Permalink
Find file Copy path
Fetching contributors…
Cannot retrieve contributors at this time
4485 lines (4484 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='non-static', save_prefix='sa-model')
Use gpu0
2320
56
Done! Tokenizing Time=0.96s, #Sentences=10662
SentimentNet(
(embedding): Embedding(18768 -> 300, float32)
(encoder): ConvolutionalEncoder(
(_convs): HybridConcurrent(
(0): HybridSequential(
(0): Conv1D(300 -> 100, kernel_size=(3,), stride=(1,))
(1): Activation(relu)
(2): HybridLambda(<lambda>)
)
(1): HybridSequential(
(0): Conv1D(300 -> 100, kernel_size=(4,), stride=(1,))
(1): Activation(relu)
(2): HybridLambda(<lambda>)
)
(2): HybridSequential(
(0): Conv1D(300 -> 100, kernel_size=(5,), stride=(1,))
(1): Activation(relu)
(2): HybridLambda(<lambda>)
)
)
)
(output): HybridSequential(
(0): Dropout(p = 0.5, axes=())
(1): Dense(None -> 2, linear)
)
)
9596 1066
[Epoch 0 Batch 30/173] avg loss 0.015203, throughput 3.6961K wps
[Epoch 0 Batch 60/173] avg loss 0.0150431, throughput 6.04342K wps
[Epoch 0 Batch 90/173] avg loss 0.0147778, throughput 6.03209K wps
[Epoch 0 Batch 120/173] avg loss 0.0143682, throughput 6.03482K wps
[Epoch 0 Batch 150/173] avg loss 0.0143305, throughput 6.03592K wps
Begin Testing...
[Epoch 0] train avg loss 0.014667, test acc 0.5604, test avg loss 0.676988, throughput 5.11379K wps
Observed Improvement.
Begin Testing...
[Epoch 1 Batch 30/173] avg loss 0.0134373, throughput 6.19251K wps
[Epoch 1 Batch 60/173] avg loss 0.0135284, throughput 6.02832K wps
[Epoch 1 Batch 90/173] avg loss 0.0136279, throughput 6.04156K wps
[Epoch 1 Batch 120/173] avg loss 0.0131238, throughput 6.03694K wps
[Epoch 1 Batch 150/173] avg loss 0.0132467, throughput 6.02131K wps
Begin Testing...
[Epoch 1] train avg loss 0.0134206, test acc 0.6115, test avg loss 0.653372, throughput 6.06195K wps
Observed Improvement.
Begin Testing...
[Epoch 2 Batch 30/173] avg loss 0.0126414, throughput 6.18294K wps
[Epoch 2 Batch 60/173] avg loss 0.0127928, throughput 6.03558K wps
[Epoch 2 Batch 90/173] avg loss 0.0125257, throughput 6.02041K wps
[Epoch 2 Batch 120/173] avg loss 0.0124697, throughput 6.04094K wps
[Epoch 2 Batch 150/173] avg loss 0.0123594, throughput 6.02556K wps
Begin Testing...
[Epoch 2] train avg loss 0.0125162, test acc 0.6188, test avg loss 0.638294, throughput 6.05623K wps
Observed Improvement.
Begin Testing...
[Epoch 3 Batch 30/173] avg loss 0.0118995, throughput 6.18414K wps
[Epoch 3 Batch 60/173] avg loss 0.0117426, throughput 6.02789K wps
[Epoch 3 Batch 90/173] avg loss 0.0116148, throughput 6.0208K wps
[Epoch 3 Batch 120/173] avg loss 0.0116146, throughput 6.02745K wps
[Epoch 3 Batch 150/173] avg loss 0.0114536, throughput 6.02593K wps
Begin Testing...
[Epoch 3] train avg loss 0.0116524, test acc 0.7063, test avg loss 0.601209, throughput 6.05377K wps
Observed Improvement.
Begin Testing...
[Epoch 4 Batch 30/173] avg loss 0.0109332, throughput 6.19075K wps
[Epoch 4 Batch 60/173] avg loss 0.0107368, throughput 6.04348K wps
[Epoch 4 Batch 90/173] avg loss 0.0106838, throughput 6.03721K wps
[Epoch 4 Batch 120/173] avg loss 0.0106015, throughput 6.02519K wps
[Epoch 4 Batch 150/173] avg loss 0.0107183, throughput 6.02631K wps
Begin Testing...
[Epoch 4] train avg loss 0.0107121, test acc 0.7417, test avg loss 0.567649, throughput 6.05912K wps
Observed Improvement.
Begin Testing...
[Epoch 5 Batch 30/173] avg loss 0.00998099, throughput 6.16524K wps
[Epoch 5 Batch 60/173] avg loss 0.00964439, throughput 6.02865K wps
[Epoch 5 Batch 90/173] avg loss 0.00950361, throughput 6.03184K wps
[Epoch 5 Batch 120/173] avg loss 0.00933593, throughput 6.03465K wps
[Epoch 5 Batch 150/173] avg loss 0.00962931, throughput 6.02544K wps
Begin Testing...
[Epoch 5] train avg loss 0.00958245, test acc 0.7677, test avg loss 0.532339, throughput 6.05151K wps
Observed Improvement.
Begin Testing...
[Epoch 6 Batch 30/173] avg loss 0.00859672, throughput 6.18755K wps
[Epoch 6 Batch 60/173] avg loss 0.00839141, throughput 6.02872K wps
[Epoch 6 Batch 90/173] avg loss 0.00853713, throughput 6.03151K wps
[Epoch 6 Batch 120/173] avg loss 0.00828238, throughput 6.01715K wps
[Epoch 6 Batch 150/173] avg loss 0.00834593, throughput 6.01663K wps
Begin Testing...
[Epoch 6] train avg loss 0.00845436, test acc 0.7740, test avg loss 0.499455, throughput 6.05246K wps
Observed Improvement.
Begin Testing...
[Epoch 7 Batch 30/173] avg loss 0.00750085, throughput 6.17438K wps
[Epoch 7 Batch 60/173] avg loss 0.00745265, throughput 6.0198K wps
[Epoch 7 Batch 90/173] avg loss 0.00727444, throughput 6.02536K wps
[Epoch 7 Batch 120/173] avg loss 0.00719763, throughput 6.01855K wps
[Epoch 7 Batch 150/173] avg loss 0.00708007, throughput 6.02694K wps
Begin Testing...
[Epoch 7] train avg loss 0.00731177, test acc 0.7760, test avg loss 0.47998, throughput 6.04869K wps
Observed Improvement.
Begin Testing...
[Epoch 8 Batch 30/173] avg loss 0.00683546, throughput 6.16926K wps
[Epoch 8 Batch 60/173] avg loss 0.00613416, throughput 6.02153K wps
[Epoch 8 Batch 90/173] avg loss 0.00631704, throughput 6.02521K wps
[Epoch 8 Batch 120/173] avg loss 0.00644446, throughput 6.02069K wps
[Epoch 8 Batch 150/173] avg loss 0.00623659, throughput 6.01092K wps
Begin Testing...
[Epoch 8] train avg loss 0.00635804, test acc 0.7823, test avg loss 0.45885, throughput 6.04652K wps
Observed Improvement.
Begin Testing...
[Epoch 9 Batch 30/173] avg loss 0.00588909, throughput 6.15466K wps
[Epoch 9 Batch 60/173] avg loss 0.00548402, throughput 6.01549K wps
[Epoch 9 Batch 90/173] avg loss 0.00543749, throughput 6.01204K wps
[Epoch 9 Batch 120/173] avg loss 0.005451, throughput 6.00838K wps
[Epoch 9 Batch 150/173] avg loss 0.00533225, throughput 6.0171K wps
Begin Testing...
[Epoch 9] train avg loss 0.00550581, test acc 0.7844, test avg loss 0.445888, throughput 6.03846K wps
Observed Improvement.
Begin Testing...
[Epoch 10 Batch 30/173] avg loss 0.0047041, throughput 6.16411K wps
[Epoch 10 Batch 60/173] avg loss 0.00488167, throughput 6.01511K wps
[Epoch 10 Batch 90/173] avg loss 0.0047591, throughput 6.02025K wps
[Epoch 10 Batch 120/173] avg loss 0.00480199, throughput 6.01438K wps
[Epoch 10 Batch 150/173] avg loss 0.00483061, throughput 6.02124K wps
Begin Testing...
[Epoch 10] train avg loss 0.00481616, test acc 0.7990, test avg loss 0.43852, throughput 6.04239K wps
Observed Improvement.
Begin Testing...
[Epoch 11 Batch 30/173] avg loss 0.00425012, throughput 6.16981K wps
[Epoch 11 Batch 60/173] avg loss 0.00403128, throughput 6.0113K wps
[Epoch 11 Batch 90/173] avg loss 0.00398855, throughput 6.03018K wps
[Epoch 11 Batch 120/173] avg loss 0.00371638, throughput 6.02651K wps
[Epoch 11 Batch 150/173] avg loss 0.00432382, throughput 6.02369K wps
Begin Testing...
[Epoch 11] train avg loss 0.00407686, test acc 0.7927, test avg loss 0.439653, throughput 6.04639K wps
[Epoch 12 Batch 30/173] avg loss 0.00357754, throughput 6.18688K wps
[Epoch 12 Batch 60/173] avg loss 0.00351782, throughput 6.02143K wps
[Epoch 12 Batch 90/173] avg loss 0.00354486, throughput 6.01239K wps
[Epoch 12 Batch 120/173] avg loss 0.00361438, throughput 6.00424K wps
[Epoch 12 Batch 150/173] avg loss 0.0035338, throughput 6.01075K wps
Begin Testing...
[Epoch 12] train avg loss 0.00356345, test acc 0.7948, test avg loss 0.44279, throughput 6.04247K wps
[Epoch 13 Batch 30/173] avg loss 0.0031997, throughput 6.16379K wps
[Epoch 13 Batch 60/173] avg loss 0.00286763, throughput 6.01068K wps
[Epoch 13 Batch 90/173] avg loss 0.00293468, throughput 6.00056K wps
[Epoch 13 Batch 120/173] avg loss 0.00294282, throughput 6.00702K wps
[Epoch 13 Batch 150/173] avg loss 0.00304026, throughput 6.00884K wps
Begin Testing...
[Epoch 13] train avg loss 0.0030097, test acc 0.7937, test avg loss 0.44738, throughput 6.03539K wps
[Epoch 14 Batch 30/173] avg loss 0.00246947, throughput 6.1594K wps
[Epoch 14 Batch 60/173] avg loss 0.00264074, throughput 6.01719K wps
[Epoch 14 Batch 90/173] avg loss 0.00260072, throughput 6.02562K wps
[Epoch 14 Batch 120/173] avg loss 0.00240941, throughput 6.02438K wps
[Epoch 14 Batch 150/173] avg loss 0.00280467, throughput 6.01965K wps
Begin Testing...
[Epoch 14] train avg loss 0.00256136, test acc 0.7927, test avg loss 0.452745, throughput 6.01055K wps
[Epoch 15 Batch 30/173] avg loss 0.00230377, throughput 6.16437K wps
[Epoch 15 Batch 60/173] avg loss 0.00230091, throughput 6.00968K wps
[Epoch 15 Batch 90/173] avg loss 0.00224599, throughput 6.01645K wps
[Epoch 15 Batch 120/173] avg loss 0.00217349, throughput 6.02193K wps
[Epoch 15 Batch 150/173] avg loss 0.00211565, throughput 6.02451K wps
Begin Testing...
[Epoch 15] train avg loss 0.00224053, test acc 0.7896, test avg loss 0.464215, throughput 6.04354K wps
[Epoch 16 Batch 30/173] avg loss 0.00185213, throughput 6.16301K wps
[Epoch 16 Batch 60/173] avg loss 0.00193051, throughput 6.01735K wps
[Epoch 16 Batch 90/173] avg loss 0.00193548, throughput 6.0182K wps
[Epoch 16 Batch 120/173] avg loss 0.00189937, throughput 6.01167K wps
[Epoch 16 Batch 150/173] avg loss 0.00180234, throughput 6.00761K wps
Begin Testing...
[Epoch 16] train avg loss 0.0018856, test acc 0.7844, test avg loss 0.478588, throughput 6.04138K wps
[Epoch 17 Batch 30/173] avg loss 0.00147367, throughput 6.16854K wps
[Epoch 17 Batch 60/173] avg loss 0.0015929, throughput 6.02227K wps
[Epoch 17 Batch 90/173] avg loss 0.00146013, throughput 6.02099K wps
[Epoch 17 Batch 120/173] avg loss 0.00153841, throughput 6.01612K wps
[Epoch 17 Batch 150/173] avg loss 0.00147114, throughput 6.00363K wps
Begin Testing...
[Epoch 17] train avg loss 0.00152661, test acc 0.7927, test avg loss 0.486245, throughput 6.04387K wps
[Epoch 18 Batch 30/173] avg loss 0.0013468, throughput 6.16288K wps
[Epoch 18 Batch 60/173] avg loss 0.00133239, throughput 6.01706K wps
[Epoch 18 Batch 90/173] avg loss 0.00144757, throughput 6.00662K wps
[Epoch 18 Batch 120/173] avg loss 0.00134155, throughput 6.00795K wps
[Epoch 18 Batch 150/173] avg loss 0.00122006, throughput 6.00994K wps
Begin Testing...
[Epoch 18] train avg loss 0.00132465, test acc 0.7917, test avg loss 0.500412, throughput 6.03596K wps
[Epoch 19 Batch 30/173] avg loss 0.0011125, throughput 6.15414K wps
[Epoch 19 Batch 60/173] avg loss 0.00104109, throughput 6.002K wps
[Epoch 19 Batch 90/173] avg loss 0.00109985, throughput 6.00195K wps
[Epoch 19 Batch 120/173] avg loss 0.00119506, throughput 6.00527K wps
[Epoch 19 Batch 150/173] avg loss 0.0011029, throughput 6.00884K wps
Begin Testing...
[Epoch 19] train avg loss 0.00113524, test acc 0.7906, test avg loss 0.513206, throughput 6.03202K wps
[Epoch 20 Batch 30/173] avg loss 0.000907174, throughput 6.15386K wps
[Epoch 20 Batch 60/173] avg loss 0.00105895, throughput 6.01653K wps
[Epoch 20 Batch 90/173] avg loss 0.000843851, throughput 6.02078K wps
[Epoch 20 Batch 120/173] avg loss 0.000914847, throughput 6.00712K wps
[Epoch 20 Batch 150/173] avg loss 0.000951288, throughput 6.01201K wps
Begin Testing...
[Epoch 20] train avg loss 0.000949168, test acc 0.7958, test avg loss 0.527149, throughput 6.03692K wps
[Epoch 21 Batch 30/173] avg loss 0.000854161, throughput 6.14191K wps
[Epoch 21 Batch 60/173] avg loss 0.000814922, throughput 6.00929K wps
[Epoch 21 Batch 90/173] avg loss 0.00090738, throughput 6.00723K wps
[Epoch 21 Batch 120/173] avg loss 0.000853009, throughput 6.00753K wps
[Epoch 21 Batch 150/173] avg loss 0.000750118, throughput 6.01767K wps
Begin Testing...
[Epoch 21] train avg loss 0.000829614, test acc 0.7948, test avg loss 0.546462, throughput 6.03447K wps
[Epoch 22 Batch 30/173] avg loss 0.000665847, throughput 6.15857K wps
[Epoch 22 Batch 60/173] avg loss 0.000693917, throughput 5.98304K wps
[Epoch 22 Batch 90/173] avg loss 0.000720603, throughput 5.99526K wps
[Epoch 22 Batch 120/173] avg loss 0.000773567, throughput 5.98302K wps
[Epoch 22 Batch 150/173] avg loss 0.000734719, throughput 6.00574K wps
Begin Testing...
[Epoch 22] train avg loss 0.000725858, test acc 0.7865, test avg loss 0.56811, throughput 6.02252K wps
[Epoch 23 Batch 30/173] avg loss 0.00062659, throughput 6.16814K wps
[Epoch 23 Batch 60/173] avg loss 0.000596615, throughput 6.01887K wps
[Epoch 23 Batch 90/173] avg loss 0.000652071, throughput 6.00818K wps
[Epoch 23 Batch 120/173] avg loss 0.000526728, throughput 6.0151K wps
[Epoch 23 Batch 150/173] avg loss 0.000639737, throughput 6.01625K wps
Begin Testing...
[Epoch 23] train avg loss 0.000604409, test acc 0.7906, test avg loss 0.581184, throughput 6.04305K wps
[Epoch 24 Batch 30/173] avg loss 0.000469036, throughput 6.15791K wps
[Epoch 24 Batch 60/173] avg loss 0.000541333, throughput 5.9997K wps
[Epoch 24 Batch 90/173] avg loss 0.000455324, throughput 5.98548K wps
[Epoch 24 Batch 120/173] avg loss 0.000545488, throughput 5.99446K wps
[Epoch 24 Batch 150/173] avg loss 0.000484691, throughput 5.9969K wps
Begin Testing...
[Epoch 24] train avg loss 0.000512611, test acc 0.7927, test avg loss 0.59735, throughput 6.02351K wps
[Epoch 25 Batch 30/173] avg loss 0.000443516, throughput 6.13244K wps
[Epoch 25 Batch 60/173] avg loss 0.000399022, throughput 6.00745K wps
[Epoch 25 Batch 90/173] avg loss 0.000481133, throughput 6.00554K wps
[Epoch 25 Batch 120/173] avg loss 0.000436668, throughput 6.00723K wps
[Epoch 25 Batch 150/173] avg loss 0.000415438, throughput 5.994K wps
Begin Testing...
[Epoch 25] train avg loss 0.000444812, test acc 0.7875, test avg loss 0.612028, throughput 6.02612K wps
[Epoch 26 Batch 30/173] avg loss 0.000379404, throughput 6.15667K wps
[Epoch 26 Batch 60/173] avg loss 0.000363374, throughput 6.00046K wps
[Epoch 26 Batch 90/173] avg loss 0.000352817, throughput 6.0024K wps
[Epoch 26 Batch 120/173] avg loss 0.000408347, throughput 6.00209K wps
[Epoch 26 Batch 150/173] avg loss 0.000402287, throughput 6.00807K wps
Begin Testing...
[Epoch 26] train avg loss 0.000386357, test acc 0.7927, test avg loss 0.629488, throughput 6.03055K wps
[Epoch 27 Batch 30/173] avg loss 0.000327579, throughput 6.15354K wps
[Epoch 27 Batch 60/173] avg loss 0.000337836, throughput 6.00515K wps
[Epoch 27 Batch 90/173] avg loss 0.00036117, throughput 6.00186K wps
[Epoch 27 Batch 120/173] avg loss 0.000368376, throughput 6.00555K wps
[Epoch 27 Batch 150/173] avg loss 0.000333458, throughput 5.99225K wps
Begin Testing...
[Epoch 27] train avg loss 0.000346507, test acc 0.7792, test avg loss 0.652539, throughput 6.0288K wps
[Epoch 28 Batch 30/173] avg loss 0.000284824, throughput 6.15006K wps
[Epoch 28 Batch 60/173] avg loss 0.000294272, throughput 6.00449K wps
[Epoch 28 Batch 90/173] avg loss 0.000276293, throughput 5.98755K wps
[Epoch 28 Batch 120/173] avg loss 0.000348308, throughput 5.98889K wps
[Epoch 28 Batch 150/173] avg loss 0.000291219, throughput 6.00171K wps
Begin Testing...
[Epoch 28] train avg loss 0.000295613, test acc 0.7844, test avg loss 0.667549, throughput 6.02311K wps
[Epoch 29 Batch 30/173] avg loss 0.000216023, throughput 6.13544K wps
[Epoch 29 Batch 60/173] avg loss 0.000238618, throughput 5.98912K wps
[Epoch 29 Batch 90/173] avg loss 0.00027413, throughput 5.98741K wps
[Epoch 29 Batch 120/173] avg loss 0.000228347, throughput 6.00477K wps
[Epoch 29 Batch 150/173] avg loss 0.000264052, throughput 6.00134K wps
Begin Testing...
