...
[Epoch 11] train avg loss 0.00401627, test acc 0.8167, test avg loss 0.425282, throughput 3.90876K wps
Observed Improvement.
Begin Testing...
[Epoch 12 Batch 30/173] avg loss 0.00347133, throughput 3.9779K wps
[Epoch 12 Batch 60/173] avg loss 0.00343408, throughput 3.90357K wps
[Epoch 12 Batch 90/173] avg loss 0.00344618, throughput 3.88313K wps
[Epoch 12 Batch 120/173] avg loss 0.00335553, throughput 3.8942K wps
[Epoch 12 Batch 150/173] avg loss 0.00344988, throughput 3.91142K wps
Begin Testing...
[Epoch 12] train avg loss 0.00344273, test acc 0.8198, test avg loss 0.427278, throughput 3.91075K wps
Observed Improvement.
Begin Testing...
[Epoch 13 Batch 30/173] avg loss 0.00283039, throughput 3.96482K wps
[Epoch 13 Batch 60/173] avg loss 0.00296621, throughput 3.89373K wps
[Epoch 13 Batch 90/173] avg loss 0.00292767, throughput 3.88099K wps
[Epoch 13 Batch 120/173] avg loss 0.00302711, throughput 3.8805K wps
[Epoch 13 Batch 150/173] avg loss 0.00310852, throughput 3.87748K wps
Begin Testing...
[Epoch 13] train avg loss 0.00296789, test acc 0.8135, test avg loss 0.433129, throughput 3.89404K wps
[Epoch 14 Batch 30/173] avg loss 0.00258808, throughput 3.97755K wps
[Epoch 14 Batch 60/173] avg loss 0.00241513, throughput 3.89108K wps
[Epoch 14 Batch 90/173] avg loss 0.00259385, throughput 3.9089K wps
[Epoch 14 Batch 120/173] avg loss 0.00241995, throughput 3.88323K wps
[Epoch 14 Batch 150/173] avg loss 0.00249623, throughput 3.8809K wps
Begin Testing...
[Epoch 14] train avg loss 0.00251608, test acc 0.8042, test avg loss 0.458477, throughput 3.90707K wps
[Epoch 15 Batch 30/173] avg loss 0.00201742, throughput 3.96208K wps
[Epoch 15 Batch 60/173] avg loss 0.00202722, throughput 3.87434K wps
[Epoch 15 Batch 90/173] avg loss 0.0021536, throughput 3.88189K wps
[Epoch 15 Batch 120/173] avg loss 0.00210351, throughput 3.87649K wps
[Epoch 15 Batch 150/173] avg loss 0.00209455, throughput 3.88338K wps
Begin Testing...
[Epoch 15] train avg loss 0.00208852, test acc 0.8146, test avg loss 0.4512, throughput 3.89448K wps
[Epoch 16 Batch 30/173] avg loss 0.00166808, throughput 3.97513K wps
[Epoch 16 Batch 60/173] avg loss 0.00175605, throughput 3.89232K wps
[Epoch 16 Batch 90/173] avg loss 0.00181174, throughput 3.91119K wps
[Epoch 16 Batch 120/173] avg loss 0.0019583, throughput 3.8907K wps
[Epoch 16 Batch 150/173] avg loss 0.00177418, throughput 3.8796K wps
Begin Testing...
[Epoch 16] train avg loss 0.00176197, test acc 0.8083, test avg loss 0.460675, throughput 3.90874K wps
[Epoch 17 Batch 30/173] avg loss 0.00146577, throughput 3.96762K wps
[Epoch 17 Batch 60/173] avg loss 0.0015795, throughput 3.89334K wps
[Epoch 17 Batch 90/173] avg loss 0.00155403, throughput 3.88863K wps
[Epoch 17 Batch 120/173] avg loss 0.0015124, throughput 3.88801K wps
[Epoch 17 Batch 150/173] avg loss 0.00138666, throughput 3.88145K wps
Begin Testing...
[Epoch 17] train avg loss 0.001513, test acc 0.8146, test avg loss 0.47669, throughput 3.90691K wps
[Epoch 18 Batch 30/173] avg loss 0.00124261, throughput 3.99044K wps
[Epoch 18 Batch 60/173] avg loss 0.00135098, throughput 3.87827K wps
[Epoch 18 Batch 90/173] avg loss 0.00141886, throughput 3.88186K wps
[Epoch 18 Batch 120/173] avg loss 0.00118941, throughput 3.87985K wps
[Epoch 18 Batch 150/173] avg loss 0.00113899, throughput 3.8955K wps
Begin Testing...
[Epoch 18] train avg loss 0.00126824, test acc 0.8125, test avg loss 0.495251, throughput 3.90328K wps
[Epoch 19 Batch 30/173] avg loss 0.000965848, throughput 3.99105K wps
[Epoch 19 Batch 60/173] avg loss 0.00107506, throughput 3.88691K wps
[Epoch 19 Batch 90/173] avg loss 0.00108537, throughput 3.89737K wps
[Epoch 19 Batch 120/173] avg loss 0.00100975, throughput 3.89045K wps
[Epoch 19 Batch 150/173] avg loss 0.00107305, throughput 3.87535K wps
Begin Testing...
[Epoch 19] train avg loss 0.00106398, test acc 0.8063, test avg loss 0.51107, throughput 3.9025K wps
[Epoch 20 Batch 30/173] avg loss 0.000947131, throughput 3.97881K wps
[Epoch 20 Batch 60/173] avg loss 0.000946625, throughput 3.87859K wps
[Epoch 20 Batch 90/173] avg loss 0.000931394, throughput 3.8758K wps
[Epoch 20 Batch 120/173] avg loss 0.000885043, throughput 3.88291K wps
[Epoch 20 Batch 150/173] avg loss 0.000852493, throughput 3.90484K wps
Begin Testing...
[Epoch 20] train avg loss 0.000908805, test acc 0.8021, test avg loss 0.528724, throughput 3.89968K wps
[Epoch 21 Batch 30/173] avg loss 0.000771165, throughput 3.9836K wps
[Epoch 21 Batch 60/173] avg loss 0.000785783, throughput 3.87127K wps
[Epoch 21 Batch 90/173] avg loss 0.00076395, throughput 3.88252K wps
[Epoch 21 Batch 120/173] avg loss 0.000784913, throughput 3.90406K wps
[Epoch 21 Batch 150/173] avg loss 0.000752576, throughput 3.9021K wps
Begin Testing...
[Epoch 21] train avg loss 0.000760967, test acc 0.8010, test avg loss 0.550986, throughput 3.90419K wps
[Epoch 22 Batch 30/173] avg loss 0.000682826, throughput 3.96423K wps
[Epoch 22 Batch 60/173] avg loss 0.000628767, throughput 3.87953K wps
[Epoch 22 Batch 90/173] avg loss 0.000625741, throughput 3.87569K wps
[Epoch 22 Batch 120/173] avg loss 0.000647122, throughput 3.88335K wps
[Epoch 22 Batch 150/173] avg loss 0.000640158, throughput 3.90834K wps
Begin Testing...
[Epoch 22] train avg loss 0.000646897, test acc 0.7948, test avg loss 0.564109, throughput 3.89963K wps
[Epoch 23 Batch 30/173] avg loss 0.00055339, throughput 3.97647K wps
[Epoch 23 Batch 60/173] avg loss 0.000535843, throughput 3.89589K wps
[Epoch 23 Batch 90/173] avg loss 0.000545959, throughput 3.88459K wps
[Epoch 23 Batch 120/173] avg loss 0.000546871, throughput 3.89403K wps
[Epoch 23 Batch 150/173] avg loss 0.000638461, throughput 3.87246K wps
Begin Testing...
[Epoch 23] train avg loss 0.000552191, test acc 0.7979, test avg loss 0.587954, throughput 3.89945K wps
[Epoch 24 Batch 30/173] avg loss 0.000466831, throughput 3.98596K wps
[Epoch 24 Batch 60/173] avg loss 0.000474101, throughput 3.89213K wps
[Epoch 24 Batch 90/173] avg loss 0.000454228, throughput 3.91061K wps
[Epoch 24 Batch 120/173] avg loss 0.000473656, throughput 3.89548K wps
[Epoch 24 Batch 150/173] avg loss 0.000501605, throughput 3.87521K wps
Begin Testing...
[Epoch 24] train avg loss 0.000475473, test acc 0.7896, test avg loss 0.600058, throughput 3.91073K wps
[Epoch 25 Batch 30/173] avg loss 0.000434042, throughput 3.97331K wps
[Epoch 25 Batch 60/173] avg loss 0.000380353, throughput 3.86019K wps
[Epoch 25 Batch 90/173] avg loss 0.000397812, throughput 3.91483K wps
[Epoch 25 Batch 120/173] avg loss 0.000413652, throughput 3.89215K wps
[Epoch 25 Batch 150/173] avg loss 0.000417811, throughput 3.88736K wps
Begin Testing...
[Epoch 25] train avg loss 0.000406388, test acc 0.7990, test avg loss 0.628346, throughput 3.90529K wps
[Epoch 26 Batch 30/173] avg loss 0.000341491, throughput 3.97374K wps
[Epoch 26 Batch 60/173] avg loss 0.00033511, throughput 3.88673K wps
[Epoch 26 Batch 90/173] avg loss 0.000318127, throughput 3.87888K wps
[Epoch 26 Batch 120/173] avg loss 0.000347006, throughput 3.86713K wps
[Epoch 26 Batch 150/173] avg loss 0.000378363, throughput 3.88388K wps
Begin Testing...
[Epoch 26] train avg loss 0.000343168, test acc 0.7896, test avg loss 0.642688, throughput 3.90009K wps
[Epoch 27 Batch 30/173] avg loss 0.000282628, throughput 4.00405K wps
[Epoch 27 Batch 60/173] avg loss 0.000287377, throughput 3.89823K wps
[Epoch 27 Batch 90/173] avg loss 0.000361136, throughput 3.87042K wps
[Epoch 27 Batch 120/173] avg loss 0.000263819, throughput 3.8757K wps
[Epoch 27 Batch 150/173] avg loss 0.000313195, throughput 3.88044K wps
Begin Testing...
[Epoch 27] train avg loss 0.000295374, test acc 0.7948, test avg loss 0.662693, throughput 3.90365K wps
[Epoch 28 Batch 30/173] avg loss 0.000243535, throughput 3.998K wps
[Epoch 28 Batch 60/173] avg loss 0.000254615, throughput 3.90285K wps
[Epoch 28 Batch 90/173] avg loss 0.000243858, throughput 3.90021K wps
[Epoch 28 Batch 120/173] avg loss 0.000256622, throughput 3.91201K wps
[Epoch 28 Batch 150/173] avg loss 0.000265946, throughput 3.88627K wps
Begin Testing...
[Epoch 28] train avg loss 0.000247989, test acc 0.7906, test avg loss 0.681206, throughput 3.9121K wps
[Epoch 29 Batch 30/173] avg loss 0.000214285, throughput 3.96949K wps
[Epoch 29 Batch 60/173] avg loss 0.000225572, throughput 3.87607K wps
[Epoch 29 Batch 90/173] avg loss 0.000207406, throughput 3.88252K wps
[Epoch 29 Batch 120/173] avg loss 0.00021033, throughput 3.91927K wps
[Epoch 29 Batch 150/173] avg loss 0.000185652, throughput 3.90227K wps
Begin Testing...
[Epoch 29] train avg loss 0.000214136, test acc 0.7896, test avg loss 0.706334, throughput 3.90964K wps
[Epoch 30 Batch 30/173] avg loss 0.00024679, throughput 3.99947K wps
[Epoch 30 Batch 60/173] avg loss 0.000222263, throughput 3.89466K wps
[Epoch 30 Batch 90/173] avg loss 0.000211634, throughput 3.88313K wps
[Epoch 30 Batch 120/173] avg loss 0.000221111, throughput 3.87769K wps
[Epoch 30 Batch 150/173] avg loss 0.000203998, throughput 3.88178K wps
Begin Testing...
[Epoch 30] train avg loss 0.000213507, test acc 0.7875, test avg loss 0.717607, throughput 3.90639K wps
[Epoch 31 Batch 30/173] avg loss 0.000167363, throughput 3.99143K wps
[Epoch 31 Batch 60/173] avg loss 0.000189846, throughput 3.89356K wps
[Epoch 31 Batch 90/173] avg loss 0.00020149, throughput 3.88735K wps
[Epoch 31 Batch 120/173] avg loss 0.000160556, throughput 3.8897K wps
[Epoch 31 Batch 150/173] avg loss 0.000170993, throughput 3.89326K wps
Begin Testing...
[Epoch 31] train avg loss 0.000177184, test acc 0.7844, test avg loss 0.73596, throughput 3.90644K wps
[Epoch 32 Batch 30/173] avg loss 0.000167411, throughput 3.97501K wps
[Epoch 32 Batch 60/173] avg loss 0.000135003, throughput 3.86515K wps
[Epoch 32 Batch 90/173] avg loss 0.000186976, throughput 3.881K wps
[Epoch 32 Batch 120/173] avg loss 0.000151979, throughput 3.86842K wps
[Epoch 32 Batch 150/173] avg loss 0.000132329, throughput 3.88825K wps
Begin Testing...
[Epoch 32] train avg loss 0.000155122, test acc 0.7885, test avg loss 0.759169, throughput 3.89771K wps
[Epoch 33 Batch 30/173] avg loss 0.00011361, throughput 3.99589K wps
[Epoch 33 Batch 60/173] avg loss 0.000108027, throughput 3.8655K wps
[Epoch 33 Batch 90/173] avg loss 0.000150631, throughput 3.87959K wps
[Epoch 33 Batch 120/173] avg loss 0.000130631, throughput 3.87997K wps
[Epoch 33 Batch 150/173] avg loss 0.000141305, throughput 3.87726K wps
Begin Testing...
[Epoch 33] train avg loss 0.000129356, test acc 0.7833, test avg loss 0.776488, throughput 3.89968K wps
[Epoch 34 Batch 30/173] avg loss 0.000104325, throughput 4.00317K wps
[Epoch 34 Batch 60/173] avg loss 0.000121498, throughput 3.89402K wps
[Epoch 34 Batch 90/173] avg loss 0.000109907, throughput 3.88922K wps
[Epoch 34 Batch 120/173] avg loss 0.00011164, throughput 3.91872K wps
[Epoch 34 Batch 150/173] avg loss 9.45713e-05, throughput 3.87425K wps
Begin Testing...
[Epoch 34] train avg loss 0.000106967, test acc 0.7833, test avg loss 0.79704, throughput 3.90989K wps
[Epoch 35 Batch 30/173] avg loss 8.94282e-05, throughput 3.97548K wps
[Epoch 35 Batch 60/173] avg loss 9.24071e-05, throughput 3.88193K wps
[Epoch 35 Batch 90/173] avg loss 0.000113211, throughput 3.88528K wps
[Epoch 35 Batch 120/173] avg loss 0.000101385, throughput 3.90979K wps
[Epoch 35 Batch 150/173] avg loss 0.00010022, throughput 3.87894K wps
Begin Testing...
[Epoch 35] train avg loss 0.000102337, test acc 0.7865, test avg loss 0.821812, throughput 3.90321K wps
[Epoch 36 Batch 30/173] avg loss 8.94634e-05, throughput 3.97882K wps
[Epoch 36 Batch 60/173] avg loss 8.24343e-05, throughput 3.87923K wps
[Epoch 36 Batch 90/173] avg loss 9.20779e-05, throughput 3.88395K wps
[Epoch 36 Batch 120/173] avg loss 0.000103173, throughput 3.88539K wps
[Epoch 36 Batch 150/173] avg loss 7.95127e-05, throughput 3.8769K wps
Begin Testing...
[Epoch 36] train avg loss 8.84795e-05, test acc 0.7833, test avg loss 0.83664, throughput 3.89606K wps
[Epoch 37 Batch 30/173] avg loss 8.77481e-05, throughput 3.98921K wps
[Epoch 37 Batch 60/173] avg loss 9.0499e-05, throughput 3.8852K wps
[Epoch 37 Batch 90/173] avg loss 7.79187e-05, throughput 3.90179K wps
[Epoch 37 Batch 120/173] avg loss 8.82546e-05, throughput 3.8931K wps
[Epoch 37 Batch 150/173] avg loss 8.86691e-05, throughput 3.87692K wps
Begin Testing...
[Epoch 37] train avg loss 8.64024e-05, test acc 0.7802, test avg loss 0.875847, throughput 3.90356K wps
[Epoch 38 Batch 30/173] avg loss 9.34063e-05, throughput 3.96636K wps
[Epoch 38 Batch 60/173] avg loss 8.21637e-05, throughput 3.87932K wps
[Epoch 38 Batch 90/173] avg loss 8.80992e-05, throughput 3.8654K wps
[Epoch 38 Batch 120/173] avg loss 7.77467e-05, throughput 3.89935K wps
[Epoch 38 Batch 150/173] avg loss 6.71774e-05, throughput 3.90167K wps
Begin Testing...
[Epoch 38] train avg loss 7.99469e-05, test acc 0.7792, test avg loss 0.879413, throughput 3.89893K wps
[Epoch 39 Batch 30/173] avg loss 6.51345e-05, throughput 3.97315K wps
[Epoch 39 Batch 60/173] avg loss 7.18474e-05, throughput 3.87596K wps
[Epoch 39 Batch 90/173] avg loss 7.51838e-05, throughput 3.88537K wps
[Epoch 39 Batch 120/173] avg loss 5.99156e-05, throughput 3.8911K wps
[Epoch 39 Batch 150/173] avg loss 5.06905e-05, throughput 3.8767K wps
Begin Testing...
[Epoch 39] train avg loss 6.55727e-05, test acc 0.7812, test avg loss 0.900666, throughput 3.89547K wps
[Epoch 40 Batch 30/173] avg loss 6.27998e-05, throughput 3.99522K wps
[Epoch 40 Batch 60/173] avg loss 5.85932e-05, throughput 3.89023K wps
[Epoch 40 Batch 90/173] avg loss 6.60988e-05, throughput 3.88581K wps
[Epoch 40 Batch 120/173] avg loss 4.50119e-05, throughput 3.90909K wps
[Epoch 40 Batch 150/173] avg loss 5.29554e-05, throughput 3.88364K wps
Begin Testing...
[Epoch 40] train avg loss 5.73592e-05, test acc 0.7802, test avg loss 0.918783, throughput 3.90814K wps
[Epoch 41 Batch 30/173] avg loss 4.53647e-05, throughput 3.97135K wps
[Epoch 41 Batch 60/173] avg loss 4.42629e-05, throughput 3.88675K wps
[Epoch 41 Batch 90/173] avg loss 4.74156e-05, throughput 3.87877K wps
[Epoch 41 Batch 120/173] avg loss 5.494e-05, throughput 3.8969K wps
[Epoch 41 Batch 150/173] avg loss 6.45954e-05, throughput 3.89628K wps
Begin Testing...