[Epoch 29] train avg loss 0.000245959, test acc 0.7896, test avg loss 0.684756, throughput 6.02072K wps
[Epoch 30 Batch 30/173] avg loss 0.000184653, throughput 6.14985K wps
[Epoch 30 Batch 60/173] avg loss 0.000206887, throughput 5.99277K wps
[Epoch 30 Batch 90/173] avg loss 0.000209808, throughput 6.00189K wps
[Epoch 30 Batch 120/173] avg loss 0.000225246, throughput 6.0077K wps
[Epoch 30 Batch 150/173] avg loss 0.000215118, throughput 6.00893K wps
Begin Testing...
[Epoch 30] train avg loss 0.000208905, test acc 0.7885, test avg loss 0.703922, throughput 6.0272K wps
[Epoch 31 Batch 30/173] avg loss 0.000182325, throughput 6.15747K wps
[Epoch 31 Batch 60/173] avg loss 0.000185297, throughput 5.98793K wps
[Epoch 31 Batch 90/173] avg loss 0.000198915, throughput 5.99487K wps
[Epoch 31 Batch 120/173] avg loss 0.000173966, throughput 6.00124K wps
[Epoch 31 Batch 150/173] avg loss 0.000213743, throughput 5.98977K wps
Begin Testing...
[Epoch 31] train avg loss 0.000191529, test acc 0.7937, test avg loss 0.71936, throughput 6.02255K wps
[Epoch 32 Batch 30/173] avg loss 0.00016951, throughput 6.141K wps
[Epoch 32 Batch 60/173] avg loss 0.000178309, throughput 5.99983K wps
[Epoch 32 Batch 90/173] avg loss 0.000200434, throughput 6.00093K wps
[Epoch 32 Batch 120/173] avg loss 0.000150266, throughput 5.99659K wps
[Epoch 32 Batch 150/173] avg loss 0.000171416, throughput 5.98252K wps
Begin Testing...
[Epoch 32] train avg loss 0.000170699, test acc 0.7896, test avg loss 0.737307, throughput 6.0189K wps
[Epoch 33 Batch 30/173] avg loss 0.000138669, throughput 6.14216K wps
[Epoch 33 Batch 60/173] avg loss 0.000143738, throughput 5.99775K wps
[Epoch 33 Batch 90/173] avg loss 0.000159212, throughput 5.99468K wps
[Epoch 33 Batch 120/173] avg loss 0.000167744, throughput 5.99411K wps
[Epoch 33 Batch 150/173] avg loss 0.000159191, throughput 5.98771K wps
Begin Testing...
[Epoch 33] train avg loss 0.000149368, test acc 0.7854, test avg loss 0.75248, throughput 6.01955K wps
[Epoch 34 Batch 30/173] avg loss 0.00015622, throughput 6.14213K wps
[Epoch 34 Batch 60/173] avg loss 0.000125286, throughput 6.00249K wps
[Epoch 34 Batch 90/173] avg loss 0.000125284, throughput 5.99315K wps
[Epoch 34 Batch 120/173] avg loss 0.00013568, throughput 5.98909K wps
[Epoch 34 Batch 150/173] avg loss 0.000137076, throughput 5.99916K wps
Begin Testing...
[Epoch 34] train avg loss 0.000136292, test acc 0.7865, test avg loss 0.76904, throughput 6.02191K wps
[Epoch 35 Batch 30/173] avg loss 0.000107147, throughput 6.14644K wps
[Epoch 35 Batch 60/173] avg loss 9.45049e-05, throughput 6.00457K wps
[Epoch 35 Batch 90/173] avg loss 0.000128017, throughput 6.00279K wps
[Epoch 35 Batch 120/173] avg loss 0.000125396, throughput 5.99562K wps
[Epoch 35 Batch 150/173] avg loss 0.000100448, throughput 5.99609K wps
Begin Testing...
[Epoch 35] train avg loss 0.000110333, test acc 0.7823, test avg loss 0.79061, throughput 6.02378K wps
[Epoch 36 Batch 30/173] avg loss 9.93802e-05, throughput 6.14405K wps
[Epoch 36 Batch 60/173] avg loss 9.53171e-05, throughput 6.00156K wps
[Epoch 36 Batch 90/173] avg loss 0.000103548, throughput 6.00053K wps
[Epoch 36 Batch 120/173] avg loss 0.000102318, throughput 5.99561K wps
[Epoch 36 Batch 150/173] avg loss 9.1194e-05, throughput 5.99707K wps
Begin Testing...
[Epoch 36] train avg loss 0.000101297, test acc 0.7854, test avg loss 0.804113, throughput 6.02375K wps
[Epoch 37 Batch 30/173] avg loss 7.27035e-05, throughput 6.1531K wps
[Epoch 37 Batch 60/173] avg loss 7.78889e-05, throughput 6.00731K wps
[Epoch 37 Batch 90/173] avg loss 0.000100501, throughput 6.00151K wps
[Epoch 37 Batch 120/173] avg loss 0.000136562, throughput 5.98915K wps
[Epoch 37 Batch 150/173] avg loss 9.5378e-05, throughput 5.99378K wps
Begin Testing...
[Epoch 37] train avg loss 9.60881e-05, test acc 0.7865, test avg loss 0.822323, throughput 6.0248K wps
[Epoch 38 Batch 30/173] avg loss 6.58222e-05, throughput 6.14363K wps
[Epoch 38 Batch 60/173] avg loss 7.28735e-05, throughput 5.98738K wps
[Epoch 38 Batch 90/173] avg loss 8.01014e-05, throughput 6.00046K wps
[Epoch 38 Batch 120/173] avg loss 7.4795e-05, throughput 5.99997K wps
[Epoch 38 Batch 150/173] avg loss 7.57246e-05, throughput 6.00836K wps
Begin Testing...
[Epoch 38] train avg loss 7.37117e-05, test acc 0.7854, test avg loss 0.83455, throughput 6.02331K wps
[Epoch 39 Batch 30/173] avg loss 6.90806e-05, throughput 6.15204K wps
[Epoch 39 Batch 60/173] avg loss 5.89135e-05, throughput 5.99602K wps
[Epoch 39 Batch 90/173] avg loss 6.47528e-05, throughput 6.0138K wps
[Epoch 39 Batch 120/173] avg loss 7.78169e-05, throughput 6.00862K wps
[Epoch 39 Batch 150/173] avg loss 6.58583e-05, throughput 6.00118K wps
Begin Testing...
[Epoch 39] train avg loss 6.84189e-05, test acc 0.7823, test avg loss 0.86132, throughput 6.03048K wps
[Epoch 40 Batch 30/173] avg loss 6.35845e-05, throughput 6.15224K wps
[Epoch 40 Batch 60/173] avg loss 5.28085e-05, throughput 6.00324K wps
[Epoch 40 Batch 90/173] avg loss 5.71693e-05, throughput 6.00249K wps
[Epoch 40 Batch 120/173] avg loss 7.28125e-05, throughput 5.99367K wps
[Epoch 40 Batch 150/173] avg loss 7.14821e-05, throughput 6.01086K wps
Begin Testing...
[Epoch 40] train avg loss 6.312e-05, test acc 0.7812, test avg loss 0.876374, throughput 6.02912K wps
[Epoch 41 Batch 30/173] avg loss 4.8586e-05, throughput 6.15195K wps
[Epoch 41 Batch 60/173] avg loss 5.81783e-05, throughput 6.00778K wps
[Epoch 41 Batch 90/173] avg loss 5.41493e-05, throughput 6.00045K wps
[Epoch 41 Batch 120/173] avg loss 8.71567e-05, throughput 6.00702K wps
[Epoch 41 Batch 150/173] avg loss 4.16295e-05, throughput 6.01384K wps
Begin Testing...
[Epoch 41] train avg loss 5.84498e-05, test acc 0.7854, test avg loss 0.897184, throughput 6.03088K wps
[Epoch 42 Batch 30/173] avg loss 5.00396e-05, throughput 6.14377K wps
[Epoch 42 Batch 60/173] avg loss 5.13779e-05, throughput 6.01144K wps
[Epoch 42 Batch 90/173] avg loss 4.33591e-05, throughput 5.98449K wps
[Epoch 42 Batch 120/173] avg loss 5.00803e-05, throughput 6.00482K wps
[Epoch 42 Batch 150/173] avg loss 3.99824e-05, throughput 5.99117K wps
Begin Testing...
[Epoch 42] train avg loss 5.10717e-05, test acc 0.7844, test avg loss 0.906774, throughput 6.02458K wps
[Epoch 43 Batch 30/173] avg loss 4.41098e-05, throughput 6.15209K wps
[Epoch 43 Batch 60/173] avg loss 3.92898e-05, throughput 6.00696K wps
[Epoch 43 Batch 90/173] avg loss 4.03988e-05, throughput 6.00303K wps
[Epoch 43 Batch 120/173] avg loss 4.74275e-05, throughput 6.00563K wps
[Epoch 43 Batch 150/173] avg loss 5.69073e-05, throughput 6.00299K wps
Begin Testing...
[Epoch 43] train avg loss 4.46938e-05, test acc 0.7802, test avg loss 0.928682, throughput 6.02934K wps
[Epoch 44 Batch 30/173] avg loss 3.13464e-05, throughput 6.13913K wps
[Epoch 44 Batch 60/173] avg loss 4.24456e-05, throughput 6.00136K wps
[Epoch 44 Batch 90/173] avg loss 4.68955e-05, throughput 6.00087K wps
[Epoch 44 Batch 120/173] avg loss 3.77969e-05, throughput 6.00004K wps
[Epoch 44 Batch 150/173] avg loss 3.41721e-05, throughput 6.01068K wps
Begin Testing...
[Epoch 44] train avg loss 3.76728e-05, test acc 0.7812, test avg loss 0.936634, throughput 6.02461K wps
[Epoch 45 Batch 30/173] avg loss 3.45482e-05, throughput 6.14359K wps
[Epoch 45 Batch 60/173] avg loss 3.28928e-05, throughput 6.00597K wps
[Epoch 45 Batch 90/173] avg loss 3.23005e-05, throughput 5.99688K wps
[Epoch 45 Batch 120/173] avg loss 3.81475e-05, throughput 5.99618K wps
[Epoch 45 Batch 150/173] avg loss 3.3793e-05, throughput 5.99928K wps
Begin Testing...
[Epoch 45] train avg loss 3.6232e-05, test acc 0.7875, test avg loss 0.956264, throughput 6.0262K wps
[Epoch 46 Batch 30/173] avg loss 2.52168e-05, throughput 6.15859K wps
[Epoch 46 Batch 60/173] avg loss 2.65744e-05, throughput 5.99403K wps
[Epoch 46 Batch 90/173] avg loss 3.24112e-05, throughput 5.99599K wps
[Epoch 46 Batch 120/173] avg loss 2.95187e-05, throughput 5.99717K wps
[Epoch 46 Batch 150/173] avg loss 2.89842e-05, throughput 6.01126K wps
Begin Testing...
[Epoch 46] train avg loss 2.78463e-05, test acc 0.7865, test avg loss 0.971349, throughput 6.02927K wps
[Epoch 47 Batch 30/173] avg loss 3.11765e-05, throughput 6.15098K wps
[Epoch 47 Batch 60/173] avg loss 5.40933e-05, throughput 5.98786K wps
[Epoch 47 Batch 90/173] avg loss 3.39135e-05, throughput 5.99031K wps
[Epoch 47 Batch 120/173] avg loss 2.86169e-05, throughput 5.99606K wps
[Epoch 47 Batch 150/173] avg loss 2.73124e-05, throughput 5.99774K wps
Begin Testing...
[Epoch 47] train avg loss 3.37418e-05, test acc 0.7833, test avg loss 0.981043, throughput 6.0222K wps
[Epoch 48 Batch 30/173] avg loss 3.03376e-05, throughput 6.13008K wps
[Epoch 48 Batch 60/173] avg loss 2.30316e-05, throughput 5.99576K wps
[Epoch 48 Batch 90/173] avg loss 3.15339e-05, throughput 5.99596K wps
[Epoch 48 Batch 120/173] avg loss 2.84027e-05, throughput 5.99396K wps
[Epoch 48 Batch 150/173] avg loss 3.91861e-05, throughput 5.99655K wps
Begin Testing...
[Epoch 48] train avg loss 3.05895e-05, test acc 0.7875, test avg loss 1.00324, throughput 6.02117K wps
[Epoch 49 Batch 30/173] avg loss 2.10485e-05, throughput 6.13456K wps
[Epoch 49 Batch 60/173] avg loss 2.72576e-05, throughput 5.99673K wps
[Epoch 49 Batch 90/173] avg loss 2.67922e-05, throughput 5.99096K wps
[Epoch 49 Batch 120/173] avg loss 3.27913e-05, throughput 5.99204K wps
[Epoch 49 Batch 150/173] avg loss 2.37288e-05, throughput 6.01134K wps
Begin Testing...
[Epoch 49] train avg loss 2.53337e-05, test acc 0.7865, test avg loss 1.02099, throughput 6.02216K wps
[Epoch 50 Batch 30/173] avg loss 2.77634e-05, throughput 6.14019K wps
[Epoch 50 Batch 60/173] avg loss 2.21572e-05, throughput 6.00329K wps
[Epoch 50 Batch 90/173] avg loss 2.5185e-05, throughput 5.98873K wps
[Epoch 50 Batch 120/173] avg loss 1.98101e-05, throughput 6.00398K wps
[Epoch 50 Batch 150/173] avg loss 2.42977e-05, throughput 5.99816K wps
Begin Testing...
[Epoch 50] train avg loss 2.69352e-05, test acc 0.7896, test avg loss 1.03766, throughput 6.02285K wps
[Epoch 51 Batch 30/173] avg loss 2.4123e-05, throughput 6.14737K wps
[Epoch 51 Batch 60/173] avg loss 1.93076e-05, throughput 6.00077K wps
[Epoch 51 Batch 90/173] avg loss 4.79508e-05, throughput 5.9876K wps
[Epoch 51 Batch 120/173] avg loss 2.08452e-05, throughput 5.99466K wps
[Epoch 51 Batch 150/173] avg loss 2.28078e-05, throughput 6.0024K wps
Begin Testing...
[Epoch 51] train avg loss 2.6337e-05, test acc 0.7906, test avg loss 1.04908, throughput 6.0232K wps
[Epoch 52 Batch 30/173] avg loss 1.89423e-05, throughput 6.15734K wps
[Epoch 52 Batch 60/173] avg loss 1.8871e-05, throughput 6.00686K wps
[Epoch 52 Batch 90/173] avg loss 1.87075e-05, throughput 5.99716K wps
[Epoch 52 Batch 120/173] avg loss 2.33211e-05, throughput 5.98583K wps
[Epoch 52 Batch 150/173] avg loss 1.5931e-05, throughput 5.97918K wps
Begin Testing...
[Epoch 52] train avg loss 2.0492e-05, test acc 0.7896, test avg loss 1.06988, throughput 6.01941K wps
[Epoch 53 Batch 30/173] avg loss 1.27196e-05, throughput 6.14672K wps
[Epoch 53 Batch 60/173] avg loss 1.80597e-05, throughput 6.00092K wps
[Epoch 53 Batch 90/173] avg loss 2.27818e-05, throughput 5.99133K wps
[Epoch 53 Batch 120/173] avg loss 1.81507e-05, throughput 5.99092K wps
[Epoch 53 Batch 150/173] avg loss 1.42931e-05, throughput 5.99537K wps
Begin Testing...
[Epoch 53] train avg loss 1.71292e-05, test acc 0.7844, test avg loss 1.07915, throughput 6.02424K wps
[Epoch 54 Batch 30/173] avg loss 1.44193e-05, throughput 6.14682K wps
[Epoch 54 Batch 60/173] avg loss 1.50234e-05, throughput 5.99325K wps
[Epoch 54 Batch 90/173] avg loss 1.33098e-05, throughput 5.99326K wps
[Epoch 54 Batch 120/173] avg loss 1.88457e-05, throughput 5.99612K wps
[Epoch 54 Batch 150/173] avg loss 2.04346e-05, throughput 5.99606K wps
Begin Testing...
[Epoch 54] train avg loss 1.63176e-05, test acc 0.7792, test avg loss 1.09747, throughput 6.02069K wps
[Epoch 55 Batch 30/173] avg loss 1.40829e-05, throughput 6.1405K wps
[Epoch 55 Batch 60/173] avg loss 2.31303e-05, throughput 5.98423K wps
[Epoch 55 Batch 90/173] avg loss 1.75542e-05, throughput 5.98245K wps
[Epoch 55 Batch 120/173] avg loss 1.27669e-05, throughput 6.00517K wps
[Epoch 55 Batch 150/173] avg loss 1.26924e-05, throughput 6.01135K wps
Begin Testing...
[Epoch 55] train avg loss 1.77661e-05, test acc 0.7792, test avg loss 1.11054, throughput 6.02202K wps
[Epoch 56 Batch 30/173] avg loss 1.21588e-05, throughput 6.14857K wps
[Epoch 56 Batch 60/173] avg loss 1.20818e-05, throughput 6.00007K wps
[Epoch 56 Batch 90/173] avg loss 1.20845e-05, throughput 5.98562K wps
[Epoch 56 Batch 120/173] avg loss 1.16143e-05, throughput 5.99339K wps
[Epoch 56 Batch 150/173] avg loss 1.43959e-05, throughput 5.98424K wps
Begin Testing...
[Epoch 56] train avg loss 1.36457e-05, test acc 0.7771, test avg loss 1.13119, throughput 6.02126K wps
[Epoch 57 Batch 30/173] avg loss 1.07449e-05, throughput 6.14635K wps
[Epoch 57 Batch 60/173] avg loss 1.01532e-05, throughput 6.005K wps
[Epoch 57 Batch 90/173] avg loss 9.42193e-06, throughput 5.99652K wps
[Epoch 57 Batch 120/173] avg loss 1.75934e-05, throughput 5.99382K wps
[Epoch 57 Batch 150/173] avg loss 1.38091e-05, throughput 6.00669K wps
Begin Testing...
[Epoch 57] train avg loss 1.25082e-05, test acc 0.7812, test avg loss 1.14691, throughput 6.02646K wps
[Epoch 58 Batch 30/173] avg loss 9.32499e-06, throughput 6.15079K wps
[Epoch 58 Batch 60/173] avg loss 1.39081e-05, throughput 5.99929K wps
[Epoch 58 Batch 90/173] avg loss 1.06306e-05, throughput 5.99155K wps
[Epoch 58 Batch 120/173] avg loss 9.21211e-06, throughput 5.99993K wps
[Epoch 58 Batch 150/173] avg loss 1.7566e-05, throughput 5.98469K wps
Begin Testing...
[Epoch 58] train avg loss 1.18444e-05, test acc 0.7802, test avg loss 1.15462, throughput 6.02157K wps
[Epoch 59 Batch 30/173] avg loss 1.03816e-05, throughput 6.14544K wps
[Epoch 59 Batch 60/173] avg loss 1.34414e-05, throughput 5.99569K wps
[Epoch 59 Batch 90/173] avg loss 9.71807e-06, throughput 5.99473K wps
[Epoch 59 Batch 120/173] avg loss 9.21753e-06, throughput 5.97845K wps
[Epoch 59 Batch 150/173] avg loss 1.29267e-05, throughput 5.98075K wps
Begin Testing...
[Epoch 59] train avg loss 1.17315e-05, test acc 0.7802, test avg loss 1.19635, throughput 6.01446K wps
Test loss 0.436088, test acc 0.7955
Total time cost 361.17s
9596 1066
[Epoch 0 Batch 30/173] avg loss 0.0150947, throughput 5.78623K wps
[Epoch 0 Batch 60/173] avg loss 0.0150307, throughput 6.00657K wps
[Epoch 0 Batch 90/173] avg loss 0.0146668, throughput 6.00215K wps
[Epoch 0 Batch 120/173] avg loss 0.0145588, throughput 6.00448K wps
[Epoch 0 Batch 150/173] avg loss 0.0141856, throughput 5.99167K wps
Begin Testing...