[Epoch 41] train avg loss 5.82616e-05, test acc 0.7760, test avg loss 0.923738, throughput 3.9064K wps
[Epoch 42 Batch 30/173] avg loss 4.43433e-05, throughput 3.98312K wps
[Epoch 42 Batch 60/173] avg loss 6.04271e-05, throughput 3.91163K wps
[Epoch 42 Batch 90/173] avg loss 4.68133e-05, throughput 3.86775K wps
[Epoch 42 Batch 120/173] avg loss 6.39734e-05, throughput 3.86769K wps
[Epoch 42 Batch 150/173] avg loss 5.64727e-05, throughput 3.88317K wps
Begin Testing...
[Epoch 42] train avg loss 5.64705e-05, test acc 0.7750, test avg loss 0.955318, throughput 3.90149K wps
[Epoch 43 Batch 30/173] avg loss 4.16396e-05, throughput 3.97528K wps
[Epoch 43 Batch 60/173] avg loss 3.89174e-05, throughput 3.90754K wps
[Epoch 43 Batch 90/173] avg loss 5.69525e-05, throughput 3.89393K wps
[Epoch 43 Batch 120/173] avg loss 3.80485e-05, throughput 3.8991K wps
[Epoch 43 Batch 150/173] avg loss 4.79827e-05, throughput 3.91995K wps
Begin Testing...
[Epoch 43] train avg loss 4.47471e-05, test acc 0.7781, test avg loss 0.977197, throughput 3.91368K wps
[Epoch 44 Batch 30/173] avg loss 3.28879e-05, throughput 3.97563K wps
[Epoch 44 Batch 60/173] avg loss 3.31473e-05, throughput 3.88335K wps
[Epoch 44 Batch 90/173] avg loss 4.13096e-05, throughput 3.87807K wps
[Epoch 44 Batch 120/173] avg loss 3.62202e-05, throughput 3.88762K wps
[Epoch 44 Batch 150/173] avg loss 4.70755e-05, throughput 3.88502K wps
Begin Testing...
[Epoch 44] train avg loss 3.86298e-05, test acc 0.7771, test avg loss 1.00088, throughput 3.90028K wps
[Epoch 45 Batch 30/173] avg loss 2.96376e-05, throughput 3.99163K wps
[Epoch 45 Batch 60/173] avg loss 3.44745e-05, throughput 3.89001K wps
[Epoch 45 Batch 90/173] avg loss 2.94269e-05, throughput 3.91416K wps
[Epoch 45 Batch 120/173] avg loss 2.8489e-05, throughput 3.87907K wps
[Epoch 45 Batch 150/173] avg loss 3.31407e-05, throughput 3.87333K wps
Begin Testing...
[Epoch 45] train avg loss 3.22309e-05, test acc 0.7812, test avg loss 1.02252, throughput 3.90591K wps
[Epoch 46 Batch 30/173] avg loss 2.44642e-05, throughput 3.97738K wps
[Epoch 46 Batch 60/173] avg loss 5.3033e-05, throughput 3.88394K wps
[Epoch 46 Batch 90/173] avg loss 4.68049e-05, throughput 3.89633K wps
[Epoch 46 Batch 120/173] avg loss 2.55881e-05, throughput 3.88904K wps
[Epoch 46 Batch 150/173] avg loss 4.07536e-05, throughput 3.88972K wps
Begin Testing...
[Epoch 46] train avg loss 3.72731e-05, test acc 0.7823, test avg loss 1.03589, throughput 3.90703K wps
[Epoch 47 Batch 30/173] avg loss 3.21365e-05, throughput 4.00245K wps
[Epoch 47 Batch 60/173] avg loss 3.07471e-05, throughput 3.87818K wps
[Epoch 47 Batch 90/173] avg loss 3.48042e-05, throughput 3.87684K wps
[Epoch 47 Batch 120/173] avg loss 3.15156e-05, throughput 3.87361K wps
[Epoch 47 Batch 150/173] avg loss 3.53329e-05, throughput 3.8792K wps
Begin Testing...
[Epoch 47] train avg loss 3.21759e-05, test acc 0.7719, test avg loss 1.05566, throughput 3.89919K wps
[Epoch 48 Batch 30/173] avg loss 3.85133e-05, throughput 3.97642K wps
[Epoch 48 Batch 60/173] avg loss 3.23953e-05, throughput 3.89737K wps
[Epoch 48 Batch 90/173] avg loss 2.35755e-05, throughput 3.90101K wps
[Epoch 48 Batch 120/173] avg loss 2.48903e-05, throughput 3.87937K wps
[Epoch 48 Batch 150/173] avg loss 2.29628e-05, throughput 3.88843K wps
Begin Testing...
[Epoch 48] train avg loss 2.80259e-05, test acc 0.7781, test avg loss 1.07117, throughput 3.90819K wps
[Epoch 49 Batch 30/173] avg loss 1.80563e-05, throughput 3.9977K wps
[Epoch 49 Batch 60/173] avg loss 2.68581e-05, throughput 3.88502K wps
[Epoch 49 Batch 90/173] avg loss 2.38464e-05, throughput 3.88084K wps
[Epoch 49 Batch 120/173] avg loss 2.87435e-05, throughput 3.87088K wps
[Epoch 49 Batch 150/173] avg loss 2.15965e-05, throughput 3.91493K wps
Begin Testing...
[Epoch 49] train avg loss 2.52661e-05, test acc 0.7760, test avg loss 1.10853, throughput 3.90544K wps
[Epoch 50 Batch 30/173] avg loss 2.99919e-05, throughput 3.98526K wps
[Epoch 50 Batch 60/173] avg loss 2.403e-05, throughput 3.89107K wps
[Epoch 50 Batch 90/173] avg loss 2.98253e-05, throughput 3.9009K wps
[Epoch 50 Batch 120/173] avg loss 1.87839e-05, throughput 3.88557K wps
[Epoch 50 Batch 150/173] avg loss 2.1373e-05, throughput 3.87939K wps
Begin Testing...
[Epoch 50] train avg loss 2.40678e-05, test acc 0.7812, test avg loss 1.10212, throughput 3.9045K wps
[Epoch 51 Batch 30/173] avg loss 1.78666e-05, throughput 3.98091K wps
[Epoch 51 Batch 60/173] avg loss 2.12292e-05, throughput 3.90969K wps
[Epoch 51 Batch 90/173] avg loss 2.71472e-05, throughput 3.9112K wps
[Epoch 51 Batch 120/173] avg loss 1.88155e-05, throughput 3.88727K wps
[Epoch 51 Batch 150/173] avg loss 2.55113e-05, throughput 3.88374K wps
Begin Testing...
[Epoch 51] train avg loss 2.09933e-05, test acc 0.7750, test avg loss 1.11887, throughput 3.9129K wps
[Epoch 52 Batch 30/173] avg loss 2.12623e-05, throughput 3.95863K wps
[Epoch 52 Batch 60/173] avg loss 2.05323e-05, throughput 3.8919K wps
[Epoch 52 Batch 90/173] avg loss 1.90979e-05, throughput 3.89164K wps
[Epoch 52 Batch 120/173] avg loss 1.52976e-05, throughput 3.88907K wps
[Epoch 52 Batch 150/173] avg loss 1.34268e-05, throughput 3.9089K wps
Begin Testing...
[Epoch 52] train avg loss 1.93173e-05, test acc 0.7750, test avg loss 1.13792, throughput 3.90527K wps
[Epoch 53 Batch 30/173] avg loss 1.23169e-05, throughput 3.97779K wps
[Epoch 53 Batch 60/173] avg loss 2.20246e-05, throughput 3.88188K wps
[Epoch 53 Batch 90/173] avg loss 1.71269e-05, throughput 3.8721K wps
[Epoch 53 Batch 120/173] avg loss 1.49035e-05, throughput 3.88799K wps
[Epoch 53 Batch 150/173] avg loss 1.70601e-05, throughput 3.90397K wps
Begin Testing...
[Epoch 53] train avg loss 1.74821e-05, test acc 0.7698, test avg loss 1.17226, throughput 3.90391K wps
[Epoch 54 Batch 30/173] avg loss 1.20855e-05, throughput 3.97854K wps
[Epoch 54 Batch 60/173] avg loss 1.33986e-05, throughput 3.87747K wps
[Epoch 54 Batch 90/173] avg loss 1.99391e-05, throughput 3.88645K wps
[Epoch 54 Batch 120/173] avg loss 1.35912e-05, throughput 3.88613K wps
[Epoch 54 Batch 150/173] avg loss 1.83934e-05, throughput 3.90906K wps
Begin Testing...
[Epoch 54] train avg loss 1.50102e-05, test acc 0.7708, test avg loss 1.19255, throughput 3.90456K wps
[Epoch 55 Batch 30/173] avg loss 1.07679e-05, throughput 3.99249K wps
[Epoch 55 Batch 60/173] avg loss 1.3758e-05, throughput 3.88123K wps
[Epoch 55 Batch 90/173] avg loss 1.06584e-05, throughput 3.89163K wps
[Epoch 55 Batch 120/173] avg loss 1.08556e-05, throughput 3.87698K wps
[Epoch 55 Batch 150/173] avg loss 1.97125e-05, throughput 3.87694K wps
Begin Testing...
[Epoch 55] train avg loss 1.32566e-05, test acc 0.7729, test avg loss 1.20141, throughput 3.898K wps
[Epoch 56 Batch 30/173] avg loss 1.52807e-05, throughput 3.99692K wps
[Epoch 56 Batch 60/173] avg loss 8.3517e-06, throughput 3.89135K wps
[Epoch 56 Batch 90/173] avg loss 1.43614e-05, throughput 3.89281K wps
[Epoch 56 Batch 120/173] avg loss 1.35144e-05, throughput 3.91485K wps
[Epoch 56 Batch 150/173] avg loss 8.03415e-06, throughput 3.87589K wps
Begin Testing...
[Epoch 56] train avg loss 1.19535e-05, test acc 0.7708, test avg loss 1.22627, throughput 3.90978K wps
[Epoch 57 Batch 30/173] avg loss 8.16505e-06, throughput 3.9712K wps
[Epoch 57 Batch 60/173] avg loss 1.16922e-05, throughput 3.87187K wps
[Epoch 57 Batch 90/173] avg loss 1.39594e-05, throughput 3.88378K wps
[Epoch 57 Batch 120/173] avg loss 2.15123e-05, throughput 3.90448K wps
[Epoch 57 Batch 150/173] avg loss 1.63571e-05, throughput 3.88928K wps
Begin Testing...
[Epoch 57] train avg loss 1.38009e-05, test acc 0.7740, test avg loss 1.2235, throughput 3.90481K wps
[Epoch 58 Batch 30/173] avg loss 1.44018e-05, throughput 3.98271K wps
[Epoch 58 Batch 60/173] avg loss 1.5314e-05, throughput 3.89803K wps
[Epoch 58 Batch 90/173] avg loss 1.30763e-05, throughput 3.86476K wps
[Epoch 58 Batch 120/173] avg loss 1.00294e-05, throughput 3.86646K wps
[Epoch 58 Batch 150/173] avg loss 1.10976e-05, throughput 3.87619K wps
Begin Testing...
[Epoch 58] train avg loss 1.26818e-05, test acc 0.7740, test avg loss 1.25272, throughput 3.89563K wps
[Epoch 59 Batch 30/173] avg loss 1.16344e-05, throughput 3.97625K wps
[Epoch 59 Batch 60/173] avg loss 9.70368e-06, throughput 3.89164K wps
[Epoch 59 Batch 90/173] avg loss 7.10263e-06, throughput 3.90936K wps
[Epoch 59 Batch 120/173] avg loss 1.55249e-05, throughput 3.87693K wps
[Epoch 59 Batch 150/173] avg loss 1.21197e-05, throughput 3.88054K wps
Begin Testing...
[Epoch 59] train avg loss 1.82007e-05, test acc 0.7583, test avg loss 1.36438, throughput 3.90657K wps
Test loss 0.414151, test acc 0.8189
Total time cost 554.60s
9596 1066
[Epoch 0 Batch 30/173] avg loss 0.0156862, throughput 3.6772K wps
[Epoch 0 Batch 60/173] avg loss 0.0146758, throughput 3.89371K wps
[Epoch 0 Batch 90/173] avg loss 0.0148351, throughput 3.87828K wps
[Epoch 0 Batch 120/173] avg loss 0.0140997, throughput 3.89155K wps
[Epoch 0 Batch 150/173] avg loss 0.0138287, throughput 3.90539K wps
Begin Testing...
[Epoch 0] train avg loss 0.0145885, test acc 0.6302, test avg loss 0.648904, throughput 3.8497K wps
Observed Improvement.
Begin Testing...
[Epoch 1 Batch 30/173] avg loss 0.0132985, throughput 3.96852K wps
[Epoch 1 Batch 60/173] avg loss 0.0128914, throughput 3.87524K wps
[Epoch 1 Batch 90/173] avg loss 0.0130141, throughput 3.8798K wps
[Epoch 1 Batch 120/173] avg loss 0.0129827, throughput 3.88697K wps
[Epoch 1 Batch 150/173] avg loss 0.0127937, throughput 3.89034K wps
Begin Testing...
[Epoch 1] train avg loss 0.0129828, test acc 0.6729, test avg loss 0.627709, throughput 3.89893K wps
Observed Improvement.
Begin Testing...
[Epoch 2 Batch 30/173] avg loss 0.0121859, throughput 3.98261K wps
[Epoch 2 Batch 60/173] avg loss 0.0121458, throughput 3.87473K wps
[Epoch 2 Batch 90/173] avg loss 0.0120112, throughput 3.8792K wps
[Epoch 2 Batch 120/173] avg loss 0.0118837, throughput 3.88627K wps
[Epoch 2 Batch 150/173] avg loss 0.0117783, throughput 3.89235K wps
Begin Testing...
[Epoch 2] train avg loss 0.0119833, test acc 0.7073, test avg loss 0.601463, throughput 3.89757K wps
Observed Improvement.
Begin Testing...
[Epoch 3 Batch 30/173] avg loss 0.0112655, throughput 3.97967K wps
[Epoch 3 Batch 60/173] avg loss 0.0112731, throughput 3.91463K wps
[Epoch 3 Batch 90/173] avg loss 0.0109321, throughput 3.88268K wps
[Epoch 3 Batch 120/173] avg loss 0.0109788, throughput 3.87758K wps
[Epoch 3 Batch 150/173] avg loss 0.0109336, throughput 3.87888K wps
Begin Testing...
[Epoch 3] train avg loss 0.0110459, test acc 0.7385, test avg loss 0.565029, throughput 3.90615K wps
Observed Improvement.
Begin Testing...
[Epoch 4 Batch 30/173] avg loss 0.0105335, throughput 3.96592K wps
[Epoch 4 Batch 60/173] avg loss 0.0100374, throughput 3.86898K wps
[Epoch 4 Batch 90/173] avg loss 0.0102404, throughput 3.88844K wps
[Epoch 4 Batch 120/173] avg loss 0.00981466, throughput 3.89421K wps
[Epoch 4 Batch 150/173] avg loss 0.0100216, throughput 3.88033K wps
Begin Testing...
[Epoch 4] train avg loss 0.0100635, test acc 0.7583, test avg loss 0.527169, throughput 3.89719K wps
Observed Improvement.
Begin Testing...
[Epoch 5 Batch 30/173] avg loss 0.00883343, throughput 3.98175K wps
[Epoch 5 Batch 60/173] avg loss 0.00938636, throughput 3.86812K wps
[Epoch 5 Batch 90/173] avg loss 0.00890139, throughput 3.88003K wps
[Epoch 5 Batch 120/173] avg loss 0.00860001, throughput 3.87412K wps
[Epoch 5 Batch 150/173] avg loss 0.00885604, throughput 3.86288K wps
Begin Testing...
[Epoch 5] train avg loss 0.008931, test acc 0.7812, test avg loss 0.490957, throughput 3.89505K wps
Observed Improvement.
Begin Testing...
[Epoch 6 Batch 30/173] avg loss 0.00814225, throughput 3.98274K wps
[Epoch 6 Batch 60/173] avg loss 0.0081958, throughput 3.92427K wps
[Epoch 6 Batch 90/173] avg loss 0.00773264, throughput 3.88196K wps
[Epoch 6 Batch 120/173] avg loss 0.00769855, throughput 3.88735K wps
[Epoch 6 Batch 150/173] avg loss 0.00767533, throughput 3.88175K wps
Begin Testing...
[Epoch 6] train avg loss 0.00788051, test acc 0.7885, test avg loss 0.461536, throughput 3.90944K wps
Observed Improvement.
Begin Testing...
[Epoch 7 Batch 30/173] avg loss 0.0075117, throughput 3.97519K wps
[Epoch 7 Batch 60/173] avg loss 0.00690286, throughput 3.86084K wps
[Epoch 7 Batch 90/173] avg loss 0.00690899, throughput 3.9024K wps
[Epoch 7 Batch 120/173] avg loss 0.00680203, throughput 3.8832K wps
[Epoch 7 Batch 150/173] avg loss 0.00690922, throughput 3.88564K wps
Begin Testing...
[Epoch 7] train avg loss 0.0069992, test acc 0.8021, test avg loss 0.442972, throughput 3.90297K wps
Observed Improvement.
Begin Testing...
[Epoch 8 Batch 30/173] avg loss 0.00612789, throughput 3.97138K wps
[Epoch 8 Batch 60/173] avg loss 0.0060099, throughput 3.87851K wps
[Epoch 8 Batch 90/173] avg loss 0.0060156, throughput 3.88061K wps
[Epoch 8 Batch 120/173] avg loss 0.00632995, throughput 3.88084K wps
[Epoch 8 Batch 150/173] avg loss 0.00595984, throughput 3.86439K wps
Begin Testing...
[Epoch 8] train avg loss 0.00605171, test acc 0.8000, test avg loss 0.428655, throughput 3.89314K wps
[Epoch 9 Batch 30/173] avg loss 0.00502566, throughput 3.97779K wps
[Epoch 9 Batch 60/173] avg loss 0.00557174, throughput 3.88814K wps
[Epoch 9 Batch 90/173] avg loss 0.00539784, throughput 3.89878K wps
[Epoch 9 Batch 120/173] avg loss 0.00526368, throughput 3.89468K wps
[Epoch 9 Batch 150/173] avg loss 0.0055063, throughput 3.86773K wps
Begin Testing...