[Epoch 0] train avg loss 0.014671, test acc 0.5979, test avg loss 0.661511, throughput 5.96163K wps
Observed Improvement.
Begin Testing...
[Epoch 1 Batch 30/173] avg loss 0.0134738, throughput 6.15499K wps
[Epoch 1 Batch 60/173] avg loss 0.0134324, throughput 6.00165K wps
[Epoch 1 Batch 90/173] avg loss 0.0135237, throughput 6.01064K wps
[Epoch 1 Batch 120/173] avg loss 0.0131719, throughput 6.01431K wps
[Epoch 1 Batch 150/173] avg loss 0.013286, throughput 6.01148K wps
Begin Testing...
[Epoch 1] train avg loss 0.0133189, test acc 0.6448, test avg loss 0.642732, throughput 6.03301K wps
Observed Improvement.
Begin Testing...
[Epoch 2 Batch 30/173] avg loss 0.0127182, throughput 6.1513K wps
[Epoch 2 Batch 60/173] avg loss 0.0127166, throughput 6.0006K wps
[Epoch 2 Batch 90/173] avg loss 0.0125131, throughput 5.99819K wps
[Epoch 2 Batch 120/173] avg loss 0.0124438, throughput 6.01086K wps
[Epoch 2 Batch 150/173] avg loss 0.0122891, throughput 6.00239K wps
Begin Testing...
[Epoch 2] train avg loss 0.0125367, test acc 0.6594, test avg loss 0.619016, throughput 6.02768K wps
Observed Improvement.
Begin Testing...
[Epoch 3 Batch 30/173] avg loss 0.0119095, throughput 6.15591K wps
[Epoch 3 Batch 60/173] avg loss 0.0117906, throughput 6.00228K wps
[Epoch 3 Batch 90/173] avg loss 0.0117125, throughput 5.99639K wps
[Epoch 3 Batch 120/173] avg loss 0.0117299, throughput 6.00047K wps
[Epoch 3 Batch 150/173] avg loss 0.011493, throughput 5.9923K wps
Begin Testing...
[Epoch 3] train avg loss 0.011676, test acc 0.6813, test avg loss 0.594967, throughput 6.02503K wps
Observed Improvement.
Begin Testing...
[Epoch 4 Batch 30/173] avg loss 0.0108399, throughput 6.13358K wps
[Epoch 4 Batch 60/173] avg loss 0.0110039, throughput 5.9882K wps
[Epoch 4 Batch 90/173] avg loss 0.0107745, throughput 5.99778K wps
[Epoch 4 Batch 120/173] avg loss 0.0106089, throughput 5.99619K wps
[Epoch 4 Batch 150/173] avg loss 0.0108605, throughput 5.99564K wps
Begin Testing...
[Epoch 4] train avg loss 0.010755, test acc 0.7292, test avg loss 0.561998, throughput 6.01841K wps
Observed Improvement.
Begin Testing...
[Epoch 5 Batch 30/173] avg loss 0.00990881, throughput 6.14656K wps
[Epoch 5 Batch 60/173] avg loss 0.00986245, throughput 6.00168K wps
[Epoch 5 Batch 90/173] avg loss 0.00991284, throughput 6.00748K wps
[Epoch 5 Batch 120/173] avg loss 0.00944491, throughput 6.00108K wps
[Epoch 5 Batch 150/173] avg loss 0.00944848, throughput 5.99855K wps
Begin Testing...
[Epoch 5] train avg loss 0.00966454, test acc 0.7542, test avg loss 0.532423, throughput 6.02971K wps
Observed Improvement.
Begin Testing...
[Epoch 6 Batch 30/173] avg loss 0.00855821, throughput 6.14084K wps
[Epoch 6 Batch 60/173] avg loss 0.00875192, throughput 5.98606K wps
[Epoch 6 Batch 90/173] avg loss 0.00845411, throughput 6.00376K wps
[Epoch 6 Batch 120/173] avg loss 0.00856686, throughput 5.99334K wps
[Epoch 6 Batch 150/173] avg loss 0.00834192, throughput 5.99028K wps
Begin Testing...
[Epoch 6] train avg loss 0.0085098, test acc 0.7562, test avg loss 0.503136, throughput 6.02101K wps
Observed Improvement.
Begin Testing...
[Epoch 7 Batch 30/173] avg loss 0.00789216, throughput 6.15222K wps
[Epoch 7 Batch 60/173] avg loss 0.00761798, throughput 5.999K wps
[Epoch 7 Batch 90/173] avg loss 0.00745391, throughput 6.00012K wps
[Epoch 7 Batch 120/173] avg loss 0.00734058, throughput 6.00623K wps
[Epoch 7 Batch 150/173] avg loss 0.00731068, throughput 5.99882K wps
Begin Testing...
[Epoch 7] train avg loss 0.00750327, test acc 0.7604, test avg loss 0.481514, throughput 6.02714K wps
Observed Improvement.
Begin Testing...
[Epoch 8 Batch 30/173] avg loss 0.00675666, throughput 6.14991K wps
[Epoch 8 Batch 60/173] avg loss 0.00660159, throughput 5.99229K wps
[Epoch 8 Batch 90/173] avg loss 0.00635057, throughput 6.00278K wps
[Epoch 8 Batch 120/173] avg loss 0.00657339, throughput 5.98989K wps
[Epoch 8 Batch 150/173] avg loss 0.00654252, throughput 5.98791K wps
Begin Testing...
[Epoch 8] train avg loss 0.00649042, test acc 0.7573, test avg loss 0.468394, throughput 6.01956K wps
[Epoch 9 Batch 30/173] avg loss 0.00571839, throughput 6.15711K wps
[Epoch 9 Batch 60/173] avg loss 0.00568071, throughput 5.98532K wps
[Epoch 9 Batch 90/173] avg loss 0.00553936, throughput 5.99905K wps
[Epoch 9 Batch 120/173] avg loss 0.0055998, throughput 6.0106K wps
[Epoch 9 Batch 150/173] avg loss 0.00579394, throughput 6.00512K wps
Begin Testing...
[Epoch 9] train avg loss 0.00569196, test acc 0.7688, test avg loss 0.45706, throughput 6.02646K wps
Observed Improvement.
Begin Testing...
[Epoch 10 Batch 30/173] avg loss 0.00505915, throughput 6.13491K wps
[Epoch 10 Batch 60/173] avg loss 0.00507704, throughput 6.00971K wps
[Epoch 10 Batch 90/173] avg loss 0.00511841, throughput 6.00071K wps
[Epoch 10 Batch 120/173] avg loss 0.00483904, throughput 5.98806K wps
[Epoch 10 Batch 150/173] avg loss 0.0047655, throughput 6.01486K wps
Begin Testing...
[Epoch 10] train avg loss 0.0049498, test acc 0.7698, test avg loss 0.45542, throughput 6.02692K wps
Observed Improvement.
Begin Testing...
[Epoch 11 Batch 30/173] avg loss 0.00429726, throughput 6.14844K wps
[Epoch 11 Batch 60/173] avg loss 0.00396709, throughput 5.99646K wps
[Epoch 11 Batch 90/173] avg loss 0.00409645, throughput 5.99985K wps
[Epoch 11 Batch 120/173] avg loss 0.00404561, throughput 5.98845K wps
[Epoch 11 Batch 150/173] avg loss 0.00444621, throughput 5.99964K wps
Begin Testing...
[Epoch 11] train avg loss 0.0041804, test acc 0.7688, test avg loss 0.46165, throughput 6.02279K wps
[Epoch 12 Batch 30/173] avg loss 0.00360587, throughput 6.14081K wps
[Epoch 12 Batch 60/173] avg loss 0.00369563, throughput 5.98905K wps
[Epoch 12 Batch 90/173] avg loss 0.00365292, throughput 6.00407K wps
[Epoch 12 Batch 120/173] avg loss 0.0035808, throughput 5.99591K wps
[Epoch 12 Batch 150/173] avg loss 0.00382702, throughput 5.99236K wps
Begin Testing...
[Epoch 12] train avg loss 0.00364734, test acc 0.7740, test avg loss 0.458778, throughput 6.0199K wps
Observed Improvement.
Begin Testing...
[Epoch 13 Batch 30/173] avg loss 0.00311186, throughput 6.14335K wps
[Epoch 13 Batch 60/173] avg loss 0.00328843, throughput 5.99527K wps
[Epoch 13 Batch 90/173] avg loss 0.00304928, throughput 5.99525K wps
[Epoch 13 Batch 120/173] avg loss 0.00317272, throughput 5.99668K wps
[Epoch 13 Batch 150/173] avg loss 0.00328561, throughput 5.99275K wps
Begin Testing...
[Epoch 13] train avg loss 0.00318892, test acc 0.7760, test avg loss 0.465634, throughput 6.02157K wps
Observed Improvement.
Begin Testing...
[Epoch 14 Batch 30/173] avg loss 0.00269999, throughput 6.15205K wps
[Epoch 14 Batch 60/173] avg loss 0.00264659, throughput 6.00195K wps
[Epoch 14 Batch 90/173] avg loss 0.00264967, throughput 5.99785K wps
[Epoch 14 Batch 120/173] avg loss 0.00266408, throughput 6.00394K wps
[Epoch 14 Batch 150/173] avg loss 0.00245068, throughput 6.00945K wps
Begin Testing...
[Epoch 14] train avg loss 0.00265635, test acc 0.7729, test avg loss 0.475777, throughput 6.03114K wps
[Epoch 15 Batch 30/173] avg loss 0.0022825, throughput 6.15255K wps
[Epoch 15 Batch 60/173] avg loss 0.00218117, throughput 5.98918K wps
[Epoch 15 Batch 90/173] avg loss 0.00221789, throughput 5.99456K wps
[Epoch 15 Batch 120/173] avg loss 0.00229574, throughput 5.99451K wps
[Epoch 15 Batch 150/173] avg loss 0.00221716, throughput 5.99355K wps
Begin Testing...
[Epoch 15] train avg loss 0.00222728, test acc 0.7719, test avg loss 0.494653, throughput 6.02072K wps
[Epoch 16 Batch 30/173] avg loss 0.00191959, throughput 6.13954K wps
[Epoch 16 Batch 60/173] avg loss 0.00165641, throughput 5.98888K wps
[Epoch 16 Batch 90/173] avg loss 0.00180211, throughput 5.99199K wps
[Epoch 16 Batch 120/173] avg loss 0.00203705, throughput 6.00397K wps
[Epoch 16 Batch 150/173] avg loss 0.00201081, throughput 6.005K wps
Begin Testing...
[Epoch 16] train avg loss 0.00191859, test acc 0.7708, test avg loss 0.505786, throughput 6.02323K wps
[Epoch 17 Batch 30/173] avg loss 0.00174903, throughput 6.14012K wps
[Epoch 17 Batch 60/173] avg loss 0.00154794, throughput 6.00233K wps
[Epoch 17 Batch 90/173] avg loss 0.00147283, throughput 6.01084K wps
[Epoch 17 Batch 120/173] avg loss 0.00151976, throughput 6.00915K wps
[Epoch 17 Batch 150/173] avg loss 0.00200551, throughput 6.00296K wps
Begin Testing...
[Epoch 17] train avg loss 0.00164575, test acc 0.7667, test avg loss 0.526151, throughput 6.0305K wps
[Epoch 18 Batch 30/173] avg loss 0.0014133, throughput 6.16233K wps
[Epoch 18 Batch 60/173] avg loss 0.00149372, throughput 6.00556K wps
[Epoch 18 Batch 90/173] avg loss 0.00132576, throughput 6.00617K wps
[Epoch 18 Batch 120/173] avg loss 0.00141245, throughput 5.99807K wps
[Epoch 18 Batch 150/173] avg loss 0.00136622, throughput 5.98823K wps
Begin Testing...
[Epoch 18] train avg loss 0.0013865, test acc 0.7688, test avg loss 0.54363, throughput 6.03015K wps
[Epoch 19 Batch 30/173] avg loss 0.00125343, throughput 6.15147K wps
[Epoch 19 Batch 60/173] avg loss 0.00112624, throughput 6.00705K wps
[Epoch 19 Batch 90/173] avg loss 0.00125009, throughput 5.98584K wps
[Epoch 19 Batch 120/173] avg loss 0.00114809, throughput 6.01427K wps
[Epoch 19 Batch 150/173] avg loss 0.00105899, throughput 5.99396K wps
Begin Testing...
[Epoch 19] train avg loss 0.00118437, test acc 0.7677, test avg loss 0.55921, throughput 6.02663K wps
[Epoch 20 Batch 30/173] avg loss 0.00103698, throughput 6.1485K wps
[Epoch 20 Batch 60/173] avg loss 0.00104537, throughput 6.00528K wps
[Epoch 20 Batch 90/173] avg loss 0.000955744, throughput 6.00038K wps
[Epoch 20 Batch 120/173] avg loss 0.000979419, throughput 5.99659K wps
[Epoch 20 Batch 150/173] avg loss 0.000928716, throughput 5.999K wps
Begin Testing...
[Epoch 20] train avg loss 0.000995319, test acc 0.7615, test avg loss 0.583679, throughput 6.02804K wps
[Epoch 21 Batch 30/173] avg loss 0.000700659, throughput 6.14625K wps
[Epoch 21 Batch 60/173] avg loss 0.000747386, throughput 6.00554K wps
[Epoch 21 Batch 90/173] avg loss 0.000837883, throughput 5.99702K wps
[Epoch 21 Batch 120/173] avg loss 0.000868892, throughput 6.00392K wps
[Epoch 21 Batch 150/173] avg loss 0.000958006, throughput 5.99451K wps
Begin Testing...
[Epoch 21] train avg loss 0.000826451, test acc 0.7635, test avg loss 0.601323, throughput 6.02609K wps
[Epoch 22 Batch 30/173] avg loss 0.000671332, throughput 6.16308K wps
[Epoch 22 Batch 60/173] avg loss 0.000669408, throughput 5.99147K wps
[Epoch 22 Batch 90/173] avg loss 0.000760882, throughput 5.99777K wps
[Epoch 22 Batch 120/173] avg loss 0.000812959, throughput 5.95926K wps
[Epoch 22 Batch 150/173] avg loss 0.000789081, throughput 5.97857K wps
Begin Testing...
[Epoch 22] train avg loss 0.000728912, test acc 0.7646, test avg loss 0.624075, throughput 6.01551K wps
[Epoch 23 Batch 30/173] avg loss 0.000554525, throughput 6.14433K wps
[Epoch 23 Batch 60/173] avg loss 0.000635745, throughput 6.00957K wps
[Epoch 23 Batch 90/173] avg loss 0.000722242, throughput 6.00666K wps
[Epoch 23 Batch 120/173] avg loss 0.000581208, throughput 6.00004K wps
[Epoch 23 Batch 150/173] avg loss 0.000680923, throughput 6.01021K wps
Begin Testing...
[Epoch 23] train avg loss 0.000629161, test acc 0.7656, test avg loss 0.646705, throughput 6.03106K wps
[Epoch 24 Batch 30/173] avg loss 0.000517376, throughput 6.14762K wps
[Epoch 24 Batch 60/173] avg loss 0.000516649, throughput 5.98963K wps
[Epoch 24 Batch 90/173] avg loss 0.000492165, throughput 5.99332K wps
[Epoch 24 Batch 120/173] avg loss 0.000501466, throughput 5.99217K wps
[Epoch 24 Batch 150/173] avg loss 0.000619398, throughput 5.99616K wps
Begin Testing...
[Epoch 24] train avg loss 0.000540928, test acc 0.7604, test avg loss 0.670814, throughput 6.02251K wps
[Epoch 25 Batch 30/173] avg loss 0.000472354, throughput 6.14939K wps
[Epoch 25 Batch 60/173] avg loss 0.00050629, throughput 6.00021K wps
[Epoch 25 Batch 90/173] avg loss 0.000447884, throughput 5.99374K wps
[Epoch 25 Batch 120/173] avg loss 0.000509149, throughput 5.99232K wps
[Epoch 25 Batch 150/173] avg loss 0.000467662, throughput 5.9966K wps
Begin Testing...
[Epoch 25] train avg loss 0.000474916, test acc 0.7625, test avg loss 0.699066, throughput 6.02249K wps
[Epoch 26 Batch 30/173] avg loss 0.000358083, throughput 6.16181K wps
[Epoch 26 Batch 60/173] avg loss 0.000378274, throughput 6.00364K wps
[Epoch 26 Batch 90/173] avg loss 0.000391378, throughput 6.00407K wps
[Epoch 26 Batch 120/173] avg loss 0.000424314, throughput 5.99398K wps
[Epoch 26 Batch 150/173] avg loss 0.0003967, throughput 5.98579K wps
Begin Testing...
[Epoch 26] train avg loss 0.000387367, test acc 0.7635, test avg loss 0.718156, throughput 6.02761K wps
[Epoch 27 Batch 30/173] avg loss 0.000365818, throughput 6.16482K wps
[Epoch 27 Batch 60/173] avg loss 0.00027604, throughput 6.00875K wps
[Epoch 27 Batch 90/173] avg loss 0.000380815, throughput 5.99622K wps
[Epoch 27 Batch 120/173] avg loss 0.000331481, throughput 6.00115K wps
[Epoch 27 Batch 150/173] avg loss 0.000368598, throughput 6.00679K wps
Begin Testing...
[Epoch 27] train avg loss 0.00034796, test acc 0.7594, test avg loss 0.739654, throughput 6.0317K wps
[Epoch 28 Batch 30/173] avg loss 0.000295419, throughput 6.14721K wps
[Epoch 28 Batch 60/173] avg loss 0.000284497, throughput 6.00684K wps
[Epoch 28 Batch 90/173] avg loss 0.000267457, throughput 6.01105K wps
[Epoch 28 Batch 120/173] avg loss 0.000287792, throughput 5.99991K wps
[Epoch 28 Batch 150/173] avg loss 0.000290274, throughput 6.01133K wps
Begin Testing...
[Epoch 28] train avg loss 0.00028778, test acc 0.7615, test avg loss 0.759488, throughput 6.03254K wps
[Epoch 29 Batch 30/173] avg loss 0.000255688, throughput 6.13606K wps
[Epoch 29 Batch 60/173] avg loss 0.000254062, throughput 5.99915K wps
[Epoch 29 Batch 90/173] avg loss 0.000225768, throughput 6.00795K wps
[Epoch 29 Batch 120/173] avg loss 0.000240798, throughput 6.01513K wps
[Epoch 29 Batch 150/173] avg loss 0.000261918, throughput 6.00426K wps
Begin Testing...
[Epoch 29] train avg loss 0.000249088, test acc 0.7542, test avg loss 0.785156, throughput 6.02737K wps
[Epoch 30 Batch 30/173] avg loss 0.000228, throughput 6.15192K wps
[Epoch 30 Batch 60/173] avg loss 0.000273611, throughput 6.01312K wps
[Epoch 30 Batch 90/173] avg loss 0.000276173, throughput 6.00366K wps
[Epoch 30 Batch 120/173] avg loss 0.000202823, throughput 6.00622K wps
[Epoch 30 Batch 150/173] avg loss 0.000199143, throughput 5.99764K wps
Begin Testing...
[Epoch 30] train avg loss 0.00023733, test acc 0.7625, test avg loss 0.804801, throughput 6.02997K wps
[Epoch 31 Batch 30/173] avg loss 0.000173973, throughput 6.1426K wps
[Epoch 31 Batch 60/173] avg loss 0.000176087, throughput 6.00839K wps
[Epoch 31 Batch 90/173] avg loss 0.00015904, throughput 5.99335K wps
[Epoch 31 Batch 120/173] avg loss 0.00024497, throughput 5.99817K wps
[Epoch 31 Batch 150/173] avg loss 0.000201032, throughput 6.00111K wps
Begin Testing...
[Epoch 31] train avg loss 0.000185367, test acc 0.7583, test avg loss 0.826364, throughput 6.02777K wps
[Epoch 32 Batch 30/173] avg loss 0.000157658, throughput 6.15624K wps
[Epoch 32 Batch 60/173] avg loss 0.000161615, throughput 5.98798K wps
[Epoch 32 Batch 90/173] avg loss 0.000210202, throughput 6.00362K wps
[Epoch 32 Batch 120/173] avg loss 0.000146233, throughput 5.99983K wps
[Epoch 32 Batch 150/173] avg loss 0.000187344, throughput 6.00989K wps
Begin Testing...
[Epoch 32] train avg loss 0.000171503, test acc 0.7594, test avg loss 0.845533, throughput 6.02841K wps
[Epoch 33 Batch 30/173] avg loss 0.000120619, throughput 6.14169K wps
[Epoch 33 Batch 60/173] avg loss 0.000136806, throughput 6.00225K wps
[Epoch 33 Batch 90/173] avg loss 0.000179983, throughput 6.0038K wps
[Epoch 33 Batch 120/173] avg loss 0.000143937, throughput 6.00347K wps
[Epoch 33 Batch 150/173] avg loss 0.000165442, throughput 6.00567K wps
Begin Testing...
[Epoch 33] train avg loss 0.000152533, test acc 0.7583, test avg loss 0.875618, throughput 6.02643K wps
[Epoch 34 Batch 30/173] avg loss 0.000109068, throughput 6.16106K wps
[Epoch 34 Batch 60/173] avg loss 0.0001492, throughput 6.01389K wps
[Epoch 34 Batch 90/173] avg loss 0.000122963, throughput 5.99542K wps
[Epoch 34 Batch 120/173] avg loss 0.000124019, throughput 5.98951K wps
[Epoch 34 Batch 150/173] avg loss 0.000143592, throughput 5.99303K wps
Begin Testing...
[Epoch 34] train avg loss 0.000130263, test acc 0.7573, test avg loss 0.900011, throughput 6.02762K wps
[Epoch 35 Batch 30/173] avg loss 8.79138e-05, throughput 6.14484K wps
[Epoch 35 Batch 60/173] avg loss 0.000102266, throughput 5.98314K wps
[Epoch 35 Batch 90/173] avg loss 0.000119591, throughput 6.00667K wps
[Epoch 35 Batch 120/173] avg loss 0.000119359, throughput 5.98926K wps
[Epoch 35 Batch 150/173] avg loss 0.000140821, throughput 5.99709K wps
Begin Testing...
[Epoch 35] train avg loss 0.000119454, test acc 0.7562, test avg loss 0.92061, throughput 6.01953K wps
[Epoch 36 Batch 30/173] avg loss 0.000101138, throughput 6.13795K wps
[Epoch 36 Batch 60/173] avg loss 9.84695e-05, throughput 5.98506K wps
[Epoch 36 Batch 90/173] avg loss 9.47715e-05, throughput 5.99956K wps
[Epoch 36 Batch 120/173] avg loss 0.000109265, throughput 5.99683K wps
[Epoch 36 Batch 150/173] avg loss 0.00011129, throughput 5.98996K wps
Begin Testing...
[Epoch 36] train avg loss 0.000103958, test acc 0.7531, test avg loss 0.937623, throughput 6.01791K wps
[Epoch 37 Batch 30/173] avg loss 8.94067e-05, throughput 6.14131K wps
[Epoch 37 Batch 60/173] avg loss 9.74077e-05, throughput 5.98875K wps
[Epoch 37 Batch 90/173] avg loss 9.70745e-05, throughput 5.98006K wps
[Epoch 37 Batch 120/173] avg loss 9.68376e-05, throughput 5.99941K wps
[Epoch 37 Batch 150/173] avg loss 9.80384e-05, throughput 5.99684K wps
Begin Testing...
[Epoch 37] train avg loss 9.66454e-05, test acc 0.7531, test avg loss 0.966286, throughput 6.01698K wps
[Epoch 38 Batch 30/173] avg loss 8.52454e-05, throughput 6.1503K wps
[Epoch 38 Batch 60/173] avg loss 7.95246e-05, throughput 6.01298K wps
[Epoch 38 Batch 90/173] avg loss 8.20853e-05, throughput 5.98934K wps
[Epoch 38 Batch 120/173] avg loss 8.08887e-05, throughput 5.99756K wps
[Epoch 38 Batch 150/173] avg loss 8.2568e-05, throughput 6.00639K wps
Begin Testing...
[Epoch 38] train avg loss 8.35502e-05, test acc 0.7510, test avg loss 0.980931, throughput 6.02833K wps
[Epoch 39 Batch 30/173] avg loss 7.60638e-05, throughput 6.1405K wps
[Epoch 39 Batch 60/173] avg loss 5.95018e-05, throughput 5.99311K wps
[Epoch 39 Batch 90/173] avg loss 7.57416e-05, throughput 5.9897K wps
[Epoch 39 Batch 120/173] avg loss 6.26133e-05, throughput 6.00685K wps
[Epoch 39 Batch 150/173] avg loss 7.73965e-05, throughput 6.00941K wps
Begin Testing...
[Epoch 39] train avg loss 6.97479e-05, test acc 0.7552, test avg loss 1.00924, throughput 6.02653K wps
[Epoch 40 Batch 30/173] avg loss 5.67167e-05, throughput 6.1463K wps
[Epoch 40 Batch 60/173] avg loss 6.19418e-05, throughput 6.00121K wps
[Epoch 40 Batch 90/173] avg loss 5.17116e-05, throughput 5.98903K wps
[Epoch 40 Batch 120/173] avg loss 8.68973e-05, throughput 5.99524K wps
[Epoch 40 Batch 150/173] avg loss 7.65449e-05, throughput 5.99845K wps
Begin Testing...
[Epoch 40] train avg loss 6.93295e-05, test acc 0.7479, test avg loss 1.0397, throughput 6.02175K wps
[Epoch 41 Batch 30/173] avg loss 6.79196e-05, throughput 6.14323K wps
[Epoch 41 Batch 60/173] avg loss 6.74736e-05, throughput 5.99818K wps
[Epoch 41 Batch 90/173] avg loss 6.13273e-05, throughput 6.00027K wps
[Epoch 41 Batch 120/173] avg loss 5.85574e-05, throughput 5.97971K wps
[Epoch 41 Batch 150/173] avg loss 6.46059e-05, throughput 6.00061K wps
Begin Testing...
[Epoch 41] train avg loss 6.33455e-05, test acc 0.7531, test avg loss 1.05563, throughput 6.02173K wps
[Epoch 42 Batch 30/173] avg loss 4.12595e-05, throughput 6.14696K wps
[Epoch 42 Batch 60/173] avg loss 7.68475e-05, throughput 6.00128K wps
[Epoch 42 Batch 90/173] avg loss 4.72968e-05, throughput 5.99667K wps
[Epoch 42 Batch 120/173] avg loss 4.13535e-05, throughput 5.9967K wps
[Epoch 42 Batch 150/173] avg loss 5.46264e-05, throughput 5.99816K wps
Begin Testing...
[Epoch 42] train avg loss 5.43032e-05, test acc 0.7521, test avg loss 1.07736, throughput 6.02425K wps
[Epoch 43 Batch 30/173] avg loss 4.17404e-05, throughput 6.13687K wps
[Epoch 43 Batch 60/173] avg loss 4.4358e-05, throughput 5.9959K wps
[Epoch 43 Batch 90/173] avg loss 4.3509e-05, throughput 5.99653K wps
[Epoch 43 Batch 120/173] avg loss 6.60048e-05, throughput 6.00677K wps
[Epoch 43 Batch 150/173] avg loss 5.49969e-05, throughput 6.00639K wps
Begin Testing...
[Epoch 43] train avg loss 4.9856e-05, test acc 0.7521, test avg loss 1.0933, throughput 6.02387K wps
[Epoch 44 Batch 30/173] avg loss 4.51831e-05, throughput 6.14081K wps
[Epoch 44 Batch 60/173] avg loss 4.22054e-05, throughput 5.99591K wps
[Epoch 44 Batch 90/173] avg loss 4.88123e-05, throughput 5.99892K wps
[Epoch 44 Batch 120/173] avg loss 4.04387e-05, throughput 5.98558K wps
[Epoch 44 Batch 150/173] avg loss 4.58332e-05, throughput 5.99013K wps
Begin Testing...
[Epoch 44] train avg loss 4.4455e-05, test acc 0.7531, test avg loss 1.1159, throughput 6.01929K wps
[Epoch 45 Batch 30/173] avg loss 3.71191e-05, throughput 6.13705K wps
[Epoch 45 Batch 60/173] avg loss 3.01632e-05, throughput 5.99913K wps
[Epoch 45 Batch 90/173] avg loss 4.00229e-05, throughput 5.99653K wps
[Epoch 45 Batch 120/173] avg loss 4.27504e-05, throughput 5.99627K wps
[Epoch 45 Batch 150/173] avg loss 3.40824e-05, throughput 5.9931K wps
Begin Testing...
[Epoch 45] train avg loss 3.88495e-05, test acc 0.7500, test avg loss 1.1445, throughput 6.0217K wps
[Epoch 46 Batch 30/173] avg loss 3.95131e-05, throughput 6.14124K wps
[Epoch 46 Batch 60/173] avg loss 2.798e-05, throughput 6.00038K wps
[Epoch 46 Batch 90/173] avg loss 3.68141e-05, throughput 6.00267K wps
[Epoch 46 Batch 120/173] avg loss 3.24614e-05, throughput 5.99944K wps
[Epoch 46 Batch 150/173] avg loss 4.19762e-05, throughput 5.98794K wps
Begin Testing...
[Epoch 46] train avg loss 3.55961e-05, test acc 0.7500, test avg loss 1.16524, throughput 6.02277K wps
[Epoch 47 Batch 30/173] avg loss 3.01588e-05, throughput 6.13378K wps
[Epoch 47 Batch 60/173] avg loss 3.3255e-05, throughput 5.99735K wps
[Epoch 47 Batch 90/173] avg loss 3.08154e-05, throughput 5.99775K wps
[Epoch 47 Batch 120/173] avg loss 3.20293e-05, throughput 5.99486K wps
[Epoch 47 Batch 150/173] avg loss 3.38651e-05, throughput 5.99976K wps
Begin Testing...
[Epoch 47] train avg loss 3.15925e-05, test acc 0.7479, test avg loss 1.18735, throughput 6.02167K wps
[Epoch 48 Batch 30/173] avg loss 2.69962e-05, throughput 6.15325K wps
[Epoch 48 Batch 60/173] avg loss 3.04995e-05, throughput 5.99305K wps
[Epoch 48 Batch 90/173] avg loss 2.72124e-05, throughput 5.99872K wps
[Epoch 48 Batch 120/173] avg loss 2.18469e-05, throughput 6.00417K wps
[Epoch 48 Batch 150/173] avg loss 3.01977e-05, throughput 5.99868K wps
Begin Testing...
[Epoch 48] train avg loss 2.98427e-05, test acc 0.7500, test avg loss 1.21463, throughput 6.02633K wps
[Epoch 49 Batch 30/173] avg loss 2.56991e-05, throughput 6.15484K wps
[Epoch 49 Batch 60/173] avg loss 2.96876e-05, throughput 6.00791K wps
[Epoch 49 Batch 90/173] avg loss 2.71136e-05, throughput 5.99343K wps
[Epoch 49 Batch 120/173] avg loss 2.50332e-05, throughput 6.00941K wps
[Epoch 49 Batch 150/173] avg loss 3.10593e-05, throughput 5.99817K wps
Begin Testing...
[Epoch 49] train avg loss 2.68738e-05, test acc 0.7490, test avg loss 1.23423, throughput 6.02683K wps
[Epoch 50 Batch 30/173] avg loss 2.80488e-05, throughput 6.15414K wps
[Epoch 50 Batch 60/173] avg loss 2.62786e-05, throughput 6.00582K wps
[Epoch 50 Batch 90/173] avg loss 2.00892e-05, throughput 5.99458K wps
[Epoch 50 Batch 120/173] avg loss 2.27986e-05, throughput 5.98645K wps
[Epoch 50 Batch 150/173] avg loss 2.05325e-05, throughput 6.00874K wps
Begin Testing...
[Epoch 50] train avg loss 2.42807e-05, test acc 0.7469, test avg loss 1.25581, throughput 6.02804K wps
[Epoch 51 Batch 30/173] avg loss 1.44786e-05, throughput 6.15616K wps
[Epoch 51 Batch 60/173] avg loss 1.76858e-05, throughput 5.99688K wps
[Epoch 51 Batch 90/173] avg loss 2.56405e-05, throughput 5.99042K wps
[Epoch 51 Batch 120/173] avg loss 2.21444e-05, throughput 5.99657K wps
[Epoch 51 Batch 150/173] avg loss 1.96551e-05, throughput 5.99003K wps
Begin Testing...
[Epoch 51] train avg loss 2.07341e-05, test acc 0.7448, test avg loss 1.27393, throughput 6.02292K wps
[Epoch 52 Batch 30/173] avg loss 2.02902e-05, throughput 6.14996K wps
[Epoch 52 Batch 60/173] avg loss 1.59254e-05, throughput 5.99375K wps
[Epoch 52 Batch 90/173] avg loss 2.7694e-05, throughput 6.00156K wps
[Epoch 52 Batch 120/173] avg loss 2.41768e-05, throughput 5.99049K wps
[Epoch 52 Batch 150/173] avg loss 1.8437e-05, throughput 5.99799K wps
Begin Testing...
[Epoch 52] train avg loss 2.10334e-05, test acc 0.7438, test avg loss 1.30916, throughput 6.02082K wps
[Epoch 53 Batch 30/173] avg loss 6.73209e-05, throughput 6.12078K wps
[Epoch 53 Batch 60/173] avg loss 1.8387e-05, throughput 5.99655K wps
[Epoch 53 Batch 90/173] avg loss 2.36295e-05, throughput 5.99022K wps
[Epoch 53 Batch 120/173] avg loss 1.94725e-05, throughput 6.0036K wps
[Epoch 53 Batch 150/173] avg loss 1.53489e-05, throughput 5.99563K wps
Begin Testing...
[Epoch 53] train avg loss 2.72365e-05, test acc 0.7427, test avg loss 1.32097, throughput 6.01749K wps
[Epoch 54 Batch 30/173] avg loss 1.45918e-05, throughput 6.15008K wps
[Epoch 54 Batch 60/173] avg loss 1.38023e-05, throughput 5.99204K wps
[Epoch 54 Batch 90/173] avg loss 1.32365e-05, throughput 6.00594K wps
[Epoch 54 Batch 120/173] avg loss 1.70705e-05, throughput 5.98982K wps
[Epoch 54 Batch 150/173] avg loss 1.817e-05, throughput 6.00659K wps
Begin Testing...
[Epoch 54] train avg loss 1.51206e-05, test acc 0.7417, test avg loss 1.33845, throughput 6.0255K wps
[Epoch 55 Batch 30/173] avg loss 1.3586e-05, throughput 6.15134K wps
[Epoch 55 Batch 60/173] avg loss 1.75084e-05, throughput 5.99235K wps
[Epoch 55 Batch 90/173] avg loss 1.63111e-05, throughput 6.00181K wps
[Epoch 55 Batch 120/173] avg loss 1.40863e-05, throughput 6.00661K wps
[Epoch 55 Batch 150/173] avg loss 1.35129e-05, throughput 5.9906K wps
Begin Testing...
[Epoch 55] train avg loss 1.50605e-05, test acc 0.7406, test avg loss 1.36114, throughput 6.02326K wps
[Epoch 56 Batch 30/173] avg loss 1.75556e-05, throughput 6.14408K wps
[Epoch 56 Batch 60/173] avg loss 1.3799e-05, throughput 6.00121K wps
[Epoch 56 Batch 90/173] avg loss 2.02958e-05, throughput 5.99754K wps
[Epoch 56 Batch 120/173] avg loss 1.62123e-05, throughput 6.00488K wps
[Epoch 56 Batch 150/173] avg loss 1.38063e-05, throughput 6.00858K wps
Begin Testing...
[Epoch 56] train avg loss 1.61422e-05, test acc 0.7406, test avg loss 1.364, throughput 6.02787K wps
[Epoch 57 Batch 30/173] avg loss 9.27337e-06, throughput 6.15397K wps
[Epoch 57 Batch 60/173] avg loss 1.30826e-05, throughput 5.99232K wps
[Epoch 57 Batch 90/173] avg loss 1.0863e-05, throughput 6.01381K wps
[Epoch 57 Batch 120/173] avg loss 1.09035e-05, throughput 5.99944K wps
[Epoch 57 Batch 150/173] avg loss 1.52407e-05, throughput 6.00203K wps
Begin Testing...
[Epoch 57] train avg loss 1.16506e-05, test acc 0.7417, test avg loss 1.38757, throughput 6.03021K wps
[Epoch 58 Batch 30/173] avg loss 9.66483e-06, throughput 6.13658K wps
[Epoch 58 Batch 60/173] avg loss 9.57192e-06, throughput 5.99917K wps
[Epoch 58 Batch 90/173] avg loss 1.60638e-05, throughput 6.01162K wps
[Epoch 58 Batch 120/173] avg loss 8.8933e-06, throughput 6.00227K wps
[Epoch 58 Batch 150/173] avg loss 1.57619e-05, throughput 6.00284K wps
Begin Testing...
[Epoch 58] train avg loss 1.17815e-05, test acc 0.7375, test avg loss 1.39881, throughput 6.02606K wps
[Epoch 59 Batch 30/173] avg loss 1.07633e-05, throughput 6.14163K wps
[Epoch 59 Batch 60/173] avg loss 1.32855e-05, throughput 5.99319K wps
[Epoch 59 Batch 90/173] avg loss 1.01769e-05, throughput 6.00149K wps
[Epoch 59 Batch 120/173] avg loss 9.57963e-06, throughput 5.99902K wps
[Epoch 59 Batch 150/173] avg loss 1.14288e-05, throughput 5.99732K wps
Begin Testing...
[Epoch 59] train avg loss 1.04713e-05, test acc 0.7375, test avg loss 1.4267, throughput 6.02185K wps
Test loss 0.425938, test acc 0.7964
Total time cost 358.63s
9596 1066
[Epoch 0 Batch 30/173] avg loss 0.0151679, throughput 5.78448K wps
[Epoch 0 Batch 60/173] avg loss 0.014922, throughput 5.98924K wps
[Epoch 0 Batch 90/173] avg loss 0.0148274, throughput 6.00065K wps
[Epoch 0 Batch 120/173] avg loss 0.0144994, throughput 6.00885K wps
[Epoch 0 Batch 150/173] avg loss 0.014345, throughput 6.00071K wps
Begin Testing...
[Epoch 0] train avg loss 0.0146757, test acc 0.5979, test avg loss 0.665816, throughput 5.96168K wps
Observed Improvement.
Begin Testing...
[Epoch 1 Batch 30/173] avg loss 0.0139262, throughput 6.15762K wps
[Epoch 1 Batch 60/173] avg loss 0.0133263, throughput 6.00588K wps
[Epoch 1 Batch 90/173] avg loss 0.0135379, throughput 6.00711K wps
[Epoch 1 Batch 120/173] avg loss 0.013387, throughput 5.98739K wps
[Epoch 1 Batch 150/173] avg loss 0.0133722, throughput 5.97552K wps
Begin Testing...
[Epoch 1] train avg loss 0.0134937, test acc 0.6427, test avg loss 0.646953, throughput 6.02142K wps
Observed Improvement.
Begin Testing...
[Epoch 2 Batch 30/173] avg loss 0.0126931, throughput 6.15847K wps
[Epoch 2 Batch 60/173] avg loss 0.0126536, throughput 5.98531K wps
[Epoch 2 Batch 90/173] avg loss 0.0126925, throughput 5.98452K wps
[Epoch 2 Batch 120/173] avg loss 0.0126621, throughput 5.99752K wps
[Epoch 2 Batch 150/173] avg loss 0.0125798, throughput 5.99764K wps
Begin Testing...
[Epoch 2] train avg loss 0.0126528, test acc 0.6792, test avg loss 0.624823, throughput 6.02156K wps
Observed Improvement.
Begin Testing...