[Epoch 9] train avg loss 0.00531596, test acc 0.8115, test avg loss 0.420386, throughput 3.89911K wps
Observed Improvement.
Begin Testing...
[Epoch 10 Batch 30/173] avg loss 0.00487105, throughput 3.95596K wps
[Epoch 10 Batch 60/173] avg loss 0.00496376, throughput 3.87373K wps
[Epoch 10 Batch 90/173] avg loss 0.00451874, throughput 3.89803K wps
[Epoch 10 Batch 120/173] avg loss 0.00458092, throughput 3.89098K wps
[Epoch 10 Batch 150/173] avg loss 0.00468845, throughput 3.88229K wps
Begin Testing...
[Epoch 10] train avg loss 0.00467953, test acc 0.8000, test avg loss 0.424841, throughput 3.90244K wps
[Epoch 11 Batch 30/173] avg loss 0.00407364, throughput 4.00182K wps
[Epoch 11 Batch 60/173] avg loss 0.00389489, throughput 3.88155K wps
[Epoch 11 Batch 90/173] avg loss 0.00421368, throughput 3.86704K wps
[Epoch 11 Batch 120/173] avg loss 0.0039297, throughput 3.86382K wps
[Epoch 11 Batch 150/173] avg loss 0.00403883, throughput 3.86794K wps
Begin Testing...
[Epoch 11] train avg loss 0.00398373, test acc 0.7990, test avg loss 0.431057, throughput 3.89436K wps
[Epoch 12 Batch 30/173] avg loss 0.00329335, throughput 3.97801K wps
[Epoch 12 Batch 60/173] avg loss 0.00362268, throughput 3.88899K wps
[Epoch 12 Batch 90/173] avg loss 0.0036326, throughput 3.91015K wps
[Epoch 12 Batch 120/173] avg loss 0.003364, throughput 3.88495K wps
[Epoch 12 Batch 150/173] avg loss 0.00326597, throughput 3.876K wps
Begin Testing...
[Epoch 12] train avg loss 0.00342353, test acc 0.7990, test avg loss 0.431363, throughput 3.90706K wps
[Epoch 13 Batch 30/173] avg loss 0.00304061, throughput 3.97572K wps
[Epoch 13 Batch 60/173] avg loss 0.0028679, throughput 3.87823K wps
[Epoch 13 Batch 90/173] avg loss 0.00283749, throughput 3.89031K wps
[Epoch 13 Batch 120/173] avg loss 0.00309014, throughput 3.87459K wps
[Epoch 13 Batch 150/173] avg loss 0.00270312, throughput 3.88067K wps
Begin Testing...
[Epoch 13] train avg loss 0.00288054, test acc 0.7990, test avg loss 0.438422, throughput 3.89918K wps
[Epoch 14 Batch 30/173] avg loss 0.0024297, throughput 3.96586K wps
[Epoch 14 Batch 60/173] avg loss 0.00227095, throughput 3.87939K wps
[Epoch 14 Batch 90/173] avg loss 0.00246689, throughput 3.89713K wps
[Epoch 14 Batch 120/173] avg loss 0.00258007, throughput 3.88235K wps
[Epoch 14 Batch 150/173] avg loss 0.00247575, throughput 3.87764K wps
Begin Testing...
[Epoch 14] train avg loss 0.00244993, test acc 0.7969, test avg loss 0.455099, throughput 3.89675K wps
[Epoch 15 Batch 30/173] avg loss 0.00184221, throughput 3.96942K wps
[Epoch 15 Batch 60/173] avg loss 0.00202357, throughput 3.87602K wps
[Epoch 15 Batch 90/173] avg loss 0.00212015, throughput 3.89021K wps
[Epoch 15 Batch 120/173] avg loss 0.00214999, throughput 3.90435K wps
[Epoch 15 Batch 150/173] avg loss 0.0020154, throughput 3.89037K wps
Begin Testing...
[Epoch 15] train avg loss 0.00204781, test acc 0.7948, test avg loss 0.471761, throughput 3.90249K wps
[Epoch 16 Batch 30/173] avg loss 0.00171096, throughput 3.97556K wps
[Epoch 16 Batch 60/173] avg loss 0.00178465, throughput 3.88295K wps
[Epoch 16 Batch 90/173] avg loss 0.00185572, throughput 3.87157K wps
[Epoch 16 Batch 120/173] avg loss 0.00168746, throughput 3.89331K wps
[Epoch 16 Batch 150/173] avg loss 0.00164499, throughput 3.87647K wps
Begin Testing...
[Epoch 16] train avg loss 0.00171719, test acc 0.7937, test avg loss 0.48843, throughput 3.89756K wps
[Epoch 17 Batch 30/173] avg loss 0.00153138, throughput 3.9838K wps
[Epoch 17 Batch 60/173] avg loss 0.00130422, throughput 3.87681K wps
[Epoch 17 Batch 90/173] avg loss 0.00128941, throughput 3.88535K wps
[Epoch 17 Batch 120/173] avg loss 0.0015117, throughput 3.88454K wps
[Epoch 17 Batch 150/173] avg loss 0.00166734, throughput 3.87215K wps
Begin Testing...
[Epoch 17] train avg loss 0.00147176, test acc 0.7875, test avg loss 0.50909, throughput 3.89906K wps
[Epoch 18 Batch 30/173] avg loss 0.00131118, throughput 3.98033K wps
[Epoch 18 Batch 60/173] avg loss 0.00124444, throughput 3.87388K wps
[Epoch 18 Batch 90/173] avg loss 0.00126123, throughput 3.87432K wps
[Epoch 18 Batch 120/173] avg loss 0.00113672, throughput 3.88905K wps
[Epoch 18 Batch 150/173] avg loss 0.00129587, throughput 3.89453K wps
Begin Testing...
[Epoch 18] train avg loss 0.00125157, test acc 0.7865, test avg loss 0.528109, throughput 3.90065K wps
[Epoch 19 Batch 30/173] avg loss 0.00108528, throughput 3.98779K wps
[Epoch 19 Batch 60/173] avg loss 0.00105347, throughput 3.87282K wps
[Epoch 19 Batch 90/173] avg loss 0.000988569, throughput 3.87361K wps
[Epoch 19 Batch 120/173] avg loss 0.00103766, throughput 3.88158K wps
[Epoch 19 Batch 150/173] avg loss 0.00102922, throughput 3.89515K wps
Begin Testing...
[Epoch 19] train avg loss 0.0010418, test acc 0.7823, test avg loss 0.562744, throughput 3.89929K wps
[Epoch 20 Batch 30/173] avg loss 0.000920046, throughput 3.99264K wps
[Epoch 20 Batch 60/173] avg loss 0.000951142, throughput 3.8944K wps
[Epoch 20 Batch 90/173] avg loss 0.000851789, throughput 3.8882K wps
[Epoch 20 Batch 120/173] avg loss 0.000845956, throughput 3.91012K wps
[Epoch 20 Batch 150/173] avg loss 0.000780001, throughput 3.88929K wps
Begin Testing...
[Epoch 20] train avg loss 0.000865401, test acc 0.7771, test avg loss 0.57701, throughput 3.90978K wps
[Epoch 21 Batch 30/173] avg loss 0.000635607, throughput 3.96775K wps
[Epoch 21 Batch 60/173] avg loss 0.000651901, throughput 3.88082K wps
[Epoch 21 Batch 90/173] avg loss 0.000722826, throughput 3.88815K wps
[Epoch 21 Batch 120/173] avg loss 0.000745689, throughput 3.89433K wps
[Epoch 21 Batch 150/173] avg loss 0.000718772, throughput 3.87574K wps
Begin Testing...
[Epoch 21] train avg loss 0.000719092, test acc 0.7781, test avg loss 0.596161, throughput 3.89663K wps
[Epoch 22 Batch 30/173] avg loss 0.00060308, throughput 3.97653K wps
[Epoch 22 Batch 60/173] avg loss 0.000551367, throughput 3.87491K wps
[Epoch 22 Batch 90/173] avg loss 0.000617936, throughput 3.87734K wps
[Epoch 22 Batch 120/173] avg loss 0.000605092, throughput 3.87348K wps
[Epoch 22 Batch 150/173] avg loss 0.000652245, throughput 3.86355K wps
Begin Testing...
[Epoch 22] train avg loss 0.000603528, test acc 0.7771, test avg loss 0.622134, throughput 3.8919K wps
[Epoch 23 Batch 30/173] avg loss 0.00051173, throughput 3.99742K wps
[Epoch 23 Batch 60/173] avg loss 0.000486819, throughput 3.90389K wps
[Epoch 23 Batch 90/173] avg loss 0.000542674, throughput 3.8858K wps
[Epoch 23 Batch 120/173] avg loss 0.00047906, throughput 3.89707K wps
[Epoch 23 Batch 150/173] avg loss 0.000580436, throughput 3.90018K wps
Begin Testing...
[Epoch 23] train avg loss 0.000523591, test acc 0.7760, test avg loss 0.643596, throughput 3.91147K wps
[Epoch 24 Batch 30/173] avg loss 0.000460903, throughput 3.97953K wps
[Epoch 24 Batch 60/173] avg loss 0.000427395, throughput 3.8721K wps
[Epoch 24 Batch 90/173] avg loss 0.000429835, throughput 3.87681K wps
[Epoch 24 Batch 120/173] avg loss 0.00043484, throughput 3.8972K wps
[Epoch 24 Batch 150/173] avg loss 0.00044269, throughput 3.89295K wps
Begin Testing...
[Epoch 24] train avg loss 0.000445753, test acc 0.7792, test avg loss 0.667961, throughput 3.89927K wps
[Epoch 25 Batch 30/173] avg loss 0.000403931, throughput 3.9783K wps
[Epoch 25 Batch 60/173] avg loss 0.00039322, throughput 3.87095K wps
[Epoch 25 Batch 90/173] avg loss 0.000371426, throughput 3.86965K wps
[Epoch 25 Batch 120/173] avg loss 0.000391905, throughput 3.87499K wps
[Epoch 25 Batch 150/173] avg loss 0.00039579, throughput 3.90396K wps
Begin Testing...
[Epoch 25] train avg loss 0.000389674, test acc 0.7750, test avg loss 0.696825, throughput 3.89995K wps
[Epoch 26 Batch 30/173] avg loss 0.000339963, throughput 3.99358K wps
[Epoch 26 Batch 60/173] avg loss 0.000339669, throughput 3.89073K wps
[Epoch 26 Batch 90/173] avg loss 0.000364713, throughput 3.91975K wps
[Epoch 26 Batch 120/173] avg loss 0.000314242, throughput 3.89144K wps
[Epoch 26 Batch 150/173] avg loss 0.000333294, throughput 3.88292K wps
Begin Testing...
[Epoch 26] train avg loss 0.000337024, test acc 0.7750, test avg loss 0.712271, throughput 3.9138K wps
[Epoch 27 Batch 30/173] avg loss 0.000272886, throughput 3.98171K wps
[Epoch 27 Batch 60/173] avg loss 0.000276818, throughput 3.894K wps
[Epoch 27 Batch 90/173] avg loss 0.000280127, throughput 3.87663K wps
[Epoch 27 Batch 120/173] avg loss 0.0002985, throughput 3.87853K wps
[Epoch 27 Batch 150/173] avg loss 0.000231136, throughput 3.89332K wps
Begin Testing...
[Epoch 27] train avg loss 0.000277483, test acc 0.7708, test avg loss 0.743244, throughput 3.90294K wps
[Epoch 28 Batch 30/173] avg loss 0.000226129, throughput 3.96414K wps
[Epoch 28 Batch 60/173] avg loss 0.000278499, throughput 3.88493K wps
[Epoch 28 Batch 90/173] avg loss 0.000231443, throughput 3.88608K wps
[Epoch 28 Batch 120/173] avg loss 0.000253953, throughput 3.89128K wps
[Epoch 28 Batch 150/173] avg loss 0.000256734, throughput 3.91047K wps
Begin Testing...
[Epoch 28] train avg loss 0.000252054, test acc 0.7688, test avg loss 0.76583, throughput 3.90489K wps
[Epoch 29 Batch 30/173] avg loss 0.000201097, throughput 3.98967K wps
[Epoch 29 Batch 60/173] avg loss 0.000187676, throughput 3.88455K wps
[Epoch 29 Batch 90/173] avg loss 0.00020876, throughput 3.88149K wps
[Epoch 29 Batch 120/173] avg loss 0.00020488, throughput 3.87438K wps
[Epoch 29 Batch 150/173] avg loss 0.000202321, throughput 3.8889K wps
Begin Testing...
[Epoch 29] train avg loss 0.000206568, test acc 0.7688, test avg loss 0.784874, throughput 3.89928K wps
[Epoch 30 Batch 30/173] avg loss 0.000176801, throughput 3.98724K wps
[Epoch 30 Batch 60/173] avg loss 0.000197963, throughput 3.87411K wps
[Epoch 30 Batch 90/173] avg loss 0.000155775, throughput 3.88187K wps
[Epoch 30 Batch 120/173] avg loss 0.000164673, throughput 3.8753K wps
[Epoch 30 Batch 150/173] avg loss 0.000176676, throughput 3.87239K wps
Begin Testing...
[Epoch 30] train avg loss 0.000175922, test acc 0.7719, test avg loss 0.81396, throughput 3.89457K wps
[Epoch 31 Batch 30/173] avg loss 0.000169602, throughput 3.95702K wps
[Epoch 31 Batch 60/173] avg loss 0.000162428, throughput 3.8737K wps
[Epoch 31 Batch 90/173] avg loss 0.000159314, throughput 3.90485K wps
[Epoch 31 Batch 120/173] avg loss 0.000163686, throughput 3.90078K wps
[Epoch 31 Batch 150/173] avg loss 0.000158809, throughput 3.88464K wps
Begin Testing...
[Epoch 31] train avg loss 0.000164906, test acc 0.7688, test avg loss 0.832437, throughput 3.90274K wps
[Epoch 32 Batch 30/173] avg loss 0.000136839, throughput 3.97967K wps
[Epoch 32 Batch 60/173] avg loss 0.00014005, throughput 3.87685K wps
[Epoch 32 Batch 90/173] avg loss 0.000139168, throughput 3.89844K wps
[Epoch 32 Batch 120/173] avg loss 0.000163179, throughput 3.88821K wps
[Epoch 32 Batch 150/173] avg loss 0.00013998, throughput 3.87921K wps
Begin Testing...
[Epoch 32] train avg loss 0.000143304, test acc 0.7688, test avg loss 0.858537, throughput 3.90113K wps
[Epoch 33 Batch 30/173] avg loss 0.0001063, throughput 3.96883K wps
[Epoch 33 Batch 60/173] avg loss 0.000112851, throughput 3.88206K wps
[Epoch 33 Batch 90/173] avg loss 0.000109828, throughput 3.8708K wps
[Epoch 33 Batch 120/173] avg loss 0.000124664, throughput 3.88579K wps
[Epoch 33 Batch 150/173] avg loss 0.000156723, throughput 3.88049K wps
Begin Testing...
[Epoch 33] train avg loss 0.000124962, test acc 0.7698, test avg loss 0.879451, throughput 3.89529K wps
[Epoch 34 Batch 30/173] avg loss 0.000152566, throughput 3.98161K wps
[Epoch 34 Batch 60/173] avg loss 0.000121271, throughput 3.88269K wps
[Epoch 34 Batch 90/173] avg loss 0.000106364, throughput 3.88703K wps
[Epoch 34 Batch 120/173] avg loss 0.000126403, throughput 3.90992K wps
[Epoch 34 Batch 150/173] avg loss 0.000114615, throughput 3.88584K wps
Begin Testing...
[Epoch 34] train avg loss 0.00012568, test acc 0.7615, test avg loss 0.912136, throughput 3.903K wps
[Epoch 35 Batch 30/173] avg loss 0.000129611, throughput 3.97425K wps
[Epoch 35 Batch 60/173] avg loss 9.89309e-05, throughput 3.88226K wps
[Epoch 35 Batch 90/173] avg loss 9.45136e-05, throughput 3.87544K wps
[Epoch 35 Batch 120/173] avg loss 9.62379e-05, throughput 3.89302K wps
[Epoch 35 Batch 150/173] avg loss 0.000102135, throughput 3.89584K wps
Begin Testing...
[Epoch 35] train avg loss 0.000102991, test acc 0.7594, test avg loss 0.930879, throughput 3.9008K wps
[Epoch 36 Batch 30/173] avg loss 9.4913e-05, throughput 3.97828K wps
[Epoch 36 Batch 60/173] avg loss 8.9809e-05, throughput 3.86987K wps
[Epoch 36 Batch 90/173] avg loss 8.4903e-05, throughput 3.87375K wps
[Epoch 36 Batch 120/173] avg loss 9.04166e-05, throughput 3.88144K wps
[Epoch 36 Batch 150/173] avg loss 0.00011385, throughput 3.90885K wps
Begin Testing...
[Epoch 36] train avg loss 9.43982e-05, test acc 0.7646, test avg loss 0.945894, throughput 3.90074K wps
[Epoch 37 Batch 30/173] avg loss 6.76273e-05, throughput 3.98553K wps
[Epoch 37 Batch 60/173] avg loss 6.8887e-05, throughput 3.89131K wps
[Epoch 37 Batch 90/173] avg loss 9.12957e-05, throughput 3.91247K wps
[Epoch 37 Batch 120/173] avg loss 7.02464e-05, throughput 3.87915K wps
[Epoch 37 Batch 150/173] avg loss 6.78752e-05, throughput 3.87968K wps
Begin Testing...
[Epoch 37] train avg loss 7.54409e-05, test acc 0.7625, test avg loss 0.970041, throughput 3.90433K wps
[Epoch 38 Batch 30/173] avg loss 7.90576e-05, throughput 3.96733K wps
[Epoch 38 Batch 60/173] avg loss 9.66625e-05, throughput 3.88501K wps
[Epoch 38 Batch 90/173] avg loss 7.11759e-05, throughput 3.88436K wps
[Epoch 38 Batch 120/173] avg loss 8.08076e-05, throughput 3.88973K wps
[Epoch 38 Batch 150/173] avg loss 6.36356e-05, throughput 3.88518K wps
Begin Testing...
[Epoch 38] train avg loss 7.75805e-05, test acc 0.7583, test avg loss 1.00941, throughput 3.89831K wps
[Epoch 39 Batch 30/173] avg loss 6.89398e-05, throughput 3.97422K wps
[Epoch 39 Batch 60/173] avg loss 4.77189e-05, throughput 3.86882K wps
[Epoch 39 Batch 90/173] avg loss 6.62838e-05, throughput 3.8836K wps
[Epoch 39 Batch 120/173] avg loss 6.62739e-05, throughput 3.89812K wps
[Epoch 39 Batch 150/173] avg loss 7.87268e-05, throughput 3.91222K wps
Begin Testing...
[Epoch 39] train avg loss 7.11892e-05, test acc 0.7698, test avg loss 1.02453, throughput 3.90282K wps
[Epoch 40 Batch 30/173] avg loss 6.20415e-05, throughput 3.98437K wps
[Epoch 40 Batch 60/173] avg loss 4.96375e-05, throughput 3.87859K wps
[Epoch 40 Batch 90/173] avg loss 4.77353e-05, throughput 3.88736K wps
[Epoch 40 Batch 120/173] avg loss 5.95608e-05, throughput 3.90163K wps
[Epoch 40 Batch 150/173] avg loss 6.44051e-05, throughput 3.87174K wps
Begin Testing...
[Epoch 40] train avg loss 5.59826e-05, test acc 0.7583, test avg loss 1.03806, throughput 3.90036K wps
[Epoch 41 Batch 30/173] avg loss 5.01799e-05, throughput 3.98469K wps
[Epoch 41 Batch 60/173] avg loss 6.13776e-05, throughput 3.87491K wps
[Epoch 41 Batch 90/173] avg loss 4.39685e-05, throughput 3.87768K wps
[Epoch 41 Batch 120/173] avg loss 4.58314e-05, throughput 3.87866K wps
[Epoch 41 Batch 150/173] avg loss 4.82254e-05, throughput 3.87697K wps
Begin Testing...
[Epoch 41] train avg loss 5.04315e-05, test acc 0.7594, test avg loss 1.05458, throughput 3.89714K wps
[Epoch 42 Batch 30/173] avg loss 5.27565e-05, throughput 4.00985K wps
[Epoch 42 Batch 60/173] avg loss 4.42998e-05, throughput 3.89105K wps
[Epoch 42 Batch 90/173] avg loss 4.85115e-05, throughput 3.90062K wps
[Epoch 42 Batch 120/173] avg loss 5.03757e-05, throughput 3.91824K wps
[Epoch 42 Batch 150/173] avg loss 3.9126e-05, throughput 3.89796K wps
Begin Testing...
[Epoch 42] train avg loss 4.80415e-05, test acc 0.7594, test avg loss 1.07946, throughput 3.92219K wps
[Epoch 43 Batch 30/173] avg loss 5.84504e-05, throughput 3.98423K wps
[Epoch 43 Batch 60/173] avg loss 4.46998e-05, throughput 3.88904K wps
[Epoch 43 Batch 90/173] avg loss 3.57738e-05, throughput 3.87579K wps
[Epoch 43 Batch 120/173] avg loss 3.8889e-05, throughput 3.87586K wps
[Epoch 43 Batch 150/173] avg loss 2.94755e-05, throughput 3.87759K wps
Begin Testing...
[Epoch 43] train avg loss 4.29592e-05, test acc 0.7583, test avg loss 1.11671, throughput 3.90089K wps
[Epoch 44 Batch 30/173] avg loss 4.42961e-05, throughput 3.98453K wps
[Epoch 44 Batch 60/173] avg loss 4.51881e-05, throughput 3.88783K wps
[Epoch 44 Batch 90/173] avg loss 3.03617e-05, throughput 3.87629K wps
[Epoch 44 Batch 120/173] avg loss 4.46913e-05, throughput 3.86295K wps
[Epoch 44 Batch 150/173] avg loss 3.81714e-05, throughput 3.89514K wps
Begin Testing...
[Epoch 44] train avg loss 4.08836e-05, test acc 0.7625, test avg loss 1.12024, throughput 3.90378K wps
[Epoch 45 Batch 30/173] avg loss 3.71313e-05, throughput 3.99436K wps
[Epoch 45 Batch 60/173] avg loss 2.86209e-05, throughput 3.88164K wps
[Epoch 45 Batch 90/173] avg loss 3.80964e-05, throughput 3.88791K wps
[Epoch 45 Batch 120/173] avg loss 4.09403e-05, throughput 3.88808K wps
[Epoch 45 Batch 150/173] avg loss 3.94108e-05, throughput 3.88154K wps
Begin Testing...
[Epoch 45] train avg loss 3.73382e-05, test acc 0.7573, test avg loss 1.14258, throughput 3.90269K wps
[Epoch 46 Batch 30/173] avg loss 2.23365e-05, throughput 3.97885K wps
[Epoch 46 Batch 60/173] avg loss 3.33022e-05, throughput 3.88602K wps
[Epoch 46 Batch 90/173] avg loss 3.4018e-05, throughput 3.87679K wps
[Epoch 46 Batch 120/173] avg loss 2.92854e-05, throughput 3.87876K wps
[Epoch 46 Batch 150/173] avg loss 3.09755e-05, throughput 3.89141K wps
Begin Testing...
[Epoch 46] train avg loss 2.92807e-05, test acc 0.7583, test avg loss 1.16277, throughput 3.90033K wps
[Epoch 47 Batch 30/173] avg loss 3.28321e-05, throughput 3.97124K wps
[Epoch 47 Batch 60/173] avg loss 2.87597e-05, throughput 3.89773K wps
[Epoch 47 Batch 90/173] avg loss 3.92958e-05, throughput 3.8932K wps
[Epoch 47 Batch 120/173] avg loss 2.37712e-05, throughput 3.89421K wps
[Epoch 47 Batch 150/173] avg loss 2.5401e-05, throughput 3.91025K wps
Begin Testing...
[Epoch 47] train avg loss 2.9438e-05, test acc 0.7615, test avg loss 1.1778, throughput 3.90836K wps
[Epoch 48 Batch 30/173] avg loss 2.42233e-05, throughput 3.98064K wps
[Epoch 48 Batch 60/173] avg loss 1.54377e-05, throughput 3.88615K wps
[Epoch 48 Batch 90/173] avg loss 1.9894e-05, throughput 3.88217K wps
[Epoch 48 Batch 120/173] avg loss 2.35846e-05, throughput 3.87997K wps
[Epoch 48 Batch 150/173] avg loss 2.1559e-05, throughput 3.8784K wps
Begin Testing...
[Epoch 48] train avg loss 2.13701e-05, test acc 0.7562, test avg loss 1.20653, throughput 3.8977K wps
[Epoch 49 Batch 30/173] avg loss 2.26469e-05, throughput 3.98903K wps
[Epoch 49 Batch 60/173] avg loss 2.57336e-05, throughput 3.87947K wps
[Epoch 49 Batch 90/173] avg loss 2.22598e-05, throughput 3.87506K wps
[Epoch 49 Batch 120/173] avg loss 2.79181e-05, throughput 3.87901K wps
[Epoch 49 Batch 150/173] avg loss 1.88066e-05, throughput 3.88426K wps
Begin Testing...
[Epoch 49] train avg loss 2.33844e-05, test acc 0.7562, test avg loss 1.22054, throughput 3.90025K wps
[Epoch 50 Batch 30/173] avg loss 1.98659e-05, throughput 4.00122K wps
[Epoch 50 Batch 60/173] avg loss 1.89587e-05, throughput 3.89579K wps
[Epoch 50 Batch 90/173] avg loss 2.17075e-05, throughput 3.90765K wps
[Epoch 50 Batch 120/173] avg loss 1.68518e-05, throughput 3.89132K wps
[Epoch 50 Batch 150/173] avg loss 1.82942e-05, throughput 3.87896K wps
Begin Testing...
[Epoch 50] train avg loss 1.92554e-05, test acc 0.7604, test avg loss 1.25568, throughput 3.90965K wps
[Epoch 51 Batch 30/173] avg loss 1.50499e-05, throughput 3.9664K wps
[Epoch 51 Batch 60/173] avg loss 1.37447e-05, throughput 3.87838K wps
[Epoch 51 Batch 90/173] avg loss 1.78551e-05, throughput 3.86144K wps
[Epoch 51 Batch 120/173] avg loss 1.73838e-05, throughput 3.88079K wps
[Epoch 51 Batch 150/173] avg loss 1.55915e-05, throughput 3.88225K wps
Begin Testing...
[Epoch 51] train avg loss 1.70484e-05, test acc 0.7594, test avg loss 1.27247, throughput 3.89088K wps
[Epoch 52 Batch 30/173] avg loss 1.42778e-05, throughput 3.98882K wps
[Epoch 52 Batch 60/173] avg loss 1.53682e-05, throughput 3.87412K wps
[Epoch 52 Batch 90/173] avg loss 1.18409e-05, throughput 3.86895K wps
[Epoch 52 Batch 120/173] avg loss 2.75873e-05, throughput 3.87259K wps
[Epoch 52 Batch 150/173] avg loss 3.84708e-05, throughput 3.89326K wps
Begin Testing...
[Epoch 52] train avg loss 2.23398e-05, test acc 0.7594, test avg loss 1.27093, throughput 3.90036K wps
[Epoch 53 Batch 30/173] avg loss 2.41705e-05, throughput 3.99843K wps
[Epoch 53 Batch 60/173] avg loss 1.85386e-05, throughput 3.88474K wps
[Epoch 53 Batch 90/173] avg loss 1.61377e-05, throughput 3.88122K wps
[Epoch 53 Batch 120/173] avg loss 1.30028e-05, throughput 3.8825K wps
[Epoch 53 Batch 150/173] avg loss 2.2086e-05, throughput 3.88865K wps
Begin Testing...
[Epoch 53] train avg loss 1.94157e-05, test acc 0.7573, test avg loss 1.3016, throughput 3.90229K wps
[Epoch 54 Batch 30/173] avg loss 2.66534e-05, throughput 3.97824K wps
[Epoch 54 Batch 60/173] avg loss 1.75295e-05, throughput 3.887K wps
[Epoch 54 Batch 90/173] avg loss 2.61806e-05, throughput 3.87668K wps
[Epoch 54 Batch 120/173] avg loss 1.41963e-05, throughput 3.87449K wps
[Epoch 54 Batch 150/173] avg loss 2.17354e-05, throughput 3.884K wps
Begin Testing...
[Epoch 54] train avg loss 2.10046e-05, test acc 0.7531, test avg loss 1.32119, throughput 3.8988K wps
[Epoch 55 Batch 30/173] avg loss 1.94836e-05, throughput 3.94766K wps
[Epoch 55 Batch 60/173] avg loss 1.49292e-05, throughput 3.88648K wps
[Epoch 55 Batch 90/173] avg loss 1.84806e-05, throughput 3.90392K wps
[Epoch 55 Batch 120/173] avg loss 1.80394e-05, throughput 3.89079K wps
[Epoch 55 Batch 150/173] avg loss 1.27731e-05, throughput 3.88515K wps
Begin Testing...
[Epoch 55] train avg loss 1.6514e-05, test acc 0.7562, test avg loss 1.35331, throughput 3.90458K wps
[Epoch 56 Batch 30/173] avg loss 1.19849e-05, throughput 3.9888K wps
[Epoch 56 Batch 60/173] avg loss 1.16182e-05, throughput 3.88675K wps
[Epoch 56 Batch 90/173] avg loss 1.58753e-05, throughput 3.87217K wps
[Epoch 56 Batch 120/173] avg loss 1.27399e-05, throughput 3.86909K wps
[Epoch 56 Batch 150/173] avg loss 1.50483e-05, throughput 3.86833K wps
Begin Testing...
[Epoch 56] train avg loss 1.29947e-05, test acc 0.7542, test avg loss 1.37741, throughput 3.89608K wps
[Epoch 57 Batch 30/173] avg loss 1.44247e-05, throughput 3.96188K wps
[Epoch 57 Batch 60/173] avg loss 1.04357e-05, throughput 3.87078K wps
[Epoch 57 Batch 90/173] avg loss 1.20439e-05, throughput 3.87821K wps
[Epoch 57 Batch 120/173] avg loss 1.16388e-05, throughput 3.86897K wps
[Epoch 57 Batch 150/173] avg loss 8.71752e-06, throughput 3.86323K wps
Begin Testing...
[Epoch 57] train avg loss 1.09903e-05, test acc 0.7510, test avg loss 1.4048, throughput 3.88804K wps
[Epoch 58 Batch 30/173] avg loss 1.26653e-05, throughput 3.99291K wps
[Epoch 58 Batch 60/173] avg loss 1.28098e-05, throughput 3.88258K wps
[Epoch 58 Batch 90/173] avg loss 8.44062e-06, throughput 3.8869K wps
[Epoch 58 Batch 120/173] avg loss 1.10535e-05, throughput 3.90895K wps
[Epoch 58 Batch 150/173] avg loss 1.16144e-05, throughput 3.88863K wps
Begin Testing...
[Epoch 58] train avg loss 1.11187e-05, test acc 0.7521, test avg loss 1.41315, throughput 3.90531K wps
[Epoch 59 Batch 30/173] avg loss 7.67523e-06, throughput 3.97068K wps
[Epoch 59 Batch 60/173] avg loss 8.50017e-06, throughput 3.86861K wps
[Epoch 59 Batch 90/173] avg loss 9.48446e-06, throughput 3.87429K wps
[Epoch 59 Batch 120/173] avg loss 1.06125e-05, throughput 3.88414K wps
[Epoch 59 Batch 150/173] avg loss 7.83002e-06, throughput 3.89271K wps
Begin Testing...
[Epoch 59] train avg loss 8.56877e-06, test acc 0.7521, test avg loss 1.43988, throughput 3.89564K wps
Test loss 0.448665, test acc 0.7739
Total time cost 553.98s
9596 1066
[Epoch 0 Batch 30/173] avg loss 0.015382, throughput 3.66965K wps
[Epoch 0 Batch 60/173] avg loss 0.0154958, throughput 3.86472K wps
[Epoch 0 Batch 90/173] avg loss 0.0146504, throughput 3.87776K wps
[Epoch 0 Batch 120/173] avg loss 0.0143133, throughput 3.87786K wps
[Epoch 0 Batch 150/173] avg loss 0.0140893, throughput 3.92144K wps
Begin Testing...
[Epoch 0] train avg loss 0.0146669, test acc 0.5833, test avg loss 0.669483, throughput 3.84648K wps
Observed Improvement.
Begin Testing...
[Epoch 1 Batch 30/173] avg loss 0.0131955, throughput 3.99373K wps
[Epoch 1 Batch 60/173] avg loss 0.0132964, throughput 3.88633K wps
[Epoch 1 Batch 90/173] avg loss 0.0129858, throughput 3.886K wps
[Epoch 1 Batch 120/173] avg loss 0.0127859, throughput 3.88076K wps
[Epoch 1 Batch 150/173] avg loss 0.0128815, throughput 3.88293K wps
Begin Testing...
[Epoch 1] train avg loss 0.0129812, test acc 0.6792, test avg loss 0.62599, throughput 3.90053K wps
Observed Improvement.
Begin Testing...
[Epoch 2 Batch 30/173] avg loss 0.0122051, throughput 3.98091K wps
[Epoch 2 Batch 60/173] avg loss 0.0122155, throughput 3.89015K wps
[Epoch 2 Batch 90/173] avg loss 0.0121304, throughput 3.88572K wps
[Epoch 2 Batch 120/173] avg loss 0.0118835, throughput 3.88073K wps
[Epoch 2 Batch 150/173] avg loss 0.0118567, throughput 3.87738K wps
Begin Testing...
[Epoch 2] train avg loss 0.0120628, test acc 0.7010, test avg loss 0.597759, throughput 3.90148K wps
Observed Improvement.
Begin Testing...
[Epoch 3 Batch 30/173] avg loss 0.0114094, throughput 3.98614K wps
[Epoch 3 Batch 60/173] avg loss 0.0112607, throughput 3.8839K wps
[Epoch 3 Batch 90/173] avg loss 0.0113659, throughput 3.90334K wps
[Epoch 3 Batch 120/173] avg loss 0.0109835, throughput 3.89755K wps
[Epoch 3 Batch 150/173] avg loss 0.0111713, throughput 3.88846K wps
Begin Testing...
[Epoch 3] train avg loss 0.0112168, test acc 0.7406, test avg loss 0.560696, throughput 3.91032K wps
Observed Improvement.
Begin Testing...
[Epoch 4 Batch 30/173] avg loss 0.010359, throughput 3.98039K wps
[Epoch 4 Batch 60/173] avg loss 0.0102949, throughput 3.88044K wps
[Epoch 4 Batch 90/173] avg loss 0.00999944, throughput 3.86926K wps
[Epoch 4 Batch 120/173] avg loss 0.010173, throughput 3.86962K wps
[Epoch 4 Batch 150/173] avg loss 0.00989089, throughput 3.88886K wps
Begin Testing...
[Epoch 4] train avg loss 0.010124, test acc 0.7552, test avg loss 0.529749, throughput 3.89719K wps
Observed Improvement.
Begin Testing...
[Epoch 5 Batch 30/173] avg loss 0.00929303, throughput 3.98232K wps
[Epoch 5 Batch 60/173] avg loss 0.0089247, throughput 3.90028K wps
[Epoch 5 Batch 90/173] avg loss 0.00925495, throughput 3.86938K wps
[Epoch 5 Batch 120/173] avg loss 0.0090342, throughput 3.87192K wps
[Epoch 5 Batch 150/173] avg loss 0.00893067, throughput 3.89207K wps
Begin Testing...
[Epoch 5] train avg loss 0.00907217, test acc 0.7729, test avg loss 0.494652, throughput 3.9039K wps
Observed Improvement.
Begin Testing...
[Epoch 6 Batch 30/173] avg loss 0.00814883, throughput 3.98102K wps
[Epoch 6 Batch 60/173] avg loss 0.00802948, throughput 3.90789K wps
[Epoch 6 Batch 90/173] avg loss 0.00806448, throughput 3.89002K wps
[Epoch 6 Batch 120/173] avg loss 0.00804318, throughput 3.87642K wps
[Epoch 6 Batch 150/173] avg loss 0.00790012, throughput 3.88215K wps
Begin Testing...
[Epoch 6] train avg loss 0.00797313, test acc 0.7865, test avg loss 0.472246, throughput 3.90599K wps
Observed Improvement.
Begin Testing...
[Epoch 7 Batch 30/173] avg loss 0.00707397, throughput 3.96382K wps
[Epoch 7 Batch 60/173] avg loss 0.00733736, throughput 3.87726K wps
[Epoch 7 Batch 90/173] avg loss 0.00686807, throughput 3.87373K wps
[Epoch 7 Batch 120/173] avg loss 0.00694048, throughput 3.88375K wps
[Epoch 7 Batch 150/173] avg loss 0.00681836, throughput 3.8839K wps
Begin Testing...