[Epoch 3 Batch 30/173] avg loss 0.0121127, throughput 6.14078K wps
[Epoch 3 Batch 60/173] avg loss 0.0119276, throughput 6.00723K wps
[Epoch 3 Batch 90/173] avg loss 0.0118131, throughput 6.00946K wps
[Epoch 3 Batch 120/173] avg loss 0.0117727, throughput 6.00472K wps
[Epoch 3 Batch 150/173] avg loss 0.011805, throughput 5.99254K wps
Begin Testing...
[Epoch 3] train avg loss 0.0118673, test acc 0.7229, test avg loss 0.597721, throughput 6.02711K wps
Observed Improvement.
Begin Testing...
[Epoch 4 Batch 30/173] avg loss 0.011348, throughput 6.15615K wps
[Epoch 4 Batch 60/173] avg loss 0.0113082, throughput 5.99574K wps
[Epoch 4 Batch 90/173] avg loss 0.0110846, throughput 6.00084K wps
[Epoch 4 Batch 120/173] avg loss 0.0108557, throughput 5.99788K wps
[Epoch 4 Batch 150/173] avg loss 0.011032, throughput 5.99085K wps
Begin Testing...
[Epoch 4] train avg loss 0.0110834, test acc 0.7438, test avg loss 0.564056, throughput 6.02379K wps
Observed Improvement.
Begin Testing...
[Epoch 5 Batch 30/173] avg loss 0.0102286, throughput 6.14755K wps
[Epoch 5 Batch 60/173] avg loss 0.0100937, throughput 5.99658K wps
[Epoch 5 Batch 90/173] avg loss 0.00998244, throughput 6.0068K wps
[Epoch 5 Batch 120/173] avg loss 0.00985386, throughput 6.00061K wps
[Epoch 5 Batch 150/173] avg loss 0.00981928, throughput 5.983K wps
Begin Testing...
[Epoch 5] train avg loss 0.00995645, test acc 0.7531, test avg loss 0.530249, throughput 6.02539K wps
Observed Improvement.
Begin Testing...
[Epoch 6 Batch 30/173] avg loss 0.00920911, throughput 6.16159K wps
[Epoch 6 Batch 60/173] avg loss 0.00892457, throughput 6.00965K wps
[Epoch 6 Batch 90/173] avg loss 0.00875222, throughput 6.00442K wps
[Epoch 6 Batch 120/173] avg loss 0.00873148, throughput 6.01138K wps
[Epoch 6 Batch 150/173] avg loss 0.00884773, throughput 6.00851K wps
Begin Testing...
[Epoch 6] train avg loss 0.00882031, test acc 0.7750, test avg loss 0.492669, throughput 6.03436K wps
Observed Improvement.
Begin Testing...
[Epoch 7 Batch 30/173] avg loss 0.00801125, throughput 6.14469K wps
[Epoch 7 Batch 60/173] avg loss 0.00753667, throughput 5.99198K wps
[Epoch 7 Batch 90/173] avg loss 0.00793098, throughput 6.00209K wps
[Epoch 7 Batch 120/173] avg loss 0.00755318, throughput 6.01082K wps
[Epoch 7 Batch 150/173] avg loss 0.00772476, throughput 5.99627K wps
Begin Testing...
[Epoch 7] train avg loss 0.00768519, test acc 0.7917, test avg loss 0.46322, throughput 6.02786K wps
Observed Improvement.
Begin Testing...
[Epoch 8 Batch 30/173] avg loss 0.00708206, throughput 6.14718K wps
[Epoch 8 Batch 60/173] avg loss 0.00682648, throughput 5.99835K wps
[Epoch 8 Batch 90/173] avg loss 0.00668424, throughput 6.00196K wps
[Epoch 8 Batch 120/173] avg loss 0.00663253, throughput 5.99823K wps
[Epoch 8 Batch 150/173] avg loss 0.00634008, throughput 6.00598K wps
Begin Testing...
[Epoch 8] train avg loss 0.00672099, test acc 0.7896, test avg loss 0.446713, throughput 6.02575K wps
[Epoch 9 Batch 30/173] avg loss 0.00603589, throughput 6.15219K wps
[Epoch 9 Batch 60/173] avg loss 0.00572573, throughput 6.00969K wps
[Epoch 9 Batch 90/173] avg loss 0.0056169, throughput 6.00021K wps
[Epoch 9 Batch 120/173] avg loss 0.00572099, throughput 6.01188K wps
[Epoch 9 Batch 150/173] avg loss 0.00580982, throughput 6.00466K wps
Begin Testing...
[Epoch 9] train avg loss 0.00577859, test acc 0.7906, test avg loss 0.438186, throughput 6.03087K wps
[Epoch 10 Batch 30/173] avg loss 0.00520029, throughput 6.15365K wps
[Epoch 10 Batch 60/173] avg loss 0.00516638, throughput 6.01865K wps
[Epoch 10 Batch 90/173] avg loss 0.00509385, throughput 5.98817K wps
[Epoch 10 Batch 120/173] avg loss 0.0051803, throughput 6.003K wps
[Epoch 10 Batch 150/173] avg loss 0.00475664, throughput 6.0043K wps
Begin Testing...
[Epoch 10] train avg loss 0.00508571, test acc 0.7948, test avg loss 0.425875, throughput 6.02897K wps
Observed Improvement.
Begin Testing...
[Epoch 11 Batch 30/173] avg loss 0.00422291, throughput 6.15398K wps
[Epoch 11 Batch 60/173] avg loss 0.0044262, throughput 6.00863K wps
[Epoch 11 Batch 90/173] avg loss 0.00436448, throughput 6.00207K wps
[Epoch 11 Batch 120/173] avg loss 0.00435442, throughput 6.01372K wps
[Epoch 11 Batch 150/173] avg loss 0.0042361, throughput 6.00313K wps
Begin Testing...
[Epoch 11] train avg loss 0.0043574, test acc 0.7979, test avg loss 0.421443, throughput 6.03254K wps
Observed Improvement.
Begin Testing...
[Epoch 12 Batch 30/173] avg loss 0.0036146, throughput 6.13927K wps
[Epoch 12 Batch 60/173] avg loss 0.00356935, throughput 5.99358K wps
[Epoch 12 Batch 90/173] avg loss 0.00364169, throughput 6.00133K wps
[Epoch 12 Batch 120/173] avg loss 0.00393446, throughput 6.0098K wps
[Epoch 12 Batch 150/173] avg loss 0.00385562, throughput 6.00203K wps
Begin Testing...
[Epoch 12] train avg loss 0.00370939, test acc 0.8031, test avg loss 0.4218, throughput 6.02552K wps
Observed Improvement.
Begin Testing...
[Epoch 13 Batch 30/173] avg loss 0.00313534, throughput 6.13477K wps
[Epoch 13 Batch 60/173] avg loss 0.0032704, throughput 6.01276K wps
[Epoch 13 Batch 90/173] avg loss 0.00311775, throughput 6.01234K wps
[Epoch 13 Batch 120/173] avg loss 0.003287, throughput 6.00847K wps
[Epoch 13 Batch 150/173] avg loss 0.00302073, throughput 6.01107K wps
Begin Testing...
[Epoch 13] train avg loss 0.00319079, test acc 0.7979, test avg loss 0.429579, throughput 6.03335K wps
[Epoch 14 Batch 30/173] avg loss 0.00280336, throughput 6.15804K wps
[Epoch 14 Batch 60/173] avg loss 0.00262123, throughput 5.99094K wps
[Epoch 14 Batch 90/173] avg loss 0.00269573, throughput 5.99287K wps
[Epoch 14 Batch 120/173] avg loss 0.00284233, throughput 5.99331K wps
[Epoch 14 Batch 150/173] avg loss 0.00270618, throughput 5.99293K wps
Begin Testing...
[Epoch 14] train avg loss 0.00277621, test acc 0.7917, test avg loss 0.439042, throughput 6.02078K wps
[Epoch 15 Batch 30/173] avg loss 0.00225519, throughput 6.14043K wps
[Epoch 15 Batch 60/173] avg loss 0.0024965, throughput 5.99534K wps
[Epoch 15 Batch 90/173] avg loss 0.00233651, throughput 5.99032K wps
[Epoch 15 Batch 120/173] avg loss 0.0023966, throughput 5.99736K wps
[Epoch 15 Batch 150/173] avg loss 0.00235169, throughput 5.9878K wps
Begin Testing...
[Epoch 15] train avg loss 0.0023499, test acc 0.7875, test avg loss 0.447015, throughput 6.01769K wps
[Epoch 16 Batch 30/173] avg loss 0.00184629, throughput 6.13273K wps
[Epoch 16 Batch 60/173] avg loss 0.00189787, throughput 6.00091K wps
[Epoch 16 Batch 90/173] avg loss 0.00186493, throughput 6.00925K wps
[Epoch 16 Batch 120/173] avg loss 0.00203727, throughput 6.00512K wps
[Epoch 16 Batch 150/173] avg loss 0.00219942, throughput 5.99042K wps
Begin Testing...
[Epoch 16] train avg loss 0.0019909, test acc 0.7896, test avg loss 0.454054, throughput 6.02363K wps
[Epoch 17 Batch 30/173] avg loss 0.00161993, throughput 6.13543K wps
[Epoch 17 Batch 60/173] avg loss 0.00140175, throughput 5.99669K wps
[Epoch 17 Batch 90/173] avg loss 0.00180512, throughput 5.99838K wps
[Epoch 17 Batch 120/173] avg loss 0.00169887, throughput 6.0009K wps
[Epoch 17 Batch 150/173] avg loss 0.00163606, throughput 6.00346K wps
Begin Testing...
[Epoch 17] train avg loss 0.00165269, test acc 0.7823, test avg loss 0.474201, throughput 6.02236K wps
[Epoch 18 Batch 30/173] avg loss 0.00148725, throughput 6.13603K wps
[Epoch 18 Batch 60/173] avg loss 0.00119155, throughput 6.00395K wps
[Epoch 18 Batch 90/173] avg loss 0.0014361, throughput 5.99162K wps
[Epoch 18 Batch 120/173] avg loss 0.0014689, throughput 5.99757K wps
[Epoch 18 Batch 150/173] avg loss 0.00147976, throughput 5.99289K wps
Begin Testing...
[Epoch 18] train avg loss 0.00143206, test acc 0.7865, test avg loss 0.481719, throughput 6.02045K wps
[Epoch 19 Batch 30/173] avg loss 0.00121163, throughput 6.13951K wps
[Epoch 19 Batch 60/173] avg loss 0.00132071, throughput 5.99576K wps
[Epoch 19 Batch 90/173] avg loss 0.0012416, throughput 5.98436K wps
[Epoch 19 Batch 120/173] avg loss 0.00114614, throughput 5.98538K wps
[Epoch 19 Batch 150/173] avg loss 0.00126761, throughput 5.99943K wps
Begin Testing...
[Epoch 19] train avg loss 0.00122965, test acc 0.7844, test avg loss 0.501592, throughput 6.01862K wps
[Epoch 20 Batch 30/173] avg loss 0.00116185, throughput 6.1515K wps
[Epoch 20 Batch 60/173] avg loss 0.0010401, throughput 5.99465K wps
[Epoch 20 Batch 90/173] avg loss 0.00105093, throughput 5.98729K wps
[Epoch 20 Batch 120/173] avg loss 0.000972319, throughput 6.00493K wps
[Epoch 20 Batch 150/173] avg loss 0.000927751, throughput 6.00374K wps
Begin Testing...
[Epoch 20] train avg loss 0.00105157, test acc 0.7875, test avg loss 0.514082, throughput 6.0266K wps
[Epoch 21 Batch 30/173] avg loss 0.000908449, throughput 6.15217K wps
[Epoch 21 Batch 60/173] avg loss 0.000840617, throughput 6.00797K wps
[Epoch 21 Batch 90/173] avg loss 0.000906757, throughput 6.00404K wps
[Epoch 21 Batch 120/173] avg loss 0.000946138, throughput 5.99505K wps
[Epoch 21 Batch 150/173] avg loss 0.000913791, throughput 5.99016K wps
Begin Testing...
[Epoch 21] train avg loss 0.000910599, test acc 0.7833, test avg loss 0.535986, throughput 6.02476K wps
[Epoch 22 Batch 30/173] avg loss 0.000719485, throughput 6.14147K wps
[Epoch 22 Batch 60/173] avg loss 0.000713407, throughput 6.00219K wps
[Epoch 22 Batch 90/173] avg loss 0.000808647, throughput 6.0056K wps
[Epoch 22 Batch 120/173] avg loss 0.000809542, throughput 6.00196K wps
[Epoch 22 Batch 150/173] avg loss 0.000797648, throughput 5.97935K wps
Begin Testing...
[Epoch 22] train avg loss 0.000763878, test acc 0.7802, test avg loss 0.54896, throughput 6.01053K wps
[Epoch 23 Batch 30/173] avg loss 0.00059341, throughput 6.1354K wps
[Epoch 23 Batch 60/173] avg loss 0.000620205, throughput 5.99749K wps
[Epoch 23 Batch 90/173] avg loss 0.000725005, throughput 5.99473K wps
[Epoch 23 Batch 120/173] avg loss 0.00062013, throughput 5.99261K wps
[Epoch 23 Batch 150/173] avg loss 0.000667575, throughput 5.9913K wps
Begin Testing...
[Epoch 23] train avg loss 0.000659472, test acc 0.7781, test avg loss 0.564318, throughput 6.01902K wps
[Epoch 24 Batch 30/173] avg loss 0.000515347, throughput 6.1482K wps
[Epoch 24 Batch 60/173] avg loss 0.000516114, throughput 5.99591K wps
[Epoch 24 Batch 90/173] avg loss 0.000574915, throughput 6.01165K wps
[Epoch 24 Batch 120/173] avg loss 0.000461714, throughput 6.00235K wps
[Epoch 24 Batch 150/173] avg loss 0.00056877, throughput 6.00702K wps
Begin Testing...
[Epoch 24] train avg loss 0.00054467, test acc 0.7740, test avg loss 0.591115, throughput 6.02938K wps
[Epoch 25 Batch 30/173] avg loss 0.000455865, throughput 6.14511K wps
[Epoch 25 Batch 60/173] avg loss 0.000471508, throughput 5.98446K wps
[Epoch 25 Batch 90/173] avg loss 0.000446787, throughput 6.00775K wps
[Epoch 25 Batch 120/173] avg loss 0.000476302, throughput 5.99008K wps
[Epoch 25 Batch 150/173] avg loss 0.000569805, throughput 6.00674K wps
Begin Testing...
[Epoch 25] train avg loss 0.000487285, test acc 0.7771, test avg loss 0.605051, throughput 6.02334K wps
[Epoch 26 Batch 30/173] avg loss 0.000420682, throughput 6.14934K wps
[Epoch 26 Batch 60/173] avg loss 0.000432169, throughput 6.00767K wps
[Epoch 26 Batch 90/173] avg loss 0.00040321, throughput 5.99474K wps
[Epoch 26 Batch 120/173] avg loss 0.000369591, throughput 5.99339K wps
[Epoch 26 Batch 150/173] avg loss 0.000395387, throughput 5.99386K wps
Begin Testing...
[Epoch 26] train avg loss 0.000411207, test acc 0.7729, test avg loss 0.623382, throughput 6.0247K wps
[Epoch 27 Batch 30/173] avg loss 0.000375495, throughput 6.13386K wps
[Epoch 27 Batch 60/173] avg loss 0.000322589, throughput 6.00458K wps
[Epoch 27 Batch 90/173] avg loss 0.000383267, throughput 5.99931K wps
[Epoch 27 Batch 120/173] avg loss 0.000385562, throughput 5.9947K wps
[Epoch 27 Batch 150/173] avg loss 0.000427685, throughput 6.01588K wps
Begin Testing...
[Epoch 27] train avg loss 0.00038276, test acc 0.7823, test avg loss 0.643942, throughput 6.02573K wps
[Epoch 28 Batch 30/173] avg loss 0.000340139, throughput 6.14959K wps
[Epoch 28 Batch 60/173] avg loss 0.000295424, throughput 5.999K wps
[Epoch 28 Batch 90/173] avg loss 0.000298399, throughput 5.99653K wps
[Epoch 28 Batch 120/173] avg loss 0.00032583, throughput 5.9902K wps
[Epoch 28 Batch 150/173] avg loss 0.000339509, throughput 5.99445K wps
Begin Testing...
[Epoch 28] train avg loss 0.000315155, test acc 0.7771, test avg loss 0.655224, throughput 6.02187K wps
[Epoch 29 Batch 30/173] avg loss 0.00024072, throughput 6.12512K wps
[Epoch 29 Batch 60/173] avg loss 0.000237234, throughput 5.98349K wps
[Epoch 29 Batch 90/173] avg loss 0.000290526, throughput 5.99139K wps
[Epoch 29 Batch 120/173] avg loss 0.000263256, throughput 5.99059K wps
[Epoch 29 Batch 150/173] avg loss 0.000236083, throughput 5.99969K wps
Begin Testing...
[Epoch 29] train avg loss 0.000258207, test acc 0.7719, test avg loss 0.674913, throughput 6.01655K wps
[Epoch 30 Batch 30/173] avg loss 0.000288203, throughput 6.14839K wps
[Epoch 30 Batch 60/173] avg loss 0.000196499, throughput 6.00993K wps
[Epoch 30 Batch 90/173] avg loss 0.000222217, throughput 5.99874K wps
[Epoch 30 Batch 120/173] avg loss 0.000243598, throughput 5.99191K wps
[Epoch 30 Batch 150/173] avg loss 0.000216675, throughput 5.99314K wps
Begin Testing...
[Epoch 30] train avg loss 0.000237826, test acc 0.7750, test avg loss 0.688644, throughput 6.02328K wps
[Epoch 31 Batch 30/173] avg loss 0.000216814, throughput 6.14566K wps
[Epoch 31 Batch 60/173] avg loss 0.00017569, throughput 5.98273K wps
[Epoch 31 Batch 90/173] avg loss 0.000204833, throughput 5.98396K wps
[Epoch 31 Batch 120/173] avg loss 0.000209037, throughput 5.99203K wps
[Epoch 31 Batch 150/173] avg loss 0.000208556, throughput 5.99815K wps
Begin Testing...
[Epoch 31] train avg loss 0.000204042, test acc 0.7792, test avg loss 0.710086, throughput 6.01506K wps
[Epoch 32 Batch 30/173] avg loss 0.000167858, throughput 6.1435K wps
[Epoch 32 Batch 60/173] avg loss 0.000167541, throughput 5.99727K wps
[Epoch 32 Batch 90/173] avg loss 0.000198812, throughput 5.99873K wps
[Epoch 32 Batch 120/173] avg loss 0.000196749, throughput 5.98835K wps
[Epoch 32 Batch 150/173] avg loss 0.000178629, throughput 6.00675K wps
Begin Testing...
[Epoch 32] train avg loss 0.000182618, test acc 0.7760, test avg loss 0.7325, throughput 6.02272K wps
[Epoch 33 Batch 30/173] avg loss 0.000158514, throughput 6.13463K wps
[Epoch 33 Batch 60/173] avg loss 0.000155034, throughput 5.9938K wps
[Epoch 33 Batch 90/173] avg loss 0.000130181, throughput 5.99861K wps
[Epoch 33 Batch 120/173] avg loss 0.000153442, throughput 6.00477K wps
[Epoch 33 Batch 150/173] avg loss 0.000200329, throughput 5.98815K wps
Begin Testing...
[Epoch 33] train avg loss 0.000159253, test acc 0.7760, test avg loss 0.748783, throughput 6.02304K wps
[Epoch 34 Batch 30/173] avg loss 0.000118553, throughput 6.14866K wps
[Epoch 34 Batch 60/173] avg loss 0.000179114, throughput 6.01068K wps
[Epoch 34 Batch 90/173] avg loss 0.000148658, throughput 5.99706K wps
[Epoch 34 Batch 120/173] avg loss 0.000152183, throughput 5.99551K wps
[Epoch 34 Batch 150/173] avg loss 0.000130745, throughput 5.99786K wps
Begin Testing...