[Epoch 7] train avg loss 0.00703572, test acc 0.7990, test avg loss 0.450459, throughput 3.89221K wps
Observed Improvement.
Begin Testing...
[Epoch 8 Batch 30/173] avg loss 0.00618696, throughput 3.9646K wps
[Epoch 8 Batch 60/173] avg loss 0.00603119, throughput 3.88561K wps
[Epoch 8 Batch 90/173] avg loss 0.00627653, throughput 3.91094K wps
[Epoch 8 Batch 120/173] avg loss 0.00630802, throughput 3.88844K wps
[Epoch 8 Batch 150/173] avg loss 0.00595721, throughput 3.88521K wps
Begin Testing...
[Epoch 8] train avg loss 0.00617362, test acc 0.7937, test avg loss 0.441232, throughput 3.90679K wps
[Epoch 9 Batch 30/173] avg loss 0.00533575, throughput 3.98192K wps
[Epoch 9 Batch 60/173] avg loss 0.00542631, throughput 3.89318K wps
[Epoch 9 Batch 90/173] avg loss 0.00549081, throughput 3.87528K wps
[Epoch 9 Batch 120/173] avg loss 0.00533091, throughput 3.87756K wps
[Epoch 9 Batch 150/173] avg loss 0.00552189, throughput 3.87293K wps
Begin Testing...
[Epoch 9] train avg loss 0.00541671, test acc 0.8000, test avg loss 0.430659, throughput 3.89664K wps
Observed Improvement.
Begin Testing...
[Epoch 10 Batch 30/173] avg loss 0.00457393, throughput 3.9668K wps
[Epoch 10 Batch 60/173] avg loss 0.00467129, throughput 3.8971K wps
[Epoch 10 Batch 90/173] avg loss 0.00463941, throughput 3.86855K wps
[Epoch 10 Batch 120/173] avg loss 0.00439482, throughput 3.86529K wps
[Epoch 10 Batch 150/173] avg loss 0.00474004, throughput 3.88741K wps
Begin Testing...
[Epoch 10] train avg loss 0.0046651, test acc 0.7979, test avg loss 0.433392, throughput 3.89984K wps
[Epoch 11 Batch 30/173] avg loss 0.00403129, throughput 3.99158K wps
[Epoch 11 Batch 60/173] avg loss 0.00422652, throughput 3.88367K wps
[Epoch 11 Batch 90/173] avg loss 0.00420834, throughput 3.88864K wps
[Epoch 11 Batch 120/173] avg loss 0.0042519, throughput 3.89703K wps
[Epoch 11 Batch 150/173] avg loss 0.00389017, throughput 3.88521K wps
Begin Testing...
[Epoch 11] train avg loss 0.00408449, test acc 0.8010, test avg loss 0.434373, throughput 3.905K wps
Observed Improvement.
Begin Testing...
[Epoch 12 Batch 30/173] avg loss 0.00345584, throughput 3.98033K wps
[Epoch 12 Batch 60/173] avg loss 0.00345633, throughput 3.87924K wps
[Epoch 12 Batch 90/173] avg loss 0.00349973, throughput 3.87441K wps
[Epoch 12 Batch 120/173] avg loss 0.00333544, throughput 3.87353K wps
[Epoch 12 Batch 150/173] avg loss 0.00324411, throughput 3.87354K wps
Begin Testing...
[Epoch 12] train avg loss 0.00342587, test acc 0.7948, test avg loss 0.438358, throughput 3.89494K wps
[Epoch 13 Batch 30/173] avg loss 0.00310492, throughput 3.96235K wps
[Epoch 13 Batch 60/173] avg loss 0.00306231, throughput 3.86479K wps
[Epoch 13 Batch 90/173] avg loss 0.00297985, throughput 3.89744K wps
[Epoch 13 Batch 120/173] avg loss 0.00287886, throughput 3.89397K wps
[Epoch 13 Batch 150/173] avg loss 0.00281178, throughput 3.88633K wps
Begin Testing...
[Epoch 13] train avg loss 0.00298636, test acc 0.8021, test avg loss 0.442612, throughput 3.89988K wps
Observed Improvement.
Begin Testing...
[Epoch 14 Batch 30/173] avg loss 0.00238376, throughput 3.98174K wps
[Epoch 14 Batch 60/173] avg loss 0.00239045, throughput 3.87777K wps
[Epoch 14 Batch 90/173] avg loss 0.00255905, throughput 3.87281K wps
[Epoch 14 Batch 120/173] avg loss 0.00235952, throughput 3.89384K wps
[Epoch 14 Batch 150/173] avg loss 0.00242988, throughput 3.88347K wps
Begin Testing...
[Epoch 14] train avg loss 0.00243747, test acc 0.7969, test avg loss 0.459001, throughput 3.89753K wps
[Epoch 15 Batch 30/173] avg loss 0.00200762, throughput 3.98295K wps
[Epoch 15 Batch 60/173] avg loss 0.00210018, throughput 3.86906K wps
[Epoch 15 Batch 90/173] avg loss 0.00212368, throughput 3.87169K wps
[Epoch 15 Batch 120/173] avg loss 0.0022385, throughput 3.8828K wps
[Epoch 15 Batch 150/173] avg loss 0.00194429, throughput 3.89285K wps
Begin Testing...
[Epoch 15] train avg loss 0.0020743, test acc 0.7885, test avg loss 0.476822, throughput 3.89685K wps
[Epoch 16 Batch 30/173] avg loss 0.00178055, throughput 3.98138K wps
[Epoch 16 Batch 60/173] avg loss 0.00178435, throughput 3.89726K wps
[Epoch 16 Batch 90/173] avg loss 0.00176904, throughput 3.88779K wps
[Epoch 16 Batch 120/173] avg loss 0.00156391, throughput 3.89816K wps
[Epoch 16 Batch 150/173] avg loss 0.00173158, throughput 3.91159K wps
Begin Testing...
[Epoch 16] train avg loss 0.00175741, test acc 0.7865, test avg loss 0.482109, throughput 3.91001K wps
[Epoch 17 Batch 30/173] avg loss 0.00147897, throughput 3.9799K wps
[Epoch 17 Batch 60/173] avg loss 0.00151604, throughput 3.87774K wps
[Epoch 17 Batch 90/173] avg loss 0.00133969, throughput 3.86466K wps
[Epoch 17 Batch 120/173] avg loss 0.00146697, throughput 3.87201K wps
[Epoch 17 Batch 150/173] avg loss 0.0016112, throughput 3.88357K wps
Begin Testing...
[Epoch 17] train avg loss 0.00147741, test acc 0.7906, test avg loss 0.500841, throughput 3.89361K wps
[Epoch 18 Batch 30/173] avg loss 0.00123161, throughput 3.96222K wps
[Epoch 18 Batch 60/173] avg loss 0.00104392, throughput 3.90025K wps
[Epoch 18 Batch 90/173] avg loss 0.00121208, throughput 3.86857K wps
[Epoch 18 Batch 120/173] avg loss 0.00122518, throughput 3.86871K wps
[Epoch 18 Batch 150/173] avg loss 0.00116807, throughput 3.87922K wps
Begin Testing...
[Epoch 18] train avg loss 0.001213, test acc 0.7948, test avg loss 0.520471, throughput 3.89855K wps
[Epoch 19 Batch 30/173] avg loss 0.00109277, throughput 4.00238K wps
[Epoch 19 Batch 60/173] avg loss 0.0011194, throughput 3.89364K wps
[Epoch 19 Batch 90/173] avg loss 0.00102863, throughput 3.88833K wps
[Epoch 19 Batch 120/173] avg loss 0.000850817, throughput 3.88765K wps
[Epoch 19 Batch 150/173] avg loss 0.00114106, throughput 3.88809K wps
Begin Testing...
[Epoch 19] train avg loss 0.00103104, test acc 0.7844, test avg loss 0.541291, throughput 3.90751K wps
[Epoch 20 Batch 30/173] avg loss 0.000762188, throughput 3.97086K wps
[Epoch 20 Batch 60/173] avg loss 0.000852881, throughput 3.87748K wps
[Epoch 20 Batch 90/173] avg loss 0.000845848, throughput 3.88142K wps
[Epoch 20 Batch 120/173] avg loss 0.000944742, throughput 3.87004K wps
[Epoch 20 Batch 150/173] avg loss 0.000858596, throughput 3.88917K wps
Begin Testing...
[Epoch 20] train avg loss 0.000878821, test acc 0.7958, test avg loss 0.55992, throughput 3.89841K wps
[Epoch 21 Batch 30/173] avg loss 0.000708967, throughput 3.97785K wps
[Epoch 21 Batch 60/173] avg loss 0.000694062, throughput 3.88422K wps
[Epoch 21 Batch 90/173] avg loss 0.000670733, throughput 3.89171K wps
[Epoch 21 Batch 120/173] avg loss 0.000655155, throughput 3.88804K wps
[Epoch 21 Batch 150/173] avg loss 0.000778992, throughput 3.89649K wps
Begin Testing...
[Epoch 21] train avg loss 0.000718396, test acc 0.7823, test avg loss 0.584325, throughput 3.90642K wps
[Epoch 22 Batch 30/173] avg loss 0.000584496, throughput 3.99449K wps
[Epoch 22 Batch 60/173] avg loss 0.000673505, throughput 3.88452K wps
[Epoch 22 Batch 90/173] avg loss 0.000644645, throughput 3.88057K wps
[Epoch 22 Batch 120/173] avg loss 0.000709361, throughput 3.89194K wps
[Epoch 22 Batch 150/173] avg loss 0.000573523, throughput 3.88464K wps
Begin Testing...
[Epoch 22] train avg loss 0.000635825, test acc 0.7854, test avg loss 0.604877, throughput 3.90154K wps
[Epoch 23 Batch 30/173] avg loss 0.000483989, throughput 3.98447K wps
[Epoch 23 Batch 60/173] avg loss 0.000488586, throughput 3.87599K wps
[Epoch 23 Batch 90/173] avg loss 0.000465938, throughput 3.8826K wps
[Epoch 23 Batch 120/173] avg loss 0.000626668, throughput 3.87878K wps
[Epoch 23 Batch 150/173] avg loss 0.000495023, throughput 3.85303K wps
Begin Testing...
[Epoch 23] train avg loss 0.000521242, test acc 0.7948, test avg loss 0.6273, throughput 3.89555K wps
[Epoch 24 Batch 30/173] avg loss 0.000456164, throughput 4.0058K wps
[Epoch 24 Batch 60/173] avg loss 0.000422763, throughput 3.87801K wps
[Epoch 24 Batch 90/173] avg loss 0.000445608, throughput 3.8774K wps
[Epoch 24 Batch 120/173] avg loss 0.00046402, throughput 3.88421K wps
[Epoch 24 Batch 150/173] avg loss 0.000453496, throughput 3.90477K wps
Begin Testing...
[Epoch 24] train avg loss 0.000465521, test acc 0.7854, test avg loss 0.651882, throughput 3.9055K wps
[Epoch 25 Batch 30/173] avg loss 0.000379946, throughput 3.98049K wps
[Epoch 25 Batch 60/173] avg loss 0.000386771, throughput 3.86974K wps
[Epoch 25 Batch 90/173] avg loss 0.000410185, throughput 3.87506K wps
[Epoch 25 Batch 120/173] avg loss 0.000395101, throughput 3.87549K wps
[Epoch 25 Batch 150/173] avg loss 0.000414052, throughput 3.87273K wps
Begin Testing...
[Epoch 25] train avg loss 0.000405136, test acc 0.7833, test avg loss 0.670122, throughput 3.897K wps
[Epoch 26 Batch 30/173] avg loss 0.000336196, throughput 3.99616K wps
[Epoch 26 Batch 60/173] avg loss 0.000288854, throughput 3.87546K wps
[Epoch 26 Batch 90/173] avg loss 0.000340503, throughput 3.86647K wps
[Epoch 26 Batch 120/173] avg loss 0.000365961, throughput 3.87875K wps
[Epoch 26 Batch 150/173] avg loss 0.000320862, throughput 3.89842K wps
Begin Testing...
[Epoch 26] train avg loss 0.000326145, test acc 0.7854, test avg loss 0.689972, throughput 3.90324K wps
[Epoch 27 Batch 30/173] avg loss 0.000273761, throughput 3.99732K wps
[Epoch 27 Batch 60/173] avg loss 0.000231288, throughput 3.88428K wps
[Epoch 27 Batch 90/173] avg loss 0.000294166, throughput 3.88688K wps
[Epoch 27 Batch 120/173] avg loss 0.000263778, throughput 3.89468K wps
[Epoch 27 Batch 150/173] avg loss 0.000252849, throughput 3.8911K wps
Begin Testing...
[Epoch 27] train avg loss 0.000276972, test acc 0.7812, test avg loss 0.714058, throughput 3.90629K wps
[Epoch 28 Batch 30/173] avg loss 0.000270385, throughput 3.97743K wps
[Epoch 28 Batch 60/173] avg loss 0.000267791, throughput 3.87649K wps
[Epoch 28 Batch 90/173] avg loss 0.000251247, throughput 3.86301K wps
[Epoch 28 Batch 120/173] avg loss 0.000248029, throughput 3.89982K wps
[Epoch 28 Batch 150/173] avg loss 0.000254187, throughput 3.88641K wps
Begin Testing...
[Epoch 28] train avg loss 0.000259078, test acc 0.7802, test avg loss 0.741005, throughput 3.89662K wps
[Epoch 29 Batch 30/173] avg loss 0.000185916, throughput 3.98876K wps
[Epoch 29 Batch 60/173] avg loss 0.000225718, throughput 3.86758K wps
[Epoch 29 Batch 90/173] avg loss 0.000232713, throughput 3.86289K wps
[Epoch 29 Batch 120/173] avg loss 0.000210025, throughput 3.863K wps
[Epoch 29 Batch 150/173] avg loss 0.000237281, throughput 3.89003K wps
Begin Testing...
[Epoch 29] train avg loss 0.000215324, test acc 0.7760, test avg loss 0.760675, throughput 3.89663K wps
[Epoch 30 Batch 30/173] avg loss 0.000174608, throughput 3.97756K wps
[Epoch 30 Batch 60/173] avg loss 0.000198799, throughput 3.89377K wps
[Epoch 30 Batch 90/173] avg loss 0.000185196, throughput 3.88927K wps
[Epoch 30 Batch 120/173] avg loss 0.000213598, throughput 3.87847K wps
[Epoch 30 Batch 150/173] avg loss 0.000181518, throughput 3.87919K wps
Begin Testing...
[Epoch 30] train avg loss 0.000189808, test acc 0.7760, test avg loss 0.778269, throughput 3.90065K wps
[Epoch 31 Batch 30/173] avg loss 0.000162324, throughput 3.9623K wps
[Epoch 31 Batch 60/173] avg loss 0.000178321, throughput 3.87953K wps
[Epoch 31 Batch 90/173] avg loss 0.0001746, throughput 3.89245K wps
[Epoch 31 Batch 120/173] avg loss 0.000155073, throughput 3.9036K wps
[Epoch 31 Batch 150/173] avg loss 0.000183824, throughput 3.88081K wps
Begin Testing...
[Epoch 31] train avg loss 0.000169939, test acc 0.7719, test avg loss 0.807747, throughput 3.89795K wps
[Epoch 32 Batch 30/173] avg loss 0.000143587, throughput 3.97576K wps
[Epoch 32 Batch 60/173] avg loss 0.000130597, throughput 3.88996K wps
[Epoch 32 Batch 90/173] avg loss 0.000163066, throughput 3.89789K wps
[Epoch 32 Batch 120/173] avg loss 0.000124845, throughput 3.91571K wps
[Epoch 32 Batch 150/173] avg loss 0.000143563, throughput 3.8896K wps
Begin Testing...
[Epoch 32] train avg loss 0.000143185, test acc 0.7812, test avg loss 0.818988, throughput 3.91404K wps
[Epoch 33 Batch 30/173] avg loss 0.000125013, throughput 3.98663K wps
[Epoch 33 Batch 60/173] avg loss 0.000125358, throughput 3.89357K wps
[Epoch 33 Batch 90/173] avg loss 0.000144317, throughput 3.87728K wps
[Epoch 33 Batch 120/173] avg loss 0.000120815, throughput 3.874K wps
[Epoch 33 Batch 150/173] avg loss 0.000119635, throughput 3.88806K wps
Begin Testing...
[Epoch 33] train avg loss 0.000125159, test acc 0.7802, test avg loss 0.838856, throughput 3.90415K wps
[Epoch 34 Batch 30/173] avg loss 9.9225e-05, throughput 3.99145K wps
[Epoch 34 Batch 60/173] avg loss 9.00868e-05, throughput 3.87864K wps
[Epoch 34 Batch 90/173] avg loss 0.000107224, throughput 3.87344K wps
[Epoch 34 Batch 120/173] avg loss 0.00010996, throughput 3.87406K wps
[Epoch 34 Batch 150/173] avg loss 0.000110585, throughput 3.87542K wps
Begin Testing...
[Epoch 34] train avg loss 0.000103996, test acc 0.7823, test avg loss 0.864937, throughput 3.90048K wps
[Epoch 35 Batch 30/173] avg loss 8.66143e-05, throughput 3.98459K wps
[Epoch 35 Batch 60/173] avg loss 8.38539e-05, throughput 3.88629K wps
[Epoch 35 Batch 90/173] avg loss 0.000102834, throughput 3.88837K wps
[Epoch 35 Batch 120/173] avg loss 0.000107006, throughput 3.90674K wps
[Epoch 35 Batch 150/173] avg loss 0.000153714, throughput 3.89159K wps
Begin Testing...
[Epoch 35] train avg loss 0.000110587, test acc 0.7750, test avg loss 0.88249, throughput 3.90632K wps
[Epoch 36 Batch 30/173] avg loss 8.40423e-05, throughput 3.96251K wps
[Epoch 36 Batch 60/173] avg loss 8.36879e-05, throughput 3.86909K wps
[Epoch 36 Batch 90/173] avg loss 9.4877e-05, throughput 3.8694K wps
[Epoch 36 Batch 120/173] avg loss 8.45374e-05, throughput 3.88355K wps
[Epoch 36 Batch 150/173] avg loss 0.000102611, throughput 3.87952K wps
Begin Testing...
[Epoch 36] train avg loss 9.01892e-05, test acc 0.7802, test avg loss 0.912239, throughput 3.89344K wps
[Epoch 37 Batch 30/173] avg loss 8.2041e-05, throughput 3.98542K wps
[Epoch 37 Batch 60/173] avg loss 8.21721e-05, throughput 3.88587K wps
[Epoch 37 Batch 90/173] avg loss 8.02179e-05, throughput 3.86688K wps
[Epoch 37 Batch 120/173] avg loss 8.02745e-05, throughput 3.89205K wps
[Epoch 37 Batch 150/173] avg loss 0.000104308, throughput 3.90995K wps
Begin Testing...
[Epoch 37] train avg loss 8.71689e-05, test acc 0.7740, test avg loss 0.927909, throughput 3.90446K wps
[Epoch 38 Batch 30/173] avg loss 7.48429e-05, throughput 4.00013K wps
[Epoch 38 Batch 60/173] avg loss 7.4918e-05, throughput 3.89417K wps
[Epoch 38 Batch 90/173] avg loss 5.66071e-05, throughput 3.89825K wps
[Epoch 38 Batch 120/173] avg loss 6.93257e-05, throughput 3.89337K wps
[Epoch 38 Batch 150/173] avg loss 9.61737e-05, throughput 3.87761K wps
Begin Testing...
[Epoch 38] train avg loss 7.34758e-05, test acc 0.7781, test avg loss 0.947878, throughput 3.90791K wps
[Epoch 39 Batch 30/173] avg loss 6.36453e-05, throughput 3.97852K wps
[Epoch 39 Batch 60/173] avg loss 5.74873e-05, throughput 3.88035K wps
[Epoch 39 Batch 90/173] avg loss 5.69185e-05, throughput 3.87751K wps
[Epoch 39 Batch 120/173] avg loss 5.41691e-05, throughput 3.8902K wps
[Epoch 39 Batch 150/173] avg loss 8.02078e-05, throughput 3.89036K wps
Begin Testing...
[Epoch 39] train avg loss 6.25342e-05, test acc 0.7760, test avg loss 0.973397, throughput 3.89761K wps
[Epoch 40 Batch 30/173] avg loss 5.6063e-05, throughput 3.9726K wps
[Epoch 40 Batch 60/173] avg loss 4.87071e-05, throughput 3.89569K wps
[Epoch 40 Batch 90/173] avg loss 6.34695e-05, throughput 3.89514K wps
[Epoch 40 Batch 120/173] avg loss 5.59605e-05, throughput 3.91197K wps
[Epoch 40 Batch 150/173] avg loss 6.18977e-05, throughput 3.89976K wps
Begin Testing...
[Epoch 40] train avg loss 5.93139e-05, test acc 0.7708, test avg loss 1.0036, throughput 3.91304K wps
[Epoch 41 Batch 30/173] avg loss 4.90849e-05, throughput 3.97575K wps
[Epoch 41 Batch 60/173] avg loss 5.32894e-05, throughput 3.88689K wps
[Epoch 41 Batch 90/173] avg loss 6.29337e-05, throughput 3.90091K wps
[Epoch 41 Batch 120/173] avg loss 6.21239e-05, throughput 3.86074K wps
[Epoch 41 Batch 150/173] avg loss 4.27482e-05, throughput 3.8742K wps
Begin Testing...
[Epoch 41] train avg loss 5.50672e-05, test acc 0.7760, test avg loss 1.01467, throughput 3.89609K wps
[Epoch 42 Batch 30/173] avg loss 3.68608e-05, throughput 3.96933K wps
[Epoch 42 Batch 60/173] avg loss 4.15613e-05, throughput 3.88335K wps
[Epoch 42 Batch 90/173] avg loss 5.39146e-05, throughput 3.87481K wps
[Epoch 42 Batch 120/173] avg loss 5.62336e-05, throughput 3.86688K wps
[Epoch 42 Batch 150/173] avg loss 4.88347e-05, throughput 3.88309K wps
Begin Testing...
[Epoch 42] train avg loss 4.68323e-05, test acc 0.7771, test avg loss 1.0329, throughput 3.89571K wps
[Epoch 43 Batch 30/173] avg loss 3.92319e-05, throughput 3.99603K wps
[Epoch 43 Batch 60/173] avg loss 4.1575e-05, throughput 3.8872K wps
[Epoch 43 Batch 90/173] avg loss 4.18185e-05, throughput 3.9016K wps
[Epoch 43 Batch 120/173] avg loss 4.27953e-05, throughput 3.90569K wps
[Epoch 43 Batch 150/173] avg loss 5.42462e-05, throughput 3.87169K wps
Begin Testing...
[Epoch 43] train avg loss 4.42392e-05, test acc 0.7740, test avg loss 1.0452, throughput 3.90601K wps
[Epoch 44 Batch 30/173] avg loss 3.67983e-05, throughput 3.97138K wps
[Epoch 44 Batch 60/173] avg loss 3.62758e-05, throughput 3.87457K wps
[Epoch 44 Batch 90/173] avg loss 4.08186e-05, throughput 3.87584K wps
[Epoch 44 Batch 120/173] avg loss 2.79967e-05, throughput 3.90462K wps
[Epoch 44 Batch 150/173] avg loss 2.76609e-05, throughput 3.88569K wps
Begin Testing...
[Epoch 44] train avg loss 3.45705e-05, test acc 0.7708, test avg loss 1.07403, throughput 3.89864K wps
[Epoch 45 Batch 30/173] avg loss 3.19698e-05, throughput 3.97639K wps
[Epoch 45 Batch 60/173] avg loss 2.68125e-05, throughput 3.87187K wps
[Epoch 45 Batch 90/173] avg loss 4.05688e-05, throughput 3.8786K wps
[Epoch 45 Batch 120/173] avg loss 2.58502e-05, throughput 3.90417K wps
[Epoch 45 Batch 150/173] avg loss 2.91821e-05, throughput 3.90237K wps
Begin Testing...
[Epoch 45] train avg loss 3.12014e-05, test acc 0.7740, test avg loss 1.09509, throughput 3.90408K wps
[Epoch 46 Batch 30/173] avg loss 3.32245e-05, throughput 3.97845K wps
[Epoch 46 Batch 60/173] avg loss 4.0703e-05, throughput 3.88311K wps
[Epoch 46 Batch 90/173] avg loss 2.89701e-05, throughput 3.89085K wps
[Epoch 46 Batch 120/173] avg loss 3.25471e-05, throughput 3.88652K wps
[Epoch 46 Batch 150/173] avg loss 2.53308e-05, throughput 3.86699K wps
Begin Testing...
[Epoch 46] train avg loss 3.17166e-05, test acc 0.7708, test avg loss 1.11772, throughput 3.89538K wps
[Epoch 47 Batch 30/173] avg loss 2.70074e-05, throughput 3.97566K wps
[Epoch 47 Batch 60/173] avg loss 2.35594e-05, throughput 3.87641K wps
[Epoch 47 Batch 90/173] avg loss 3.06238e-05, throughput 3.87299K wps
[Epoch 47 Batch 120/173] avg loss 3.41051e-05, throughput 3.87697K wps
[Epoch 47 Batch 150/173] avg loss 2.47687e-05, throughput 3.87737K wps
Begin Testing...
[Epoch 47] train avg loss 2.8354e-05, test acc 0.7708, test avg loss 1.12733, throughput 3.89403K wps
[Epoch 48 Batch 30/173] avg loss 1.92392e-05, throughput 3.97585K wps
[Epoch 48 Batch 60/173] avg loss 2.65497e-05, throughput 3.91317K wps
[Epoch 48 Batch 90/173] avg loss 3.06349e-05, throughput 3.8909K wps
[Epoch 48 Batch 120/173] avg loss 5.09739e-05, throughput 3.89832K wps
[Epoch 48 Batch 150/173] avg loss 2.55267e-05, throughput 3.91437K wps
Begin Testing...
[Epoch 48] train avg loss 3.03292e-05, test acc 0.7677, test avg loss 1.13861, throughput 3.91263K wps
[Epoch 49 Batch 30/173] avg loss 3.76254e-05, throughput 3.97052K wps
[Epoch 49 Batch 60/173] avg loss 2.046e-05, throughput 3.87341K wps
[Epoch 49 Batch 90/173] avg loss 1.99374e-05, throughput 3.88144K wps
[Epoch 49 Batch 120/173] avg loss 2.42619e-05, throughput 3.88107K wps
[Epoch 49 Batch 150/173] avg loss 2.03e-05, throughput 3.89086K wps
Begin Testing...
[Epoch 49] train avg loss 2.49113e-05, test acc 0.7615, test avg loss 1.17433, throughput 3.89739K wps
[Epoch 50 Batch 30/173] avg loss 1.86183e-05, throughput 3.97785K wps
[Epoch 50 Batch 60/173] avg loss 2.3369e-05, throughput 3.89122K wps
[Epoch 50 Batch 90/173] avg loss 2.23022e-05, throughput 3.8736K wps
[Epoch 50 Batch 120/173] avg loss 2.03475e-05, throughput 3.86545K wps
[Epoch 50 Batch 150/173] avg loss 2.62821e-05, throughput 3.88166K wps
Begin Testing...
[Epoch 50] train avg loss 2.20642e-05, test acc 0.7677, test avg loss 1.1792, throughput 3.89974K wps
[Epoch 51 Batch 30/173] avg loss 1.70102e-05, throughput 3.98945K wps
[Epoch 51 Batch 60/173] avg loss 1.30155e-05, throughput 3.89044K wps
[Epoch 51 Batch 90/173] avg loss 2.43059e-05, throughput 3.88862K wps
[Epoch 51 Batch 120/173] avg loss 2.6283e-05, throughput 3.89748K wps
[Epoch 51 Batch 150/173] avg loss 2.15322e-05, throughput 3.89122K wps
Begin Testing...
[Epoch 51] train avg loss 2.19578e-05, test acc 0.7667, test avg loss 1.21043, throughput 3.90659K wps
[Epoch 52 Batch 30/173] avg loss 1.77005e-05, throughput 3.97888K wps
[Epoch 52 Batch 60/173] avg loss 1.67074e-05, throughput 3.87016K wps
[Epoch 52 Batch 90/173] avg loss 2.43866e-05, throughput 3.88189K wps
[Epoch 52 Batch 120/173] avg loss 1.51952e-05, throughput 3.88326K wps
[Epoch 52 Batch 150/173] avg loss 1.44342e-05, throughput 3.88995K wps
Begin Testing...
[Epoch 52] train avg loss 1.74616e-05, test acc 0.7688, test avg loss 1.22013, throughput 3.89826K wps
[Epoch 53 Batch 30/173] avg loss 1.5411e-05, throughput 3.97767K wps
[Epoch 53 Batch 60/173] avg loss 3.2372e-05, throughput 3.87468K wps
[Epoch 53 Batch 90/173] avg loss 2.96112e-05, throughput 3.88493K wps
[Epoch 53 Batch 120/173] avg loss 1.71517e-05, throughput 3.88779K wps
[Epoch 53 Batch 150/173] avg loss 1.44046e-05, throughput 3.91585K wps
Begin Testing...
[Epoch 53] train avg loss 2.07989e-05, test acc 0.7646, test avg loss 1.23916, throughput 3.90392K wps
[Epoch 54 Batch 30/173] avg loss 1.54702e-05, throughput 3.98086K wps
[Epoch 54 Batch 60/173] avg loss 1.15729e-05, throughput 3.88434K wps
[Epoch 54 Batch 90/173] avg loss 1.22304e-05, throughput 3.86973K wps
[Epoch 54 Batch 120/173] avg loss 1.1939e-05, throughput 3.89142K wps
[Epoch 54 Batch 150/173] avg loss 2.70942e-05, throughput 3.88351K wps
Begin Testing...
[Epoch 54] train avg loss 1.61066e-05, test acc 0.7635, test avg loss 1.24416, throughput 3.89866K wps
[Epoch 55 Batch 30/173] avg loss 1.10746e-05, throughput 3.97557K wps
[Epoch 55 Batch 60/173] avg loss 1.50145e-05, throughput 3.87756K wps
[Epoch 55 Batch 90/173] avg loss 1.03175e-05, throughput 3.88878K wps
[Epoch 55 Batch 120/173] avg loss 1.92808e-05, throughput 3.90834K wps
[Epoch 55 Batch 150/173] avg loss 1.50213e-05, throughput 3.87324K wps
Begin Testing...
[Epoch 55] train avg loss 1.36496e-05, test acc 0.7677, test avg loss 1.26902, throughput 3.90045K wps
[Epoch 56 Batch 30/173] avg loss 1.75631e-05, throughput 3.9892K wps
[Epoch 56 Batch 60/173] avg loss 1.61564e-05, throughput 3.8882K wps
[Epoch 56 Batch 90/173] avg loss 2.3464e-05, throughput 3.90726K wps
[Epoch 56 Batch 120/173] avg loss 1.34124e-05, throughput 3.90152K wps
[Epoch 56 Batch 150/173] avg loss 1.90032e-05, throughput 3.89028K wps
Begin Testing...
[Epoch 56] train avg loss 1.78772e-05, test acc 0.7646, test avg loss 1.29347, throughput 3.91303K wps
[Epoch 57 Batch 30/173] avg loss 9.17349e-06, throughput 3.97295K wps
[Epoch 57 Batch 60/173] avg loss 1.1036e-05, throughput 3.8864K wps
[Epoch 57 Batch 90/173] avg loss 1.08899e-05, throughput 3.87844K wps
[Epoch 57 Batch 120/173] avg loss 1.14201e-05, throughput 3.87747K wps
[Epoch 57 Batch 150/173] avg loss 1.55012e-05, throughput 3.87102K wps
Begin Testing...
[Epoch 57] train avg loss 1.14938e-05, test acc 0.7656, test avg loss 1.32553, throughput 3.89282K wps
[Epoch 58 Batch 30/173] avg loss 1.33664e-05, throughput 3.9657K wps
[Epoch 58 Batch 60/173] avg loss 1.07849e-05, throughput 3.8677K wps
[Epoch 58 Batch 90/173] avg loss 2.41821e-05, throughput 3.87673K wps
[Epoch 58 Batch 120/173] avg loss 1.25847e-05, throughput 3.87414K wps
[Epoch 58 Batch 150/173] avg loss 1.10211e-05, throughput 3.92184K wps
Begin Testing...
[Epoch 58] train avg loss 1.39588e-05, test acc 0.7667, test avg loss 1.33261, throughput 3.89747K wps
[Epoch 59 Batch 30/173] avg loss 1.86113e-05, throughput 3.99168K wps
[Epoch 59 Batch 60/173] avg loss 1.33863e-05, throughput 3.88721K wps
[Epoch 59 Batch 90/173] avg loss 8.64826e-06, throughput 3.91123K wps
[Epoch 59 Batch 120/173] avg loss 1.82698e-05, throughput 3.88894K wps
[Epoch 59 Batch 150/173] avg loss 1.01095e-05, throughput 3.87755K wps
Begin Testing...
[Epoch 59] train avg loss 1.33498e-05, test acc 0.7656, test avg loss 1.35175, throughput 3.9053K wps
Test loss 0.436703, test acc 0.7936
Total time cost 554.61s
9596 1066
[Epoch 0 Batch 30/173] avg loss 0.0152207, throughput 3.67064K wps
[Epoch 0 Batch 60/173] avg loss 0.0154197, throughput 3.88291K wps
[Epoch 0 Batch 90/173] avg loss 0.0145289, throughput 3.88735K wps
[Epoch 0 Batch 120/173] avg loss 0.0144761, throughput 3.88826K wps
[Epoch 0 Batch 150/173] avg loss 0.0140197, throughput 3.87513K wps
Begin Testing...
[Epoch 0] train avg loss 0.0147001, test acc 0.6062, test avg loss 0.64986, throughput 3.84408K wps
Observed Improvement.
Begin Testing...
[Epoch 1 Batch 30/173] avg loss 0.0133723, throughput 3.96156K wps
[Epoch 1 Batch 60/173] avg loss 0.0129288, throughput 3.88306K wps
[Epoch 1 Batch 90/173] avg loss 0.0129986, throughput 3.89773K wps
[Epoch 1 Batch 120/173] avg loss 0.0131099, throughput 3.89278K wps
[Epoch 1 Batch 150/173] avg loss 0.0128869, throughput 3.90637K wps
Begin Testing...
[Epoch 1] train avg loss 0.0130747, test acc 0.6865, test avg loss 0.620869, throughput 3.90495K wps
Observed Improvement.
Begin Testing...
[Epoch 2 Batch 30/173] avg loss 0.0123327, throughput 4.00564K wps
[Epoch 2 Batch 60/173] avg loss 0.0120235, throughput 3.86914K wps
[Epoch 2 Batch 90/173] avg loss 0.0119241, throughput 3.87244K wps
[Epoch 2 Batch 120/173] avg loss 0.0120275, throughput 3.8772K wps
[Epoch 2 Batch 150/173] avg loss 0.012174, throughput 3.86849K wps
Begin Testing...
[Epoch 2] train avg loss 0.012094, test acc 0.7198, test avg loss 0.597725, throughput 3.89748K wps
Observed Improvement.
Begin Testing...
[Epoch 3 Batch 30/173] avg loss 0.0114583, throughput 3.96609K wps
[Epoch 3 Batch 60/173] avg loss 0.0112813, throughput 3.87957K wps
[Epoch 3 Batch 90/173] avg loss 0.0112856, throughput 3.88874K wps
[Epoch 3 Batch 120/173] avg loss 0.0109737, throughput 3.88731K wps
[Epoch 3 Batch 150/173] avg loss 0.0108761, throughput 3.88041K wps
Begin Testing...
[Epoch 3] train avg loss 0.0111534, test acc 0.7448, test avg loss 0.563338, throughput 3.89431K wps
Observed Improvement.
Begin Testing...
[Epoch 4 Batch 30/173] avg loss 0.0104149, throughput 4.0068K wps
[Epoch 4 Batch 60/173] avg loss 0.0104004, throughput 3.88504K wps
[Epoch 4 Batch 90/173] avg loss 0.0100116, throughput 3.87644K wps
[Epoch 4 Batch 120/173] avg loss 0.00999906, throughput 3.89367K wps
[Epoch 4 Batch 150/173] avg loss 0.0102818, throughput 3.90824K wps
Begin Testing...
[Epoch 4] train avg loss 0.0101338, test acc 0.7698, test avg loss 0.52631, throughput 3.90965K wps
Observed Improvement.
Begin Testing...
[Epoch 5 Batch 30/173] avg loss 0.00930186, throughput 3.9813K wps
[Epoch 5 Batch 60/173] avg loss 0.00930695, throughput 3.86719K wps
[Epoch 5 Batch 90/173] avg loss 0.00919353, throughput 3.86995K wps
[Epoch 5 Batch 120/173] avg loss 0.00901354, throughput 3.88354K wps
[Epoch 5 Batch 150/173] avg loss 0.0087592, throughput 3.87194K wps
Begin Testing...