[Epoch 34] train avg loss 0.000146781, test acc 0.7688, test avg loss 0.762755, throughput 6.02671K wps
[Epoch 35 Batch 30/173] avg loss 0.000114171, throughput 6.13655K wps
[Epoch 35 Batch 60/173] avg loss 0.000123017, throughput 5.99451K wps
[Epoch 35 Batch 90/173] avg loss 0.000124415, throughput 5.99399K wps
[Epoch 35 Batch 120/173] avg loss 0.000125108, throughput 5.99552K wps
[Epoch 35 Batch 150/173] avg loss 0.000128127, throughput 6.00827K wps
Begin Testing...
[Epoch 35] train avg loss 0.000124101, test acc 0.7708, test avg loss 0.781585, throughput 6.02082K wps
[Epoch 36 Batch 30/173] avg loss 0.000112507, throughput 6.13681K wps
[Epoch 36 Batch 60/173] avg loss 0.000103304, throughput 5.99867K wps
[Epoch 36 Batch 90/173] avg loss 0.000111012, throughput 5.98864K wps
[Epoch 36 Batch 120/173] avg loss 0.000115807, throughput 6.00046K wps
[Epoch 36 Batch 150/173] avg loss 0.000133154, throughput 5.99546K wps
Begin Testing...
[Epoch 36] train avg loss 0.000114445, test acc 0.7729, test avg loss 0.807002, throughput 6.0207K wps
[Epoch 37 Batch 30/173] avg loss 8.65082e-05, throughput 6.15573K wps
[Epoch 37 Batch 60/173] avg loss 0.000105927, throughput 6.00032K wps
[Epoch 37 Batch 90/173] avg loss 9.70918e-05, throughput 6.00435K wps
[Epoch 37 Batch 120/173] avg loss 0.000119995, throughput 6.00637K wps
[Epoch 37 Batch 150/173] avg loss 9.35392e-05, throughput 6.00513K wps
Begin Testing...
[Epoch 37] train avg loss 9.86084e-05, test acc 0.7740, test avg loss 0.831461, throughput 6.03146K wps
[Epoch 38 Batch 30/173] avg loss 8.31267e-05, throughput 6.13119K wps
[Epoch 38 Batch 60/173] avg loss 8.08283e-05, throughput 6.00158K wps
[Epoch 38 Batch 90/173] avg loss 0.000105628, throughput 5.99255K wps
[Epoch 38 Batch 120/173] avg loss 7.1484e-05, throughput 6.00474K wps
[Epoch 38 Batch 150/173] avg loss 8.93564e-05, throughput 6.00726K wps
Begin Testing...
[Epoch 38] train avg loss 8.72553e-05, test acc 0.7729, test avg loss 0.855371, throughput 6.025K wps
[Epoch 39 Batch 30/173] avg loss 0.000103099, throughput 6.16675K wps
[Epoch 39 Batch 60/173] avg loss 7.19477e-05, throughput 6.00824K wps
[Epoch 39 Batch 90/173] avg loss 7.48888e-05, throughput 6.01068K wps
[Epoch 39 Batch 120/173] avg loss 7.09811e-05, throughput 6.00683K wps
[Epoch 39 Batch 150/173] avg loss 8.94178e-05, throughput 5.99786K wps
Begin Testing...
[Epoch 39] train avg loss 8.14007e-05, test acc 0.7656, test avg loss 0.856098, throughput 6.03375K wps
[Epoch 40 Batch 30/173] avg loss 5.51839e-05, throughput 6.13523K wps
[Epoch 40 Batch 60/173] avg loss 7.16611e-05, throughput 6.00125K wps
[Epoch 40 Batch 90/173] avg loss 7.31127e-05, throughput 5.98821K wps
[Epoch 40 Batch 120/173] avg loss 7.29248e-05, throughput 5.99787K wps
[Epoch 40 Batch 150/173] avg loss 6.43779e-05, throughput 5.9912K wps
Begin Testing...
[Epoch 40] train avg loss 6.76895e-05, test acc 0.7698, test avg loss 0.874498, throughput 6.01847K wps
[Epoch 41 Batch 30/173] avg loss 7.1401e-05, throughput 6.14351K wps
[Epoch 41 Batch 60/173] avg loss 5.77872e-05, throughput 5.9982K wps
[Epoch 41 Batch 90/173] avg loss 5.30281e-05, throughput 6.00193K wps
[Epoch 41 Batch 120/173] avg loss 5.93168e-05, throughput 5.99106K wps
[Epoch 41 Batch 150/173] avg loss 5.36144e-05, throughput 5.98947K wps
Begin Testing...
[Epoch 41] train avg loss 6.2196e-05, test acc 0.7677, test avg loss 0.900005, throughput 6.02012K wps
[Epoch 42 Batch 30/173] avg loss 6.62124e-05, throughput 6.14927K wps
[Epoch 42 Batch 60/173] avg loss 5.69166e-05, throughput 5.99468K wps
[Epoch 42 Batch 90/173] avg loss 5.24907e-05, throughput 6.003K wps
[Epoch 42 Batch 120/173] avg loss 6.00895e-05, throughput 6.00029K wps
[Epoch 42 Batch 150/173] avg loss 5.65372e-05, throughput 6.00535K wps
Begin Testing...
[Epoch 42] train avg loss 5.70309e-05, test acc 0.7708, test avg loss 0.913252, throughput 6.02806K wps
[Epoch 43 Batch 30/173] avg loss 4.5009e-05, throughput 6.15226K wps
[Epoch 43 Batch 60/173] avg loss 4.69787e-05, throughput 5.99807K wps
[Epoch 43 Batch 90/173] avg loss 3.9626e-05, throughput 5.99307K wps
[Epoch 43 Batch 120/173] avg loss 4.82095e-05, throughput 6.00687K wps
[Epoch 43 Batch 150/173] avg loss 4.95844e-05, throughput 6.00047K wps
Begin Testing...
[Epoch 43] train avg loss 4.63819e-05, test acc 0.7708, test avg loss 0.930763, throughput 6.02405K wps
[Epoch 44 Batch 30/173] avg loss 4.16353e-05, throughput 6.13824K wps
[Epoch 44 Batch 60/173] avg loss 4.13406e-05, throughput 5.98357K wps
[Epoch 44 Batch 90/173] avg loss 4.29996e-05, throughput 6.00327K wps
[Epoch 44 Batch 120/173] avg loss 3.89572e-05, throughput 5.9968K wps
[Epoch 44 Batch 150/173] avg loss 5.13097e-05, throughput 5.99386K wps
Begin Testing...
[Epoch 44] train avg loss 4.40049e-05, test acc 0.7719, test avg loss 0.956705, throughput 6.01935K wps
[Epoch 45 Batch 30/173] avg loss 3.38202e-05, throughput 6.13566K wps
[Epoch 45 Batch 60/173] avg loss 5.14419e-05, throughput 5.98386K wps
[Epoch 45 Batch 90/173] avg loss 4.46879e-05, throughput 5.98191K wps
[Epoch 45 Batch 120/173] avg loss 3.27404e-05, throughput 5.98635K wps
[Epoch 45 Batch 150/173] avg loss 3.84316e-05, throughput 5.98728K wps
Begin Testing...
[Epoch 45] train avg loss 4.02954e-05, test acc 0.7750, test avg loss 0.956887, throughput 6.01201K wps
[Epoch 46 Batch 30/173] avg loss 3.25735e-05, throughput 6.15159K wps
[Epoch 46 Batch 60/173] avg loss 4.97693e-05, throughput 6.00763K wps
[Epoch 46 Batch 90/173] avg loss 2.79243e-05, throughput 5.98177K wps
[Epoch 46 Batch 120/173] avg loss 3.30032e-05, throughput 5.99127K wps
[Epoch 46 Batch 150/173] avg loss 5.09833e-05, throughput 5.99715K wps
Begin Testing...
[Epoch 46] train avg loss 3.78899e-05, test acc 0.7750, test avg loss 0.977744, throughput 6.02156K wps
[Epoch 47 Batch 30/173] avg loss 2.65511e-05, throughput 6.14865K wps
[Epoch 47 Batch 60/173] avg loss 2.46398e-05, throughput 6.01175K wps
[Epoch 47 Batch 90/173] avg loss 2.35137e-05, throughput 6.01025K wps
[Epoch 47 Batch 120/173] avg loss 3.67228e-05, throughput 6.0058K wps
[Epoch 47 Batch 150/173] avg loss 4.61561e-05, throughput 6.00275K wps
Begin Testing...
[Epoch 47] train avg loss 3.11463e-05, test acc 0.7719, test avg loss 1.00431, throughput 6.03254K wps
[Epoch 48 Batch 30/173] avg loss 5.09996e-05, throughput 6.14746K wps
[Epoch 48 Batch 60/173] avg loss 3.96344e-05, throughput 6.01135K wps
[Epoch 48 Batch 90/173] avg loss 2.93851e-05, throughput 6.00821K wps
[Epoch 48 Batch 120/173] avg loss 2.82347e-05, throughput 6.01564K wps
[Epoch 48 Batch 150/173] avg loss 3.83916e-05, throughput 6.00706K wps
Begin Testing...
[Epoch 48] train avg loss 3.69513e-05, test acc 0.7729, test avg loss 1.00372, throughput 6.03463K wps
[Epoch 49 Batch 30/173] avg loss 2.28981e-05, throughput 6.13901K wps
[Epoch 49 Batch 60/173] avg loss 2.98679e-05, throughput 5.98595K wps
[Epoch 49 Batch 90/173] avg loss 2.16042e-05, throughput 5.99239K wps
[Epoch 49 Batch 120/173] avg loss 2.26435e-05, throughput 5.98941K wps
[Epoch 49 Batch 150/173] avg loss 2.70652e-05, throughput 5.99287K wps
Begin Testing...
[Epoch 49] train avg loss 2.49774e-05, test acc 0.7729, test avg loss 1.02626, throughput 6.0164K wps
[Epoch 50 Batch 30/173] avg loss 2.06826e-05, throughput 6.16157K wps
[Epoch 50 Batch 60/173] avg loss 2.68891e-05, throughput 5.99658K wps
[Epoch 50 Batch 90/173] avg loss 2.12996e-05, throughput 5.9928K wps
[Epoch 50 Batch 120/173] avg loss 2.24712e-05, throughput 5.9894K wps
[Epoch 50 Batch 150/173] avg loss 2.018e-05, throughput 5.99066K wps
Begin Testing...
[Epoch 50] train avg loss 2.23749e-05, test acc 0.7750, test avg loss 1.04003, throughput 6.02328K wps
[Epoch 51 Batch 30/173] avg loss 2.49419e-05, throughput 6.15064K wps
[Epoch 51 Batch 60/173] avg loss 2.15747e-05, throughput 5.98455K wps
[Epoch 51 Batch 90/173] avg loss 2.45592e-05, throughput 5.99158K wps
[Epoch 51 Batch 120/173] avg loss 2.58989e-05, throughput 5.98914K wps
[Epoch 51 Batch 150/173] avg loss 3.00606e-05, throughput 5.99799K wps
Begin Testing...
[Epoch 51] train avg loss 2.45621e-05, test acc 0.7729, test avg loss 1.05196, throughput 6.01937K wps
[Epoch 52 Batch 30/173] avg loss 2.16292e-05, throughput 6.13867K wps
[Epoch 52 Batch 60/173] avg loss 1.96723e-05, throughput 5.98358K wps
[Epoch 52 Batch 90/173] avg loss 2.58075e-05, throughput 5.99035K wps
[Epoch 52 Batch 120/173] avg loss 3.42251e-05, throughput 6.00854K wps
[Epoch 52 Batch 150/173] avg loss 1.7945e-05, throughput 5.98909K wps
Begin Testing...
[Epoch 52] train avg loss 2.25602e-05, test acc 0.7708, test avg loss 1.07175, throughput 6.01837K wps
[Epoch 53 Batch 30/173] avg loss 2.65475e-05, throughput 6.13026K wps
[Epoch 53 Batch 60/173] avg loss 1.59614e-05, throughput 5.97643K wps
[Epoch 53 Batch 90/173] avg loss 1.48087e-05, throughput 5.97941K wps
[Epoch 53 Batch 120/173] avg loss 2.5599e-05, throughput 5.98294K wps
[Epoch 53 Batch 150/173] avg loss 1.66867e-05, throughput 5.98779K wps
Begin Testing...
[Epoch 53] train avg loss 1.94089e-05, test acc 0.7729, test avg loss 1.08414, throughput 6.00825K wps
[Epoch 54 Batch 30/173] avg loss 1.92e-05, throughput 6.13565K wps
[Epoch 54 Batch 60/173] avg loss 1.73498e-05, throughput 5.99654K wps
[Epoch 54 Batch 90/173] avg loss 1.50514e-05, throughput 5.99636K wps
[Epoch 54 Batch 120/173] avg loss 1.54806e-05, throughput 5.98247K wps
[Epoch 54 Batch 150/173] avg loss 1.95478e-05, throughput 6.00913K wps
Begin Testing...
[Epoch 54] train avg loss 1.65973e-05, test acc 0.7677, test avg loss 1.10896, throughput 6.02192K wps
[Epoch 55 Batch 30/173] avg loss 1.49224e-05, throughput 6.14284K wps
[Epoch 55 Batch 60/173] avg loss 1.27845e-05, throughput 5.98959K wps
[Epoch 55 Batch 90/173] avg loss 1.51791e-05, throughput 5.98736K wps
[Epoch 55 Batch 120/173] avg loss 2.12861e-05, throughput 5.98583K wps
[Epoch 55 Batch 150/173] avg loss 1.7762e-05, throughput 6.01245K wps
Begin Testing...
[Epoch 55] train avg loss 1.62819e-05, test acc 0.7750, test avg loss 1.11752, throughput 6.0223K wps
[Epoch 56 Batch 30/173] avg loss 1.71584e-05, throughput 6.14052K wps
[Epoch 56 Batch 60/173] avg loss 1.15805e-05, throughput 5.98412K wps
[Epoch 56 Batch 90/173] avg loss 1.82015e-05, throughput 6.00095K wps
[Epoch 56 Batch 120/173] avg loss 1.24224e-05, throughput 5.99279K wps
[Epoch 56 Batch 150/173] avg loss 1.30086e-05, throughput 5.99682K wps
Begin Testing...
[Epoch 56] train avg loss 1.38637e-05, test acc 0.7698, test avg loss 1.14312, throughput 6.02104K wps
[Epoch 57 Batch 30/173] avg loss 1.3147e-05, throughput 6.15788K wps
[Epoch 57 Batch 60/173] avg loss 1.19746e-05, throughput 6.01478K wps
[Epoch 57 Batch 90/173] avg loss 1.28657e-05, throughput 6.00452K wps
[Epoch 57 Batch 120/173] avg loss 1.666e-05, throughput 5.99201K wps
[Epoch 57 Batch 150/173] avg loss 1.08608e-05, throughput 6.00183K wps
Begin Testing...
[Epoch 57] train avg loss 1.27298e-05, test acc 0.7688, test avg loss 1.16732, throughput 6.03084K wps
[Epoch 58 Batch 30/173] avg loss 1.78025e-05, throughput 6.16425K wps
[Epoch 58 Batch 60/173] avg loss 1.29048e-05, throughput 6.00838K wps
[Epoch 58 Batch 90/173] avg loss 8.34132e-06, throughput 6.00675K wps
[Epoch 58 Batch 120/173] avg loss 1.97687e-05, throughput 6.00535K wps
[Epoch 58 Batch 150/173] avg loss 1.22267e-05, throughput 5.98643K wps
Begin Testing...
[Epoch 58] train avg loss 1.4034e-05, test acc 0.7656, test avg loss 1.16572, throughput 6.0282K wps
[Epoch 59 Batch 30/173] avg loss 1.0666e-05, throughput 6.145K wps
[Epoch 59 Batch 60/173] avg loss 8.08107e-06, throughput 5.99114K wps
[Epoch 59 Batch 90/173] avg loss 8.73227e-06, throughput 6.00593K wps
[Epoch 59 Batch 120/173] avg loss 1.71458e-05, throughput 6.0161K wps
[Epoch 59 Batch 150/173] avg loss 1.23963e-05, throughput 5.99804K wps
Begin Testing...
[Epoch 59] train avg loss 1.16273e-05, test acc 0.7656, test avg loss 1.20131, throughput 6.02746K wps
Test loss 0.443202, test acc 0.7964
Total time cost 358.52s
9596 1066
[Epoch 0 Batch 30/173] avg loss 0.0158435, throughput 5.78367K wps
[Epoch 0 Batch 60/173] avg loss 0.0153895, throughput 5.98213K wps
[Epoch 0 Batch 90/173] avg loss 0.0149871, throughput 6.00498K wps
[Epoch 0 Batch 120/173] avg loss 0.0146208, throughput 6.00556K wps
[Epoch 0 Batch 150/173] avg loss 0.0141016, throughput 5.98683K wps
Begin Testing...
[Epoch 0] train avg loss 0.0149053, test acc 0.5667, test avg loss 0.681236, throughput 5.95829K wps
Observed Improvement.
Begin Testing...
[Epoch 1 Batch 30/173] avg loss 0.013808, throughput 6.13385K wps
[Epoch 1 Batch 60/173] avg loss 0.0138138, throughput 5.99683K wps
[Epoch 1 Batch 90/173] avg loss 0.0132794, throughput 5.99628K wps
[Epoch 1 Batch 120/173] avg loss 0.0136665, throughput 5.99569K wps
[Epoch 1 Batch 150/173] avg loss 0.0131609, throughput 5.99211K wps
Begin Testing...
[Epoch 1] train avg loss 0.0135276, test acc 0.5917, test avg loss 0.655866, throughput 6.02125K wps
Observed Improvement.
Begin Testing...
[Epoch 2 Batch 30/173] avg loss 0.0128377, throughput 6.15289K wps
[Epoch 2 Batch 60/173] avg loss 0.012891, throughput 5.99407K wps
[Epoch 2 Batch 90/173] avg loss 0.0128315, throughput 5.98855K wps
[Epoch 2 Batch 120/173] avg loss 0.0127199, throughput 5.99473K wps
[Epoch 2 Batch 150/173] avg loss 0.0125245, throughput 5.98541K wps
Begin Testing...
[Epoch 2] train avg loss 0.0127438, test acc 0.5958, test avg loss 0.640478, throughput 6.0196K wps
Observed Improvement.
Begin Testing...
[Epoch 3 Batch 30/173] avg loss 0.0121319, throughput 6.14097K wps
[Epoch 3 Batch 60/173] avg loss 0.0119711, throughput 5.99585K wps
[Epoch 3 Batch 90/173] avg loss 0.0118962, throughput 5.99956K wps
[Epoch 3 Batch 120/173] avg loss 0.0118118, throughput 5.99569K wps
[Epoch 3 Batch 150/173] avg loss 0.0117535, throughput 5.99871K wps
Begin Testing...
[Epoch 3] train avg loss 0.0118559, test acc 0.6583, test avg loss 0.607856, throughput 6.02187K wps
Observed Improvement.
Begin Testing...
[Epoch 4 Batch 30/173] avg loss 0.0110734, throughput 6.1453K wps
[Epoch 4 Batch 60/173] avg loss 0.010793, throughput 5.99796K wps
[Epoch 4 Batch 90/173] avg loss 0.010951, throughput 6.0033K wps
[Epoch 4 Batch 120/173] avg loss 0.0108798, throughput 5.99995K wps
[Epoch 4 Batch 150/173] avg loss 0.0109441, throughput 5.9983K wps
Begin Testing...
[Epoch 4] train avg loss 0.0109076, test acc 0.7042, test avg loss 0.572057, throughput 6.0237K wps
Observed Improvement.
Begin Testing...
[Epoch 5 Batch 30/173] avg loss 0.0103996, throughput 6.14854K wps
[Epoch 5 Batch 60/173] avg loss 0.0100441, throughput 6.00123K wps
[Epoch 5 Batch 90/173] avg loss 0.0099558, throughput 5.99106K wps
[Epoch 5 Batch 120/173] avg loss 0.00971262, throughput 5.98574K wps
[Epoch 5 Batch 150/173] avg loss 0.00962347, throughput 6.00305K wps
Begin Testing...
[Epoch 5] train avg loss 0.00991587, test acc 0.7583, test avg loss 0.529701, throughput 6.02235K wps
Observed Improvement.