[Epoch 5] train avg loss 0.00910246, test acc 0.7698, test avg loss 0.493443, throughput 3.8954K wps
Observed Improvement.
Begin Testing...
[Epoch 6 Batch 30/173] avg loss 0.00832069, throughput 3.96844K wps
[Epoch 6 Batch 60/173] avg loss 0.00804643, throughput 3.89103K wps
[Epoch 6 Batch 90/173] avg loss 0.00808698, throughput 3.87159K wps
[Epoch 6 Batch 120/173] avg loss 0.00786729, throughput 3.88462K wps
[Epoch 6 Batch 150/173] avg loss 0.00781521, throughput 3.89137K wps
Begin Testing...
[Epoch 6] train avg loss 0.00800796, test acc 0.7656, test avg loss 0.47645, throughput 3.90245K wps
[Epoch 7 Batch 30/173] avg loss 0.00737615, throughput 4.00126K wps
[Epoch 7 Batch 60/173] avg loss 0.00732365, throughput 3.88726K wps
[Epoch 7 Batch 90/173] avg loss 0.0073219, throughput 3.88005K wps
[Epoch 7 Batch 120/173] avg loss 0.00713147, throughput 3.86639K wps
[Epoch 7 Batch 150/173] avg loss 0.00681091, throughput 3.8716K wps
Begin Testing...
[Epoch 7] train avg loss 0.00713286, test acc 0.7812, test avg loss 0.44747, throughput 3.89946K wps
Observed Improvement.
Begin Testing...
[Epoch 8 Batch 30/173] avg loss 0.00629355, throughput 3.97592K wps
[Epoch 8 Batch 60/173] avg loss 0.00622392, throughput 3.87834K wps
[Epoch 8 Batch 90/173] avg loss 0.00633213, throughput 3.87126K wps
[Epoch 8 Batch 120/173] avg loss 0.00647829, throughput 3.87626K wps
[Epoch 8 Batch 150/173] avg loss 0.00602554, throughput 3.89455K wps
Begin Testing...
[Epoch 8] train avg loss 0.00623946, test acc 0.7906, test avg loss 0.435569, throughput 3.8975K wps
Observed Improvement.
Begin Testing...
[Epoch 9 Batch 30/173] avg loss 0.00535103, throughput 3.96863K wps
[Epoch 9 Batch 60/173] avg loss 0.00533304, throughput 3.89412K wps
[Epoch 9 Batch 90/173] avg loss 0.00556658, throughput 3.91568K wps
[Epoch 9 Batch 120/173] avg loss 0.00540147, throughput 3.89792K wps
[Epoch 9 Batch 150/173] avg loss 0.00558661, throughput 3.88772K wps
Begin Testing...
[Epoch 9] train avg loss 0.00547854, test acc 0.7812, test avg loss 0.429122, throughput 3.91318K wps
[Epoch 10 Batch 30/173] avg loss 0.00499246, throughput 3.98484K wps
[Epoch 10 Batch 60/173] avg loss 0.00464331, throughput 3.88848K wps
[Epoch 10 Batch 90/173] avg loss 0.0047001, throughput 3.86825K wps
[Epoch 10 Batch 120/173] avg loss 0.00461158, throughput 3.87226K wps
[Epoch 10 Batch 150/173] avg loss 0.00490131, throughput 3.88677K wps
Begin Testing...
[Epoch 10] train avg loss 0.0047897, test acc 0.7885, test avg loss 0.432253, throughput 3.90035K wps
[Epoch 11 Batch 30/173] avg loss 0.00458673, throughput 3.98731K wps
[Epoch 11 Batch 60/173] avg loss 0.00390871, throughput 3.87741K wps
[Epoch 11 Batch 90/173] avg loss 0.00417923, throughput 3.8715K wps
[Epoch 11 Batch 120/173] avg loss 0.00420609, throughput 3.8704K wps
[Epoch 11 Batch 150/173] avg loss 0.00417324, throughput 3.87266K wps
Begin Testing...
[Epoch 11] train avg loss 0.00418616, test acc 0.7917, test avg loss 0.426875, throughput 3.8974K wps
Observed Improvement.
Begin Testing...
[Epoch 12 Batch 30/173] avg loss 0.00348752, throughput 3.98496K wps
[Epoch 12 Batch 60/173] avg loss 0.00365549, throughput 3.89654K wps
[Epoch 12 Batch 90/173] avg loss 0.00354852, throughput 3.893K wps
[Epoch 12 Batch 120/173] avg loss 0.00362063, throughput 3.88529K wps
[Epoch 12 Batch 150/173] avg loss 0.00354059, throughput 3.87943K wps
Begin Testing...
[Epoch 12] train avg loss 0.00358435, test acc 0.7875, test avg loss 0.428697, throughput 3.90631K wps
[Epoch 13 Batch 30/173] avg loss 0.00311094, throughput 3.96448K wps
[Epoch 13 Batch 60/173] avg loss 0.00320733, throughput 3.87799K wps
[Epoch 13 Batch 90/173] avg loss 0.00283781, throughput 3.89097K wps
[Epoch 13 Batch 120/173] avg loss 0.00300321, throughput 3.87601K wps
[Epoch 13 Batch 150/173] avg loss 0.0029028, throughput 3.87505K wps
Begin Testing...
[Epoch 13] train avg loss 0.00304297, test acc 0.7833, test avg loss 0.430735, throughput 3.89312K wps
[Epoch 14 Batch 30/173] avg loss 0.0026374, throughput 3.96187K wps
[Epoch 14 Batch 60/173] avg loss 0.00258485, throughput 3.87074K wps
[Epoch 14 Batch 90/173] avg loss 0.00268249, throughput 3.88582K wps
[Epoch 14 Batch 120/173] avg loss 0.00241569, throughput 3.8857K wps
[Epoch 14 Batch 150/173] avg loss 0.00269753, throughput 3.88298K wps
Begin Testing...
[Epoch 14] train avg loss 0.00260874, test acc 0.7885, test avg loss 0.429803, throughput 3.89873K wps
[Epoch 15 Batch 30/173] avg loss 0.00221085, throughput 4.00071K wps
[Epoch 15 Batch 60/173] avg loss 0.00234381, throughput 3.87643K wps
[Epoch 15 Batch 90/173] avg loss 0.00211121, throughput 3.88015K wps
[Epoch 15 Batch 120/173] avg loss 0.00211488, throughput 3.87298K wps
[Epoch 15 Batch 150/173] avg loss 0.00224122, throughput 3.88051K wps
Begin Testing...
[Epoch 15] train avg loss 0.00220392, test acc 0.7812, test avg loss 0.443768, throughput 3.89978K wps
[Epoch 16 Batch 30/173] avg loss 0.00169125, throughput 3.98341K wps
[Epoch 16 Batch 60/173] avg loss 0.00193485, throughput 3.87965K wps
[Epoch 16 Batch 90/173] avg loss 0.00198446, throughput 3.87724K wps
[Epoch 16 Batch 120/173] avg loss 0.0018726, throughput 3.88095K wps
[Epoch 16 Batch 150/173] avg loss 0.00189114, throughput 3.87861K wps
Begin Testing...
[Epoch 16] train avg loss 0.00189088, test acc 0.7917, test avg loss 0.45576, throughput 3.89924K wps
Observed Improvement.
Begin Testing...
[Epoch 17 Batch 30/173] avg loss 0.00154652, throughput 3.97991K wps
[Epoch 17 Batch 60/173] avg loss 0.00164679, throughput 3.91321K wps
[Epoch 17 Batch 90/173] avg loss 0.00152901, throughput 3.89069K wps
[Epoch 17 Batch 120/173] avg loss 0.00158544, throughput 3.88929K wps
[Epoch 17 Batch 150/173] avg loss 0.00148408, throughput 3.8948K wps
Begin Testing...
[Epoch 17] train avg loss 0.00156831, test acc 0.7927, test avg loss 0.471038, throughput 3.91014K wps
Observed Improvement.
Begin Testing...
[Epoch 18 Batch 30/173] avg loss 0.00142907, throughput 3.96829K wps
[Epoch 18 Batch 60/173] avg loss 0.00127367, throughput 3.88169K wps
[Epoch 18 Batch 90/173] avg loss 0.00134759, throughput 3.88601K wps
[Epoch 18 Batch 120/173] avg loss 0.00136881, throughput 3.87618K wps
[Epoch 18 Batch 150/173] avg loss 0.00123192, throughput 3.8744K wps
Begin Testing...
[Epoch 18] train avg loss 0.00135265, test acc 0.7885, test avg loss 0.481688, throughput 3.89571K wps
[Epoch 19 Batch 30/173] avg loss 0.00115957, throughput 3.96597K wps
[Epoch 19 Batch 60/173] avg loss 0.00122449, throughput 3.86414K wps
[Epoch 19 Batch 90/173] avg loss 0.00120099, throughput 3.87065K wps
[Epoch 19 Batch 120/173] avg loss 0.00104074, throughput 3.89376K wps
[Epoch 19 Batch 150/173] avg loss 0.00101376, throughput 3.90968K wps
Begin Testing...
[Epoch 19] train avg loss 0.00112862, test acc 0.7854, test avg loss 0.499412, throughput 3.89751K wps
[Epoch 20 Batch 30/173] avg loss 0.000984417, throughput 3.98215K wps
[Epoch 20 Batch 60/173] avg loss 0.000857137, throughput 3.88748K wps
[Epoch 20 Batch 90/173] avg loss 0.000923408, throughput 3.87296K wps
[Epoch 20 Batch 120/173] avg loss 0.000870605, throughput 3.87981K wps
[Epoch 20 Batch 150/173] avg loss 0.00104432, throughput 3.8682K wps
Begin Testing...
[Epoch 20] train avg loss 0.000954553, test acc 0.7823, test avg loss 0.512576, throughput 3.89459K wps
[Epoch 21 Batch 30/173] avg loss 0.000717737, throughput 3.97407K wps
[Epoch 21 Batch 60/173] avg loss 0.00084297, throughput 3.89404K wps
[Epoch 21 Batch 90/173] avg loss 0.000764595, throughput 3.88413K wps
[Epoch 21 Batch 120/173] avg loss 0.000845878, throughput 3.8811K wps
[Epoch 21 Batch 150/173] avg loss 0.000904835, throughput 3.87477K wps
Begin Testing...
[Epoch 21] train avg loss 0.00081021, test acc 0.7802, test avg loss 0.530592, throughput 3.90044K wps
[Epoch 22 Batch 30/173] avg loss 0.00068504, throughput 3.96699K wps
[Epoch 22 Batch 60/173] avg loss 0.000596769, throughput 3.91619K wps
[Epoch 22 Batch 90/173] avg loss 0.000558693, throughput 3.88534K wps
[Epoch 22 Batch 120/173] avg loss 0.00075623, throughput 3.87625K wps
[Epoch 22 Batch 150/173] avg loss 0.000666561, throughput 3.91643K wps
Begin Testing...
[Epoch 22] train avg loss 0.000671435, test acc 0.7812, test avg loss 0.55384, throughput 3.90824K wps
[Epoch 23 Batch 30/173] avg loss 0.000530321, throughput 3.9853K wps
[Epoch 23 Batch 60/173] avg loss 0.000555574, throughput 3.88134K wps
[Epoch 23 Batch 90/173] avg loss 0.000596903, throughput 3.87208K wps
[Epoch 23 Batch 120/173] avg loss 0.000674315, throughput 3.87439K wps
[Epoch 23 Batch 150/173] avg loss 0.000649473, throughput 3.89166K wps
Begin Testing...
[Epoch 23] train avg loss 0.000598946, test acc 0.7740, test avg loss 0.572441, throughput 3.8989K wps
[Epoch 24 Batch 30/173] avg loss 0.00045127, throughput 3.98691K wps
[Epoch 24 Batch 60/173] avg loss 0.000551797, throughput 3.88203K wps
[Epoch 24 Batch 90/173] avg loss 0.000499493, throughput 3.8765K wps
[Epoch 24 Batch 120/173] avg loss 0.000495215, throughput 3.87983K wps
[Epoch 24 Batch 150/173] avg loss 0.000526411, throughput 3.88934K wps
Begin Testing...
[Epoch 24] train avg loss 0.000512315, test acc 0.7760, test avg loss 0.591827, throughput 3.90325K wps
[Epoch 25 Batch 30/173] avg loss 0.000448111, throughput 3.99594K wps
[Epoch 25 Batch 60/173] avg loss 0.000501526, throughput 3.88443K wps
[Epoch 25 Batch 90/173] avg loss 0.000394852, throughput 3.8891K wps
[Epoch 25 Batch 120/173] avg loss 0.000449927, throughput 3.89775K wps
[Epoch 25 Batch 150/173] avg loss 0.000418044, throughput 3.87703K wps
Begin Testing...
[Epoch 25] train avg loss 0.000438782, test acc 0.7771, test avg loss 0.609996, throughput 3.90338K wps
[Epoch 26 Batch 30/173] avg loss 0.00040328, throughput 3.97678K wps
[Epoch 26 Batch 60/173] avg loss 0.000368752, throughput 3.87326K wps
[Epoch 26 Batch 90/173] avg loss 0.000370029, throughput 3.87859K wps
[Epoch 26 Batch 120/173] avg loss 0.000418824, throughput 3.87727K wps
[Epoch 26 Batch 150/173] avg loss 0.000364919, throughput 3.88107K wps
Begin Testing...
[Epoch 26] train avg loss 0.000380928, test acc 0.7750, test avg loss 0.630056, throughput 3.8969K wps
[Epoch 27 Batch 30/173] avg loss 0.000280991, throughput 3.96814K wps
[Epoch 27 Batch 60/173] avg loss 0.000317298, throughput 3.87613K wps
[Epoch 27 Batch 90/173] avg loss 0.000310905, throughput 3.89687K wps
[Epoch 27 Batch 120/173] avg loss 0.00031001, throughput 3.88834K wps
[Epoch 27 Batch 150/173] avg loss 0.000338886, throughput 3.89038K wps
Begin Testing...
[Epoch 27] train avg loss 0.000314821, test acc 0.7688, test avg loss 0.646002, throughput 3.90152K wps
[Epoch 28 Batch 30/173] avg loss 0.000260538, throughput 3.98102K wps
[Epoch 28 Batch 60/173] avg loss 0.000248816, throughput 3.90526K wps
[Epoch 28 Batch 90/173] avg loss 0.00026205, throughput 3.87651K wps
[Epoch 28 Batch 120/173] avg loss 0.000298556, throughput 3.87927K wps
[Epoch 28 Batch 150/173] avg loss 0.000318005, throughput 3.87594K wps
Begin Testing...
[Epoch 28] train avg loss 0.000273378, test acc 0.7677, test avg loss 0.663448, throughput 3.90326K wps
[Epoch 29 Batch 30/173] avg loss 0.00023902, throughput 3.98065K wps
[Epoch 29 Batch 60/173] avg loss 0.000251417, throughput 3.8883K wps
[Epoch 29 Batch 90/173] avg loss 0.000227798, throughput 3.88287K wps
[Epoch 29 Batch 120/173] avg loss 0.000233099, throughput 3.88329K wps
[Epoch 29 Batch 150/173] avg loss 0.000272606, throughput 3.87354K wps
Begin Testing...
[Epoch 29] train avg loss 0.000244729, test acc 0.7615, test avg loss 0.682463, throughput 3.90127K wps
[Epoch 30 Batch 30/173] avg loss 0.000204957, throughput 3.97943K wps
[Epoch 30 Batch 60/173] avg loss 0.000201801, throughput 3.91802K wps
[Epoch 30 Batch 90/173] avg loss 0.000222823, throughput 3.89445K wps
[Epoch 30 Batch 120/173] avg loss 0.00021048, throughput 3.8858K wps
[Epoch 30 Batch 150/173] avg loss 0.000202136, throughput 3.91942K wps
Begin Testing...
[Epoch 30] train avg loss 0.000208252, test acc 0.7677, test avg loss 0.700523, throughput 3.91407K wps
[Epoch 31 Batch 30/173] avg loss 0.000164882, throughput 3.97696K wps
[Epoch 31 Batch 60/173] avg loss 0.0002025, throughput 3.87951K wps
[Epoch 31 Batch 90/173] avg loss 0.000167338, throughput 3.87207K wps
[Epoch 31 Batch 120/173] avg loss 0.000169257, throughput 3.87072K wps
[Epoch 31 Batch 150/173] avg loss 0.000197363, throughput 3.88612K wps
Begin Testing...
[Epoch 31] train avg loss 0.000184297, test acc 0.7646, test avg loss 0.724188, throughput 3.89597K wps
[Epoch 32 Batch 30/173] avg loss 0.000164625, throughput 3.97317K wps
[Epoch 32 Batch 60/173] avg loss 0.000174755, throughput 3.8827K wps
[Epoch 32 Batch 90/173] avg loss 0.00018941, throughput 3.86932K wps
[Epoch 32 Batch 120/173] avg loss 0.000171995, throughput 3.87448K wps
[Epoch 32 Batch 150/173] avg loss 0.000156733, throughput 3.88328K wps
Begin Testing...
[Epoch 32] train avg loss 0.00016956, test acc 0.7729, test avg loss 0.738982, throughput 3.89851K wps
[Epoch 33 Batch 30/173] avg loss 0.000134508, throughput 3.99207K wps
[Epoch 33 Batch 60/173] avg loss 0.000130834, throughput 3.88615K wps
[Epoch 33 Batch 90/173] avg loss 0.000154851, throughput 3.87849K wps
[Epoch 33 Batch 120/173] avg loss 0.000142548, throughput 3.88117K wps
[Epoch 33 Batch 150/173] avg loss 0.000151221, throughput 3.88492K wps
Begin Testing...
[Epoch 33] train avg loss 0.000145881, test acc 0.7646, test avg loss 0.763186, throughput 3.90184K wps
[Epoch 34 Batch 30/173] avg loss 0.00013623, throughput 3.99033K wps
[Epoch 34 Batch 60/173] avg loss 0.00012209, throughput 3.86985K wps
[Epoch 34 Batch 90/173] avg loss 0.000134735, throughput 3.87967K wps
[Epoch 34 Batch 120/173] avg loss 0.000121967, throughput 3.87699K wps
[Epoch 34 Batch 150/173] avg loss 0.000139348, throughput 3.8862K wps
Begin Testing...
[Epoch 34] train avg loss 0.000134449, test acc 0.7656, test avg loss 0.780415, throughput 3.90007K wps
[Epoch 35 Batch 30/173] avg loss 0.000132676, throughput 3.98771K wps
[Epoch 35 Batch 60/173] avg loss 0.000103544, throughput 3.87728K wps
[Epoch 35 Batch 90/173] avg loss 9.73802e-05, throughput 3.877K wps
[Epoch 35 Batch 120/173] avg loss 0.000115368, throughput 3.87462K wps
[Epoch 35 Batch 150/173] avg loss 9.48065e-05, throughput 3.8914K wps
Begin Testing...
[Epoch 35] train avg loss 0.000108266, test acc 0.7604, test avg loss 0.798836, throughput 3.90075K wps
[Epoch 36 Batch 30/173] avg loss 0.000103468, throughput 3.99672K wps
[Epoch 36 Batch 60/173] avg loss 0.000106375, throughput 3.88347K wps
[Epoch 36 Batch 90/173] avg loss 8.79255e-05, throughput 3.90629K wps
[Epoch 36 Batch 120/173] avg loss 9.66171e-05, throughput 3.90257K wps
[Epoch 36 Batch 150/173] avg loss 7.99233e-05, throughput 3.88156K wps
Begin Testing...
[Epoch 36] train avg loss 9.69908e-05, test acc 0.7615, test avg loss 0.817171, throughput 3.90876K wps
[Epoch 37 Batch 30/173] avg loss 7.86558e-05, throughput 3.96968K wps
[Epoch 37 Batch 60/173] avg loss 7.18843e-05, throughput 3.87858K wps
[Epoch 37 Batch 90/173] avg loss 0.000101281, throughput 3.8813K wps
[Epoch 37 Batch 120/173] avg loss 0.000100165, throughput 3.89039K wps
[Epoch 37 Batch 150/173] avg loss 0.000102278, throughput 3.8803K wps
Begin Testing...
[Epoch 37] train avg loss 9.16587e-05, test acc 0.7646, test avg loss 0.839478, throughput 3.89713K wps
[Epoch 38 Batch 30/173] avg loss 7.87455e-05, throughput 3.96309K wps
[Epoch 38 Batch 60/173] avg loss 7.66208e-05, throughput 3.87732K wps
[Epoch 38 Batch 90/173] avg loss 7.29528e-05, throughput 3.87997K wps
[Epoch 38 Batch 120/173] avg loss 6.57789e-05, throughput 3.8877K wps
[Epoch 38 Batch 150/173] avg loss 9.21406e-05, throughput 3.90948K wps
Begin Testing...
[Epoch 38] train avg loss 7.55016e-05, test acc 0.7573, test avg loss 0.85458, throughput 3.90055K wps
[Epoch 39 Batch 30/173] avg loss 7.21554e-05, throughput 3.98574K wps
[Epoch 39 Batch 60/173] avg loss 7.91432e-05, throughput 3.87792K wps
[Epoch 39 Batch 90/173] avg loss 8.4032e-05, throughput 3.87317K wps
[Epoch 39 Batch 120/173] avg loss 8.44221e-05, throughput 3.88994K wps
[Epoch 39 Batch 150/173] avg loss 6.53961e-05, throughput 3.87055K wps
Begin Testing...
[Epoch 39] train avg loss 7.77425e-05, test acc 0.7531, test avg loss 0.881466, throughput 3.89658K wps
[Epoch 40 Batch 30/173] avg loss 6.06591e-05, throughput 3.98328K wps
[Epoch 40 Batch 60/173] avg loss 5.86784e-05, throughput 3.8712K wps
[Epoch 40 Batch 90/173] avg loss 6.39912e-05, throughput 3.8705K wps
[Epoch 40 Batch 120/173] avg loss 5.53288e-05, throughput 3.87001K wps
[Epoch 40 Batch 150/173] avg loss 6.43725e-05, throughput 3.8722K wps
Begin Testing...
[Epoch 40] train avg loss 6.22645e-05, test acc 0.7531, test avg loss 0.89908, throughput 3.8916K wps
[Epoch 41 Batch 30/173] avg loss 5.02082e-05, throughput 3.97263K wps
[Epoch 41 Batch 60/173] avg loss 4.09002e-05, throughput 3.89987K wps
[Epoch 41 Batch 90/173] avg loss 5.52535e-05, throughput 3.90248K wps
[Epoch 41 Batch 120/173] avg loss 5.84533e-05, throughput 3.87784K wps
[Epoch 41 Batch 150/173] avg loss 5.15932e-05, throughput 3.88646K wps
Begin Testing...
[Epoch 41] train avg loss 5.26155e-05, test acc 0.7531, test avg loss 0.91606, throughput 3.90814K wps
[Epoch 42 Batch 30/173] avg loss 3.75389e-05, throughput 3.98394K wps
[Epoch 42 Batch 60/173] avg loss 4.71025e-05, throughput 3.89513K wps
[Epoch 42 Batch 90/173] avg loss 5.86883e-05, throughput 3.87183K wps
[Epoch 42 Batch 120/173] avg loss 5.17523e-05, throughput 3.87978K wps
[Epoch 42 Batch 150/173] avg loss 4.7767e-05, throughput 3.88052K wps
Begin Testing...
[Epoch 42] train avg loss 5.14737e-05, test acc 0.7552, test avg loss 0.936333, throughput 3.90243K wps
[Epoch 43 Batch 30/173] avg loss 4.82955e-05, throughput 3.98175K wps
[Epoch 43 Batch 60/173] avg loss 3.93413e-05, throughput 3.90183K wps
[Epoch 43 Batch 90/173] avg loss 4.9471e-05, throughput 3.86849K wps
[Epoch 43 Batch 120/173] avg loss 4.83557e-05, throughput 3.86447K wps
[Epoch 43 Batch 150/173] avg loss 3.63305e-05, throughput 3.88636K wps
Begin Testing...
[Epoch 43] train avg loss 4.42197e-05, test acc 0.7562, test avg loss 0.951895, throughput 3.90186K wps
[Epoch 44 Batch 30/173] avg loss 3.12721e-05, throughput 4.01361K wps
[Epoch 44 Batch 60/173] avg loss 4.11314e-05, throughput 3.88471K wps
[Epoch 44 Batch 90/173] avg loss 3.58128e-05, throughput 3.89277K wps
[Epoch 44 Batch 120/173] avg loss 3.48145e-05, throughput 3.91008K wps
[Epoch 44 Batch 150/173] avg loss 5.30924e-05, throughput 3.88538K wps
Begin Testing...
[Epoch 44] train avg loss 4.04332e-05, test acc 0.7594, test avg loss 0.953165, throughput 3.91043K wps
[Epoch 45 Batch 30/173] avg loss 2.79959e-05, throughput 3.97529K wps
[Epoch 45 Batch 60/173] avg loss 4.1728e-05, throughput 3.87033K wps
[Epoch 45 Batch 90/173] avg loss 2.66075e-05, throughput 3.88188K wps
[Epoch 45 Batch 120/173] avg loss 3.99101e-05, throughput 3.88219K wps
[Epoch 45 Batch 150/173] avg loss 3.9318e-05, throughput 3.89461K wps
Begin Testing...
[Epoch 45] train avg loss 3.51182e-05, test acc 0.7542, test avg loss 0.98203, throughput 3.89744K wps
[Epoch 46 Batch 30/173] avg loss 2.70983e-05, throughput 3.9791K wps
[Epoch 46 Batch 60/173] avg loss 3.36491e-05, throughput 3.86295K wps
[Epoch 46 Batch 90/173] avg loss 4.51537e-05, throughput 3.87631K wps
[Epoch 46 Batch 120/173] avg loss 3.51403e-05, throughput 3.88922K wps
[Epoch 46 Batch 150/173] avg loss 3.9552e-05, throughput 3.90481K wps
Begin Testing...
[Epoch 46] train avg loss 3.54627e-05, test acc 0.7479, test avg loss 1.01005, throughput 3.90239K wps
[Epoch 47 Batch 30/173] avg loss 2.46989e-05, throughput 3.98363K wps
[Epoch 47 Batch 60/173] avg loss 2.84184e-05, throughput 3.88992K wps
[Epoch 47 Batch 90/173] avg loss 2.75964e-05, throughput 3.87403K wps
[Epoch 47 Batch 120/173] avg loss 3.21166e-05, throughput 3.84117K wps
[Epoch 47 Batch 150/173] avg loss 3.45152e-05, throughput 3.87192K wps
Begin Testing...
[Epoch 47] train avg loss 2.96429e-05, test acc 0.7510, test avg loss 1.03059, throughput 3.89224K wps
[Epoch 48 Batch 30/173] avg loss 2.80233e-05, throughput 3.958K wps
[Epoch 48 Batch 60/173] avg loss 2.8072e-05, throughput 3.8703K wps
[Epoch 48 Batch 90/173] avg loss 2.30354e-05, throughput 3.8866K wps
[Epoch 48 Batch 120/173] avg loss 2.3097e-05, throughput 3.88308K wps
[Epoch 48 Batch 150/173] avg loss 2.8894e-05, throughput 3.87501K wps
Begin Testing...
[Epoch 48] train avg loss 2.58503e-05, test acc 0.7521, test avg loss 1.05419, throughput 3.89128K wps
[Epoch 49 Batch 30/173] avg loss 2.44326e-05, throughput 3.97904K wps
[Epoch 49 Batch 60/173] avg loss 5.01902e-05, throughput 3.87691K wps
[Epoch 49 Batch 90/173] avg loss 3.31344e-05, throughput 3.89053K wps
[Epoch 49 Batch 120/173] avg loss 3.36448e-05, throughput 3.90358K wps
[Epoch 49 Batch 150/173] avg loss 4.01652e-05, throughput 3.90659K wps
Begin Testing...
[Epoch 49] train avg loss 3.49665e-05, test acc 0.7510, test avg loss 1.07607, throughput 3.90806K wps
[Epoch 50 Batch 30/173] avg loss 2.04711e-05, throughput 3.96938K wps
[Epoch 50 Batch 60/173] avg loss 2.1695e-05, throughput 3.88654K wps
[Epoch 50 Batch 90/173] avg loss 2.29378e-05, throughput 3.88151K wps
[Epoch 50 Batch 120/173] avg loss 1.9476e-05, throughput 3.88696K wps
[Epoch 50 Batch 150/173] avg loss 2.82863e-05, throughput 3.88257K wps
Begin Testing...
[Epoch 50] train avg loss 2.52994e-05, test acc 0.7510, test avg loss 1.08186, throughput 3.8966K wps
[Epoch 51 Batch 30/173] avg loss 1.50081e-05, throughput 3.97806K wps
[Epoch 51 Batch 60/173] avg loss 2.46591e-05, throughput 3.87702K wps
[Epoch 51 Batch 90/173] avg loss 1.76934e-05, throughput 3.8761K wps
[Epoch 51 Batch 120/173] avg loss 2.17689e-05, throughput 3.86655K wps
[Epoch 51 Batch 150/173] avg loss 1.8923e-05, throughput 3.88639K wps
Begin Testing...
[Epoch 51] train avg loss 1.93454e-05, test acc 0.7510, test avg loss 1.09744, throughput 3.89579K wps
[Epoch 52 Batch 30/173] avg loss 1.85628e-05, throughput 4.00388K wps
[Epoch 52 Batch 60/173] avg loss 2.25766e-05, throughput 3.88986K wps
[Epoch 52 Batch 90/173] avg loss 1.84142e-05, throughput 3.89496K wps
[Epoch 52 Batch 120/173] avg loss 1.52067e-05, throughput 3.9048K wps
[Epoch 52 Batch 150/173] avg loss 3.16771e-05, throughput 3.87861K wps
Begin Testing...
[Epoch 52] train avg loss 2.15711e-05, test acc 0.7469, test avg loss 1.10105, throughput 3.91083K wps
[Epoch 53 Batch 30/173] avg loss 2.09682e-05, throughput 3.96062K wps
[Epoch 53 Batch 60/173] avg loss 2.10788e-05, throughput 3.86794K wps
[Epoch 53 Batch 90/173] avg loss 1.62685e-05, throughput 3.87396K wps
[Epoch 53 Batch 120/173] avg loss 1.74421e-05, throughput 3.87195K wps
[Epoch 53 Batch 150/173] avg loss 1.48762e-05, throughput 3.89001K wps
Begin Testing...
[Epoch 53] train avg loss 1.89712e-05, test acc 0.7500, test avg loss 1.12414, throughput 3.89087K wps
[Epoch 54 Batch 30/173] avg loss 1.14972e-05, throughput 3.96799K wps
[Epoch 54 Batch 60/173] avg loss 2.28185e-05, throughput 3.86895K wps
[Epoch 54 Batch 90/173] avg loss 1.88422e-05, throughput 3.86353K wps
[Epoch 54 Batch 120/173] avg loss 1.76251e-05, throughput 3.88569K wps
[Epoch 54 Batch 150/173] avg loss 1.52201e-05, throughput 3.88679K wps
Begin Testing...
[Epoch 54] train avg loss 1.68532e-05, test acc 0.7500, test avg loss 1.13947, throughput 3.89662K wps
[Epoch 55 Batch 30/173] avg loss 2.09563e-05, throughput 4.00686K wps
[Epoch 55 Batch 60/173] avg loss 1.34961e-05, throughput 3.8906K wps
[Epoch 55 Batch 90/173] avg loss 1.79942e-05, throughput 3.88151K wps
[Epoch 55 Batch 120/173] avg loss 1.22745e-05, throughput 3.8751K wps
[Epoch 55 Batch 150/173] avg loss 1.77326e-05, throughput 3.88523K wps
Begin Testing...
[Epoch 55] train avg loss 1.58515e-05, test acc 0.7448, test avg loss 1.19847, throughput 3.90453K wps
[Epoch 56 Batch 30/173] avg loss 8.37726e-06, throughput 3.97339K wps
[Epoch 56 Batch 60/173] avg loss 1.3175e-05, throughput 3.89177K wps
[Epoch 56 Batch 90/173] avg loss 1.01159e-05, throughput 3.87351K wps
[Epoch 56 Batch 120/173] avg loss 9.66483e-06, throughput 3.87618K wps
[Epoch 56 Batch 150/173] avg loss 1.31376e-05, throughput 3.87681K wps
Begin Testing...
[Epoch 56] train avg loss 1.07783e-05, test acc 0.7479, test avg loss 1.19328, throughput 3.89866K wps
[Epoch 57 Batch 30/173] avg loss 1.03892e-05, throughput 3.9705K wps
[Epoch 57 Batch 60/173] avg loss 9.77892e-06, throughput 3.89785K wps
[Epoch 57 Batch 90/173] avg loss 9.58562e-06, throughput 3.89725K wps
[Epoch 57 Batch 120/173] avg loss 7.4506e-06, throughput 3.89075K wps
[Epoch 57 Batch 150/173] avg loss 1.41077e-05, throughput 3.89996K wps
Begin Testing...
[Epoch 57] train avg loss 1.0654e-05, test acc 0.7448, test avg loss 1.22341, throughput 3.91007K wps
[Epoch 58 Batch 30/173] avg loss 9.01026e-06, throughput 3.98184K wps
[Epoch 58 Batch 60/173] avg loss 9.92769e-06, throughput 3.8732K wps
[Epoch 58 Batch 90/173] avg loss 2.76054e-05, throughput 3.86962K wps
[Epoch 58 Batch 120/173] avg loss 2.01407e-05, throughput 3.86912K wps
[Epoch 58 Batch 150/173] avg loss 1.31781e-05, throughput 3.87643K wps
Begin Testing...
[Epoch 58] train avg loss 1.62755e-05, test acc 0.7448, test avg loss 1.22904, throughput 3.89294K wps
[Epoch 59 Batch 30/173] avg loss 1.79766e-05, throughput 3.95983K wps
[Epoch 59 Batch 60/173] avg loss 1.09662e-05, throughput 3.87485K wps
[Epoch 59 Batch 90/173] avg loss 1.26937e-05, throughput 3.87414K wps
[Epoch 59 Batch 120/173] avg loss 1.32343e-05, throughput 3.87283K wps
[Epoch 59 Batch 150/173] avg loss 1.54085e-05, throughput 3.89652K wps
Begin Testing...
[Epoch 59] train avg loss 1.37923e-05, test acc 0.7448, test avg loss 1.23364, throughput 3.89437K wps
Test loss 0.457078, test acc 0.7955
Total time cost 554.72s
9596 1066
[Epoch 0 Batch 30/173] avg loss 0.015454, throughput 3.69123K wps
[Epoch 0 Batch 60/173] avg loss 0.0146144, throughput 3.89112K wps
[Epoch 0 Batch 90/173] avg loss 0.0146187, throughput 3.88055K wps
[Epoch 0 Batch 120/173] avg loss 0.0139715, throughput 3.86966K wps
[Epoch 0 Batch 150/173] avg loss 0.013854, throughput 3.87786K wps
Begin Testing...
[Epoch 0] train avg loss 0.0144218, test acc 0.6396, test avg loss 0.644561, throughput 3.84782K wps
Observed Improvement.
Begin Testing...
[Epoch 1 Batch 30/173] avg loss 0.0133874, throughput 3.95162K wps
[Epoch 1 Batch 60/173] avg loss 0.0131781, throughput 3.8757K wps
[Epoch 1 Batch 90/173] avg loss 0.0128281, throughput 3.86902K wps
[Epoch 1 Batch 120/173] avg loss 0.0129649, throughput 3.86642K wps
[Epoch 1 Batch 150/173] avg loss 0.0130949, throughput 3.88017K wps
Begin Testing...
[Epoch 1] train avg loss 0.0130102, test acc 0.6292, test avg loss 0.626091, throughput 3.88917K wps
[Epoch 2 Batch 30/173] avg loss 0.0121679, throughput 3.95811K wps
[Epoch 2 Batch 60/173] avg loss 0.0124439, throughput 3.87729K wps
[Epoch 2 Batch 90/173] avg loss 0.0121238, throughput 3.8957K wps
[Epoch 2 Batch 120/173] avg loss 0.0118613, throughput 3.90618K wps
[Epoch 2 Batch 150/173] avg loss 0.0119275, throughput 3.88812K wps
Begin Testing...
[Epoch 2] train avg loss 0.0120578, test acc 0.7208, test avg loss 0.591875, throughput 3.90685K wps
Observed Improvement.
Begin Testing...
[Epoch 3 Batch 30/173] avg loss 0.011351, throughput 3.98302K wps
[Epoch 3 Batch 60/173] avg loss 0.0110858, throughput 3.90085K wps
[Epoch 3 Batch 90/173] avg loss 0.0112973, throughput 3.89095K wps
[Epoch 3 Batch 120/173] avg loss 0.0110689, throughput 3.87051K wps
[Epoch 3 Batch 150/173] avg loss 0.0109252, throughput 3.86529K wps
Begin Testing...
[Epoch 3] train avg loss 0.0111164, test acc 0.7562, test avg loss 0.553456, throughput 3.89998K wps
Observed Improvement.