Begin Testing...
[Epoch 6 Batch 30/173] avg loss 0.00885845, throughput 6.13321K wps
[Epoch 6 Batch 60/173] avg loss 0.00884777, throughput 5.99784K wps
[Epoch 6 Batch 90/173] avg loss 0.00878407, throughput 6.0096K wps
[Epoch 6 Batch 120/173] avg loss 0.00840287, throughput 5.98413K wps
[Epoch 6 Batch 150/173] avg loss 0.0088139, throughput 5.98935K wps
Begin Testing...
[Epoch 6] train avg loss 0.00867712, test acc 0.7656, test avg loss 0.494522, throughput 6.02204K wps
Observed Improvement.
Begin Testing...
[Epoch 7 Batch 30/173] avg loss 0.00789573, throughput 6.15812K wps
[Epoch 7 Batch 60/173] avg loss 0.00790906, throughput 6.00148K wps
[Epoch 7 Batch 90/173] avg loss 0.00751374, throughput 5.99274K wps
[Epoch 7 Batch 120/173] avg loss 0.00771023, throughput 5.99214K wps
[Epoch 7 Batch 150/173] avg loss 0.00757431, throughput 5.99617K wps
Begin Testing...
[Epoch 7] train avg loss 0.00767571, test acc 0.7906, test avg loss 0.464775, throughput 6.02495K wps
Observed Improvement.
Begin Testing...
[Epoch 8 Batch 30/173] avg loss 0.00663987, throughput 6.14602K wps
[Epoch 8 Batch 60/173] avg loss 0.00710273, throughput 6.0041K wps
[Epoch 8 Batch 90/173] avg loss 0.00661469, throughput 5.99857K wps
[Epoch 8 Batch 120/173] avg loss 0.00641725, throughput 6.00382K wps
[Epoch 8 Batch 150/173] avg loss 0.00660601, throughput 6.00124K wps
Begin Testing...
[Epoch 8] train avg loss 0.00667777, test acc 0.8010, test avg loss 0.442144, throughput 6.02856K wps
Observed Improvement.
Begin Testing...
[Epoch 9 Batch 30/173] avg loss 0.00567678, throughput 6.14346K wps
[Epoch 9 Batch 60/173] avg loss 0.00577933, throughput 6.00608K wps
[Epoch 9 Batch 90/173] avg loss 0.00561931, throughput 5.99967K wps
[Epoch 9 Batch 120/173] avg loss 0.0061136, throughput 6.00951K wps
[Epoch 9 Batch 150/173] avg loss 0.00573564, throughput 6.0093K wps
Begin Testing...
[Epoch 9] train avg loss 0.00577969, test acc 0.7990, test avg loss 0.427539, throughput 6.02906K wps
[Epoch 10 Batch 30/173] avg loss 0.00493566, throughput 6.15307K wps
[Epoch 10 Batch 60/173] avg loss 0.00523769, throughput 5.99136K wps
[Epoch 10 Batch 90/173] avg loss 0.00519011, throughput 6.00138K wps
[Epoch 10 Batch 120/173] avg loss 0.00485507, throughput 5.97691K wps
[Epoch 10 Batch 150/173] avg loss 0.00480001, throughput 5.998K wps
Begin Testing...
[Epoch 10] train avg loss 0.00499918, test acc 0.8052, test avg loss 0.419077, throughput 6.02158K wps
Observed Improvement.
Begin Testing...
[Epoch 11 Batch 30/173] avg loss 0.00456043, throughput 6.14166K wps
[Epoch 11 Batch 60/173] avg loss 0.00430885, throughput 6.00847K wps
[Epoch 11 Batch 90/173] avg loss 0.0041594, throughput 5.98981K wps
[Epoch 11 Batch 120/173] avg loss 0.00437315, throughput 5.99454K wps
[Epoch 11 Batch 150/173] avg loss 0.00416214, throughput 6.0023K wps
Begin Testing...
[Epoch 11] train avg loss 0.0043084, test acc 0.8042, test avg loss 0.416575, throughput 6.02449K wps
[Epoch 12 Batch 30/173] avg loss 0.00392394, throughput 6.1487K wps
[Epoch 12 Batch 60/173] avg loss 0.0039144, throughput 5.98807K wps
[Epoch 12 Batch 90/173] avg loss 0.00376028, throughput 5.99117K wps
[Epoch 12 Batch 120/173] avg loss 0.00357737, throughput 6.00699K wps
[Epoch 12 Batch 150/173] avg loss 0.00340043, throughput 5.99398K wps
Begin Testing...
[Epoch 12] train avg loss 0.00366773, test acc 0.8052, test avg loss 0.417872, throughput 6.02267K wps
Observed Improvement.
Begin Testing...
[Epoch 13 Batch 30/173] avg loss 0.00310148, throughput 6.14463K wps
[Epoch 13 Batch 60/173] avg loss 0.00318371, throughput 5.98326K wps
[Epoch 13 Batch 90/173] avg loss 0.0033047, throughput 5.99747K wps
[Epoch 13 Batch 120/173] avg loss 0.00308454, throughput 5.99702K wps
[Epoch 13 Batch 150/173] avg loss 0.0030913, throughput 5.98664K wps
Begin Testing...
[Epoch 13] train avg loss 0.00314182, test acc 0.8083, test avg loss 0.423146, throughput 6.02039K wps
Observed Improvement.
Begin Testing...
[Epoch 14 Batch 30/173] avg loss 0.00266662, throughput 6.136K wps
[Epoch 14 Batch 60/173] avg loss 0.00250757, throughput 5.9993K wps
[Epoch 14 Batch 90/173] avg loss 0.00280197, throughput 6.00029K wps
[Epoch 14 Batch 120/173] avg loss 0.00266338, throughput 5.99532K wps
[Epoch 14 Batch 150/173] avg loss 0.00261393, throughput 5.98843K wps
Begin Testing...
[Epoch 14] train avg loss 0.00267334, test acc 0.8083, test avg loss 0.430268, throughput 6.02154K wps
Observed Improvement.
Begin Testing...
[Epoch 15 Batch 30/173] avg loss 0.00219751, throughput 6.13788K wps
[Epoch 15 Batch 60/173] avg loss 0.00232743, throughput 6.00094K wps
[Epoch 15 Batch 90/173] avg loss 0.0022562, throughput 5.98922K wps
[Epoch 15 Batch 120/173] avg loss 0.0022939, throughput 5.98809K wps
[Epoch 15 Batch 150/173] avg loss 0.00227166, throughput 6.00542K wps
Begin Testing...
[Epoch 15] train avg loss 0.00225966, test acc 0.8042, test avg loss 0.439972, throughput 6.02112K wps
[Epoch 16 Batch 30/173] avg loss 0.00200517, throughput 6.15083K wps
[Epoch 16 Batch 60/173] avg loss 0.00187654, throughput 5.97701K wps
[Epoch 16 Batch 90/173] avg loss 0.00198021, throughput 5.97869K wps
[Epoch 16 Batch 120/173] avg loss 0.00201509, throughput 5.99323K wps
[Epoch 16 Batch 150/173] avg loss 0.00175291, throughput 5.997K wps
Begin Testing...
[Epoch 16] train avg loss 0.0019407, test acc 0.8000, test avg loss 0.452803, throughput 6.01573K wps
[Epoch 17 Batch 30/173] avg loss 0.0014937, throughput 6.13618K wps
[Epoch 17 Batch 60/173] avg loss 0.00165712, throughput 5.97955K wps
[Epoch 17 Batch 90/173] avg loss 0.00157267, throughput 5.98457K wps
[Epoch 17 Batch 120/173] avg loss 0.00172852, throughput 5.99107K wps
[Epoch 17 Batch 150/173] avg loss 0.00172397, throughput 5.98476K wps
Begin Testing...
[Epoch 17] train avg loss 0.00164415, test acc 0.8042, test avg loss 0.466566, throughput 6.01206K wps
[Epoch 18 Batch 30/173] avg loss 0.00140285, throughput 6.14507K wps
[Epoch 18 Batch 60/173] avg loss 0.00140044, throughput 5.99625K wps
[Epoch 18 Batch 90/173] avg loss 0.00139486, throughput 5.97636K wps
[Epoch 18 Batch 120/173] avg loss 0.00132088, throughput 5.99229K wps
[Epoch 18 Batch 150/173] avg loss 0.00144974, throughput 5.99622K wps
Begin Testing...
[Epoch 18] train avg loss 0.00137905, test acc 0.8031, test avg loss 0.477957, throughput 6.0171K wps
[Epoch 19 Batch 30/173] avg loss 0.00116531, throughput 6.12746K wps
[Epoch 19 Batch 60/173] avg loss 0.00122101, throughput 5.98085K wps
[Epoch 19 Batch 90/173] avg loss 0.00109427, throughput 6.00187K wps
[Epoch 19 Batch 120/173] avg loss 0.00128482, throughput 5.991K wps
[Epoch 19 Batch 150/173] avg loss 0.00113075, throughput 5.99137K wps
Begin Testing...
[Epoch 19] train avg loss 0.001164, test acc 0.7969, test avg loss 0.499009, throughput 6.01757K wps
[Epoch 20 Batch 30/173] avg loss 0.000977873, throughput 6.13062K wps
[Epoch 20 Batch 60/173] avg loss 0.00101602, throughput 5.99635K wps
[Epoch 20 Batch 90/173] avg loss 0.000872978, throughput 5.99245K wps
[Epoch 20 Batch 120/173] avg loss 0.000983699, throughput 5.99484K wps
[Epoch 20 Batch 150/173] avg loss 0.00106616, throughput 5.99179K wps
Begin Testing...
[Epoch 20] train avg loss 0.000982108, test acc 0.7927, test avg loss 0.516616, throughput 6.0177K wps
[Epoch 21 Batch 30/173] avg loss 0.000785568, throughput 6.14503K wps
[Epoch 21 Batch 60/173] avg loss 0.00075596, throughput 5.99437K wps
[Epoch 21 Batch 90/173] avg loss 0.000817744, throughput 5.99016K wps
[Epoch 21 Batch 120/173] avg loss 0.000925075, throughput 5.993K wps
[Epoch 21 Batch 150/173] avg loss 0.000907763, throughput 6.00225K wps
Begin Testing...
[Epoch 21] train avg loss 0.000848113, test acc 0.7802, test avg loss 0.539583, throughput 6.02079K wps
[Epoch 22 Batch 30/173] avg loss 0.000770749, throughput 6.15061K wps
[Epoch 22 Batch 60/173] avg loss 0.000629936, throughput 6.00024K wps
[Epoch 22 Batch 90/173] avg loss 0.000764366, throughput 5.9935K wps
[Epoch 22 Batch 120/173] avg loss 0.000730998, throughput 6.00433K wps
[Epoch 22 Batch 150/173] avg loss 0.000742521, throughput 5.9933K wps
Begin Testing...
[Epoch 22] train avg loss 0.000736286, test acc 0.7865, test avg loss 0.553739, throughput 6.02462K wps
[Epoch 23 Batch 30/173] avg loss 0.000573724, throughput 6.09311K wps
[Epoch 23 Batch 60/173] avg loss 0.000624053, throughput 5.97987K wps
[Epoch 23 Batch 90/173] avg loss 0.00057735, throughput 5.9922K wps
[Epoch 23 Batch 120/173] avg loss 0.000600096, throughput 5.99282K wps
[Epoch 23 Batch 150/173] avg loss 0.000596952, throughput 5.99271K wps
Begin Testing...
[Epoch 23] train avg loss 0.000598268, test acc 0.7854, test avg loss 0.577994, throughput 6.00893K wps
[Epoch 24 Batch 30/173] avg loss 0.000528057, throughput 6.14508K wps
[Epoch 24 Batch 60/173] avg loss 0.000577625, throughput 6.00661K wps
[Epoch 24 Batch 90/173] avg loss 0.000462484, throughput 6.00492K wps
[Epoch 24 Batch 120/173] avg loss 0.000438912, throughput 5.99767K wps
[Epoch 24 Batch 150/173] avg loss 0.000559132, throughput 5.98491K wps
Begin Testing...
[Epoch 24] train avg loss 0.000519145, test acc 0.7854, test avg loss 0.59634, throughput 6.02275K wps
[Epoch 25 Batch 30/173] avg loss 0.000410007, throughput 6.15241K wps
[Epoch 25 Batch 60/173] avg loss 0.000442305, throughput 5.99318K wps
[Epoch 25 Batch 90/173] avg loss 0.000454673, throughput 5.99389K wps
[Epoch 25 Batch 120/173] avg loss 0.000447849, throughput 5.98716K wps
[Epoch 25 Batch 150/173] avg loss 0.000431391, throughput 5.9922K wps
Begin Testing...
[Epoch 25] train avg loss 0.000435518, test acc 0.7792, test avg loss 0.620181, throughput 6.01942K wps
[Epoch 26 Batch 30/173] avg loss 0.000346889, throughput 6.14302K wps
[Epoch 26 Batch 60/173] avg loss 0.000387277, throughput 5.99076K wps
[Epoch 26 Batch 90/173] avg loss 0.000362495, throughput 5.99231K wps
[Epoch 26 Batch 120/173] avg loss 0.000351326, throughput 5.98557K wps
[Epoch 26 Batch 150/173] avg loss 0.000414665, throughput 5.98596K wps
Begin Testing...
[Epoch 26] train avg loss 0.000377399, test acc 0.7844, test avg loss 0.640463, throughput 6.01587K wps
[Epoch 27 Batch 30/173] avg loss 0.000325767, throughput 6.1405K wps
[Epoch 27 Batch 60/173] avg loss 0.000351124, throughput 5.99471K wps
[Epoch 27 Batch 90/173] avg loss 0.000322925, throughput 5.99053K wps
[Epoch 27 Batch 120/173] avg loss 0.000321151, throughput 5.98861K wps
[Epoch 27 Batch 150/173] avg loss 0.000361472, throughput 5.99918K wps
Begin Testing...
[Epoch 27] train avg loss 0.000336079, test acc 0.7885, test avg loss 0.661509, throughput 6.02098K wps
[Epoch 28 Batch 30/173] avg loss 0.000248157, throughput 6.13938K wps
[Epoch 28 Batch 60/173] avg loss 0.000255141, throughput 5.98987K wps
[Epoch 28 Batch 90/173] avg loss 0.000251391, throughput 6.00296K wps
[Epoch 28 Batch 120/173] avg loss 0.00031067, throughput 5.99068K wps
[Epoch 28 Batch 150/173] avg loss 0.000312505, throughput 5.98887K wps
Begin Testing...
[Epoch 28] train avg loss 0.00028399, test acc 0.7781, test avg loss 0.681619, throughput 6.0186K wps
[Epoch 29 Batch 30/173] avg loss 0.000235079, throughput 6.16394K wps
[Epoch 29 Batch 60/173] avg loss 0.00023429, throughput 6.01553K wps
[Epoch 29 Batch 90/173] avg loss 0.000251803, throughput 6.01266K wps
[Epoch 29 Batch 120/173] avg loss 0.000283442, throughput 6.00826K wps
[Epoch 29 Batch 150/173] avg loss 0.000223779, throughput 5.9924K wps
Begin Testing...
[Epoch 29] train avg loss 0.000244953, test acc 0.7792, test avg loss 0.700551, throughput 6.03325K wps
[Epoch 30 Batch 30/173] avg loss 0.000216628, throughput 6.14319K wps
[Epoch 30 Batch 60/173] avg loss 0.000198385, throughput 5.98882K wps
[Epoch 30 Batch 90/173] avg loss 0.000209393, throughput 5.99667K wps
[Epoch 30 Batch 120/173] avg loss 0.000222519, throughput 5.9855K wps
[Epoch 30 Batch 150/173] avg loss 0.00022006, throughput 6.00608K wps
Begin Testing...
[Epoch 30] train avg loss 0.000210899, test acc 0.7771, test avg loss 0.725783, throughput 6.0215K wps
[Epoch 31 Batch 30/173] avg loss 0.000178931, throughput 6.14227K wps
[Epoch 31 Batch 60/173] avg loss 0.000239617, throughput 5.99863K wps
[Epoch 31 Batch 90/173] avg loss 0.000195263, throughput 5.99085K wps
[Epoch 31 Batch 120/173] avg loss 0.000186732, throughput 6.00675K wps
[Epoch 31 Batch 150/173] avg loss 0.000204188, throughput 6.00885K wps
Begin Testing...
[Epoch 31] train avg loss 0.000200896, test acc 0.7729, test avg loss 0.754008, throughput 6.02788K wps
[Epoch 32 Batch 30/173] avg loss 0.000139391, throughput 6.13472K wps
[Epoch 32 Batch 60/173] avg loss 0.000189001, throughput 5.99664K wps
[Epoch 32 Batch 90/173] avg loss 0.000178663, throughput 6.00909K wps
[Epoch 32 Batch 120/173] avg loss 0.000154506, throughput 6.00304K wps
[Epoch 32 Batch 150/173] avg loss 0.000209269, throughput 6.00181K wps
Begin Testing...
[Epoch 32] train avg loss 0.000177956, test acc 0.7740, test avg loss 0.770764, throughput 6.02741K wps
[Epoch 33 Batch 30/173] avg loss 0.000201704, throughput 6.15694K wps
[Epoch 33 Batch 60/173] avg loss 0.000136095, throughput 5.99028K wps
[Epoch 33 Batch 90/173] avg loss 0.000142873, throughput 5.98875K wps
[Epoch 33 Batch 120/173] avg loss 0.000134121, throughput 6.00567K wps
[Epoch 33 Batch 150/173] avg loss 0.000160754, throughput 6.00486K wps
Begin Testing...
[Epoch 33] train avg loss 0.000155174, test acc 0.7792, test avg loss 0.786095, throughput 6.02424K wps
[Epoch 34 Batch 30/173] avg loss 0.000145092, throughput 6.14701K wps
[Epoch 34 Batch 60/173] avg loss 0.00013044, throughput 6.00814K wps
[Epoch 34 Batch 90/173] avg loss 0.000154902, throughput 5.99905K wps
[Epoch 34 Batch 120/173] avg loss 0.000108595, throughput 6.00324K wps
[Epoch 34 Batch 150/173] avg loss 0.000153172, throughput 5.99018K wps
Begin Testing...
[Epoch 34] train avg loss 0.00013846, test acc 0.7760, test avg loss 0.804617, throughput 6.02612K wps
[Epoch 35 Batch 30/173] avg loss 0.000119199, throughput 6.14288K wps
[Epoch 35 Batch 60/173] avg loss 0.000132306, throughput 5.98616K wps
[Epoch 35 Batch 90/173] avg loss 0.000107658, throughput 5.98787K wps
[Epoch 35 Batch 120/173] avg loss 0.000101996, throughput 5.99614K wps
[Epoch 35 Batch 150/173] avg loss 0.000109279, throughput 6.00503K wps
Begin Testing...
[Epoch 35] train avg loss 0.000114956, test acc 0.7771, test avg loss 0.828904, throughput 6.02267K wps
[Epoch 36 Batch 30/173] avg loss 0.000129106, throughput 6.16108K wps
[Epoch 36 Batch 60/173] avg loss 0.000111309, throughput 5.99766K wps
[Epoch 36 Batch 90/173] avg loss 9.74761e-05, throughput 5.99956K wps
[Epoch 36 Batch 120/173] avg loss 0.000136829, throughput 5.99449K wps
[Epoch 36 Batch 150/173] avg loss 0.000117146, throughput 5.98588K wps
Begin Testing...
[Epoch 36] train avg loss 0.000119173, test acc 0.7750, test avg loss 0.849141, throughput 6.02521K wps
[Epoch 37 Batch 30/173] avg loss 9.96762e-05, throughput 6.14921K wps
[Epoch 37 Batch 60/173] avg loss 8.05194e-05, throughput 5.99798K wps
[Epoch 37 Batch 90/173] avg loss 8.27619e-05, throughput 5.99378K wps
[Epoch 37 Batch 120/173] avg loss 9.32823e-05, throughput 5.99756K wps
[Epoch 37 Batch 150/173] avg loss 0.000105468, throughput 5.99547K wps
Begin Testing...
[Epoch 37] train avg loss 9.22321e-05, test acc 0.7750, test avg loss 0.866972, throughput 6.02444K wps
[Epoch 38 Batch 30/173] avg loss 8.27235e-05, throughput 6.13363K wps
[Epoch 38 Batch 60/173] avg loss 8.75627e-05, throughput 5.99297K wps
[Epoch 38 Batch 90/173] avg loss 7.73934e-05, throughput 6.00451K wps
[Epoch 38 Batch 120/173] avg loss 6.54011e-05, throughput 6.01291K wps
[Epoch 38 Batch 150/173] avg loss 8.82998e-05, throughput 6.00563K wps
Begin Testing...
[Epoch 38] train avg loss 7.8614e-05, test acc 0.7719, test avg loss 0.88825, throughput 6.02644K wps
[Epoch 39 Batch 30/173] avg loss 6.48263e-05, throughput 6.13453K wps
[Epoch 39 Batch 60/173] avg loss 7.58554e-05, throughput 6.00865K wps
[Epoch 39 Batch 90/173] avg loss 5.95301e-05, throughput 6.007K wps
[Epoch 39 Batch 120/173] avg loss 7.17561e-05, throughput 6.00838K wps
[Epoch 39 Batch 150/173] avg loss 7.16064e-05, throughput 5.99894K wps
Begin Testing...
[Epoch 39] train avg loss 6.84813e-05, test acc 0.7740, test avg loss 0.903326, throughput 6.02917K wps
[Epoch 40 Batch 30/173] avg loss 6.51135e-05, throughput 6.1471K wps
[Epoch 40 Batch 60/173] avg loss 6.49432e-05, throughput 6.00044K wps
[Epoch 40 Batch 90/173] avg loss 7.55112e-05, throughput 6.00805K wps
[Epoch 40 Batch 120/173] avg loss 5.93198e-05, throughput 5.99927K wps
[Epoch 40 Batch 150/173] avg loss 6.39048e-05, throughput 6.00042K wps
Begin Testing...
[Epoch 40] train avg loss 6.59486e-05, test acc 0.7771, test avg loss 0.922373, throughput 6.02533K wps
[Epoch 41 Batch 30/173] avg loss 5.64537e-05, throughput 6.14344K wps
[Epoch 41 Batch 60/173] avg loss 6.91602e-05, throughput 6.00747K wps
[Epoch 41 Batch 90/173] avg loss 5.51889e-05, throughput 5.99507K wps
[Epoch 41 Batch 120/173] avg loss 6.21222e-05, throughput 6.00484K wps
[Epoch 41 Batch 150/173] avg loss 5.64632e-05, throughput 5.99582K wps
Begin Testing...
[Epoch 41] train avg loss 5.93405e-05, test acc 0.7729, test avg loss 0.949168, throughput 6.02572K wps
[Epoch 42 Batch 30/173] avg loss 5.54065e-05, throughput 6.13591K wps
[Epoch 42 Batch 60/173] avg loss 4.55822e-05, throughput 5.98482K wps
[Epoch 42 Batch 90/173] avg loss 6.02914e-05, throughput 5.9946K wps
[Epoch 42 Batch 120/173] avg loss 4.51836e-05, throughput 6.00841K wps
[Epoch 42 Batch 150/173] avg loss 5.89177e-05, throughput 6.00574K wps
Begin Testing...
[Epoch 42] train avg loss 5.22847e-05, test acc 0.7719, test avg loss 0.967166, throughput 6.02195K wps
[Epoch 43 Batch 30/173] avg loss 3.73712e-05, throughput 6.13948K wps
[Epoch 43 Batch 60/173] avg loss 4.45865e-05, throughput 5.98353K wps
[Epoch 43 Batch 90/173] avg loss 4.58517e-05, throughput 5.99815K wps
[Epoch 43 Batch 120/173] avg loss 3.51622e-05, throughput 5.98922K wps
[Epoch 43 Batch 150/173] avg loss 5.35408e-05, throughput 5.99938K wps
Begin Testing...
[Epoch 43] train avg loss 4.366e-05, test acc 0.7740, test avg loss 0.988551, throughput 6.01886K wps
[Epoch 44 Batch 30/173] avg loss 5.5306e-05, throughput 6.13794K wps
[Epoch 44 Batch 60/173] avg loss 5.3466e-05, throughput 6.00145K wps
[Epoch 44 Batch 90/173] avg loss 5.2449e-05, throughput 5.99704K wps
[Epoch 44 Batch 120/173] avg loss 3.92045e-05, throughput 6.00421K wps
[Epoch 44 Batch 150/173] avg loss 4.33957e-05, throughput 5.99015K wps
Begin Testing...
[Epoch 44] train avg loss 4.92237e-05, test acc 0.7708, test avg loss 1.0006, throughput 6.02326K wps
[Epoch 45 Batch 30/173] avg loss 4.28865e-05, throughput 6.14947K wps
[Epoch 45 Batch 60/173] avg loss 3.80935e-05, throughput 5.98703K wps
[Epoch 45 Batch 90/173] avg loss 3.42299e-05, throughput 6.01171K wps
[Epoch 45 Batch 120/173] avg loss 3.04071e-05, throughput 5.98868K wps
[Epoch 45 Batch 150/173] avg loss 3.32122e-05, throughput 5.99611K wps
Begin Testing...
[Epoch 45] train avg loss 3.58624e-05, test acc 0.7656, test avg loss 1.02433, throughput 6.02183K wps
[Epoch 46 Batch 30/173] avg loss 3.58294e-05, throughput 6.14094K wps
[Epoch 46 Batch 60/173] avg loss 3.36243e-05, throughput 5.99481K wps
[Epoch 46 Batch 90/173] avg loss 3.65621e-05, throughput 5.98575K wps
[Epoch 46 Batch 120/173] avg loss 3.44332e-05, throughput 5.99943K wps
[Epoch 46 Batch 150/173] avg loss 3.37708e-05, throughput 5.98767K wps
Begin Testing...
[Epoch 46] train avg loss 3.54192e-05, test acc 0.7688, test avg loss 1.04012, throughput 6.01694K wps
[Epoch 47 Batch 30/173] avg loss 3.12786e-05, throughput 6.14074K wps
[Epoch 47 Batch 60/173] avg loss 2.71497e-05, throughput 6.00848K wps
[Epoch 47 Batch 90/173] avg loss 2.95806e-05, throughput 6.0065K wps
[Epoch 47 Batch 120/173] avg loss 2.80808e-05, throughput 5.98397K wps
[Epoch 47 Batch 150/173] avg loss 3.13342e-05, throughput 5.98751K wps
Begin Testing...
[Epoch 47] train avg loss 3.16079e-05, test acc 0.7729, test avg loss 1.06033, throughput 6.02108K wps
[Epoch 48 Batch 30/173] avg loss 2.90539e-05, throughput 6.14001K wps
[Epoch 48 Batch 60/173] avg loss 3.87556e-05, throughput 6.00198K wps
[Epoch 48 Batch 90/173] avg loss 3.85765e-05, throughput 6.00338K wps
[Epoch 48 Batch 120/173] avg loss 3.55583e-05, throughput 5.98998K wps
[Epoch 48 Batch 150/173] avg loss 3.86196e-05, throughput 5.99446K wps
Begin Testing...
[Epoch 48] train avg loss 3.49042e-05, test acc 0.7708, test avg loss 1.07777, throughput 6.02476K wps
[Epoch 49 Batch 30/173] avg loss 2.3997e-05, throughput 6.15442K wps
[Epoch 49 Batch 60/173] avg loss 2.82337e-05, throughput 5.99977K wps
[Epoch 49 Batch 90/173] avg loss 3.83549e-05, throughput 5.99896K wps
[Epoch 49 Batch 120/173] avg loss 2.55589e-05, throughput 5.98906K wps
[Epoch 49 Batch 150/173] avg loss 2.68551e-05, throughput 5.98529K wps
Begin Testing...
[Epoch 49] train avg loss 2.82222e-05, test acc 0.7740, test avg loss 1.09705, throughput 6.02026K wps
[Epoch 50 Batch 30/173] avg loss 2.94204e-05, throughput 6.15489K wps
[Epoch 50 Batch 60/173] avg loss 2.15538e-05, throughput 5.98641K wps
[Epoch 50 Batch 90/173] avg loss 2.47211e-05, throughput 6.00323K wps
[Epoch 50 Batch 120/173] avg loss 2.3099e-05, throughput 5.98218K wps
[Epoch 50 Batch 150/173] avg loss 2.01792e-05, throughput 6.00037K wps
Begin Testing...
[Epoch 50] train avg loss 2.37294e-05, test acc 0.7698, test avg loss 1.12263, throughput 6.02256K wps
[Epoch 51 Batch 30/173] avg loss 1.60947e-05, throughput 6.15668K wps
[Epoch 51 Batch 60/173] avg loss 2.18558e-05, throughput 5.99501K wps
[Epoch 51 Batch 90/173] avg loss 1.60582e-05, throughput 5.99898K wps
[Epoch 51 Batch 120/173] avg loss 1.8569e-05, throughput 5.9997K wps
[Epoch 51 Batch 150/173] avg loss 1.95334e-05, throughput 5.9988K wps
Begin Testing...
[Epoch 51] train avg loss 1.90578e-05, test acc 0.7698, test avg loss 1.14222, throughput 6.02599K wps
[Epoch 52 Batch 30/173] avg loss 2.40987e-05, throughput 6.12731K wps
[Epoch 52 Batch 60/173] avg loss 1.75985e-05, throughput 5.98996K wps
[Epoch 52 Batch 90/173] avg loss 1.95568e-05, throughput 6.00908K wps
[Epoch 52 Batch 120/173] avg loss 2.82866e-05, throughput 5.99744K wps
[Epoch 52 Batch 150/173] avg loss 1.49524e-05, throughput 6.00857K wps
Begin Testing...
[Epoch 52] train avg loss 2.01333e-05, test acc 0.7656, test avg loss 1.16067, throughput 6.02274K wps
[Epoch 53 Batch 30/173] avg loss 2.1346e-05, throughput 6.13846K wps
[Epoch 53 Batch 60/173] avg loss 1.6996e-05, throughput 5.98794K wps
[Epoch 53 Batch 90/173] avg loss 1.25868e-05, throughput 5.97268K wps
[Epoch 53 Batch 120/173] avg loss 1.60472e-05, throughput 5.98059K wps
[Epoch 53 Batch 150/173] avg loss 2.31528e-05, throughput 6.00337K wps
Begin Testing...
[Epoch 53] train avg loss 1.75362e-05, test acc 0.7667, test avg loss 1.17962, throughput 6.01397K wps
[Epoch 54 Batch 30/173] avg loss 1.38121e-05, throughput 6.14463K wps
[Epoch 54 Batch 60/173] avg loss 1.69588e-05, throughput 5.99922K wps
[Epoch 54 Batch 90/173] avg loss 1.44549e-05, throughput 5.98638K wps
[Epoch 54 Batch 120/173] avg loss 1.76762e-05, throughput 5.98614K wps
[Epoch 54 Batch 150/173] avg loss 3.6402e-05, throughput 6.0068K wps
Begin Testing...
[Epoch 54] train avg loss 2.00683e-05, test acc 0.7667, test avg loss 1.20157, throughput 6.02123K wps
[Epoch 55 Batch 30/173] avg loss 1.2765e-05, throughput 6.14532K wps
[Epoch 55 Batch 60/173] avg loss 1.30005e-05, throughput 5.99059K wps
[Epoch 55 Batch 90/173] avg loss 1.54631e-05, throughput 6.0021K wps
[Epoch 55 Batch 120/173] avg loss 1.73585e-05, throughput 6.00375K wps
[Epoch 55 Batch 150/173] avg loss 1.61115e-05, throughput 5.98132K wps
Begin Testing...
[Epoch 55] train avg loss 1.48052e-05, test acc 0.7708, test avg loss 1.20158, throughput 6.02036K wps
[Epoch 56 Batch 30/173] avg loss 1.15938e-05, throughput 6.12865K wps
[Epoch 56 Batch 60/173] avg loss 1.74811e-05, throughput 5.99593K wps
[Epoch 56 Batch 90/173] avg loss 1.23871e-05, throughput 5.99904K wps
[Epoch 56 Batch 120/173] avg loss 1.8414e-05, throughput 5.99525K wps
[Epoch 56 Batch 150/173] avg loss 1.10154e-05, throughput 6.003K wps
Begin Testing...
[Epoch 56] train avg loss 1.72747e-05, test acc 0.7677, test avg loss 1.23346, throughput 6.02063K wps
[Epoch 57 Batch 30/173] avg loss 1.51904e-05, throughput 6.13996K wps
[Epoch 57 Batch 60/173] avg loss 1.66298e-05, throughput 5.99606K wps
[Epoch 57 Batch 90/173] avg loss 1.0598e-05, throughput 5.99777K wps
[Epoch 57 Batch 120/173] avg loss 1.33194e-05, throughput 6.00206K wps
[Epoch 57 Batch 150/173] avg loss 2.23003e-05, throughput 6.00002K wps
Begin Testing...
[Epoch 57] train avg loss 1.69702e-05, test acc 0.7656, test avg loss 1.23601, throughput 6.02368K wps
[Epoch 58 Batch 30/173] avg loss 1.51304e-05, throughput 6.13551K wps
[Epoch 58 Batch 60/173] avg loss 1.35974e-05, throughput 5.99507K wps
[Epoch 58 Batch 90/173] avg loss 1.74561e-05, throughput 5.99765K wps
[Epoch 58 Batch 120/173] avg loss 1.43266e-05, throughput 5.99878K wps
[Epoch 58 Batch 150/173] avg loss 9.74073e-06, throughput 6.0016K wps
Begin Testing...
[Epoch 58] train avg loss 1.39989e-05, test acc 0.7656, test avg loss 1.2609, throughput 6.0223K wps
[Epoch 59 Batch 30/173] avg loss 1.04777e-05, throughput 6.13972K wps
[Epoch 59 Batch 60/173] avg loss 1.4328e-05, throughput 6.00069K wps
[Epoch 59 Batch 90/173] avg loss 9.55036e-06, throughput 6.00284K wps
[Epoch 59 Batch 120/173] avg loss 1.29692e-05, throughput 5.99644K wps
[Epoch 59 Batch 150/173] avg loss 1.90273e-05, throughput 5.99411K wps
Begin Testing...
[Epoch 59] train avg loss 1.33765e-05, test acc 0.7615, test avg loss 1.27968, throughput 6.02244K wps
Test loss 0.423362, test acc 0.8114
Total time cost 358.98s
9596 1066
[Epoch 0 Batch 30/173] avg loss 0.0150579, throughput 5.7864K wps
[Epoch 0 Batch 60/173] avg loss 0.0149814, throughput 5.99137K wps
[Epoch 0 Batch 90/173] avg loss 0.0149812, throughput 5.99641K wps
[Epoch 0 Batch 120/173] avg loss 0.0142937, throughput 5.98947K wps
[Epoch 0 Batch 150/173] avg loss 0.0146167, throughput 6.00369K wps
Begin Testing...
[Epoch 0] train avg loss 0.0147254, test acc 0.5948, test avg loss 0.670547, throughput 5.9576K wps
Observed Improvement.
Begin Testing...
[Epoch 1 Batch 30/173] avg loss 0.0137541, throughput 6.14179K wps
[Epoch 1 Batch 60/173] avg loss 0.0136074, throughput 6.00335K wps
[Epoch 1 Batch 90/173] avg loss 0.0133117, throughput 5.99209K wps
[Epoch 1 Batch 120/173] avg loss 0.0133094, throughput 5.98629K wps
[Epoch 1 Batch 150/173] avg loss 0.0131097, throughput 5.99055K wps
Begin Testing...
[Epoch 1] train avg loss 0.0134267, test acc 0.6240, test avg loss 0.650727, throughput 6.01812K wps
Observed Improvement.
Begin Testing...
[Epoch 2 Batch 30/173] avg loss 0.0129041, throughput 6.16117K wps
[Epoch 2 Batch 60/173] avg loss 0.0125469, throughput 5.99519K wps
[Epoch 2 Batch 90/173] avg loss 0.0125569, throughput 5.98365K wps
[Epoch 2 Batch 120/173] avg loss 0.0125693, throughput 5.98659K wps
[Epoch 2 Batch 150/173] avg loss 0.0124013, throughput 5.98022K wps
Begin Testing...
[Epoch 2] train avg loss 0.0125811, test acc 0.6552, test avg loss 0.632994, throughput 6.01727K wps
Observed Improvement.
Begin Testing...
[Epoch 3 Batch 30/173] avg loss 0.0118815, throughput 6.16046K wps
[Epoch 3 Batch 60/173] avg loss 0.0117304, throughput 5.99728K wps
[Epoch 3 Batch 90/173] avg loss 0.0117149, throughput 6.00394K wps
[Epoch 3 Batch 120/173] avg loss 0.0117134, throughput 5.97249K wps
[Epoch 3 Batch 150/173] avg loss 0.0116201, throughput 5.98425K wps
Begin Testing...
[Epoch 3] train avg loss 0.0116848, test acc 0.6823, test avg loss 0.607226, throughput 6.02035K wps
Observed Improvement.
Begin Testing...
[Epoch 4 Batch 30/173] avg loss 0.0109588, throughput 6.1434K wps
[Epoch 4 Batch 60/173] avg loss 0.0107386, throughput 6.00415K wps
[Epoch 4 Batch 90/173] avg loss 0.0108179, throughput 5.99175K wps
[Epoch 4 Batch 120/173] avg loss 0.0106565, throughput 6.00201K wps
[Epoch 4 Batch 150/173] avg loss 0.0105014, throughput 5.99396K wps
Begin Testing...
[Epoch 4] train avg loss 0.010715, test acc 0.7344, test avg loss 0.571681, throughput 6.02064K wps
Observed Improvement.
Begin Testing...
[Epoch 5 Batch 30/173] avg loss 0.00980948, throughput 6.15852K wps
[Epoch 5 Batch 60/173] avg loss 0.00969736, throughput 5.97834K wps
[Epoch 5 Batch 90/173] avg loss 0.00976812, throughput 6.01558K wps
[Epoch 5 Batch 120/173] avg loss 0.00961278, throughput 6.00501K wps
[Epoch 5 Batch 150/173] avg loss 0.00941369, throughput 5.99803K wps
Begin Testing...
[Epoch 5] train avg loss 0.00963159, test acc 0.7615, test avg loss 0.537182, throughput 6.02705K wps
Observed Improvement.
Begin Testing...
[Epoch 6 Batch 30/173] avg loss 0.00877894, throughput 6.14133K wps
[Epoch 6 Batch 60/173] avg loss 0.00864027, throughput 6.00479K wps
[Epoch 6 Batch 90/173] avg loss 0.00853735, throughput 6.00549K wps
[Epoch 6 Batch 120/173] avg loss 0.00834066, throughput 6.01585K wps
[Epoch 6 Batch 150/173] avg loss 0.00830288, throughput 6.00503K wps
Begin Testing...
[Epoch 6] train avg loss 0.00850051, test acc 0.7802, test avg loss 0.510037, throughput 6.03232K wps
Observed Improvement.
Begin Testing...
[Epoch 7 Batch 30/173] avg loss 0.00774714, throughput 6.15885K wps
[Epoch 7 Batch 60/173] avg loss 0.00749889, throughput 6.01465K wps
[Epoch 7 Batch 90/173] avg loss 0.00764457, throughput 6.01479K wps
[Epoch 7 Batch 120/173] avg loss 0.00698109, throughput 6.00464K wps
[Epoch 7 Batch 150/173] avg loss 0.00761535, throughput 6.00633K wps
Begin Testing...
[Epoch 7] train avg loss 0.00747782, test acc 0.8063, test avg loss 0.481648, throughput 6.03538K wps
Observed Improvement.
Begin Testing...
[Epoch 8 Batch 30/173] avg loss 0.00684158, throughput 6.14465K wps
[Epoch 8 Batch 60/173] avg loss 0.0066966, throughput 5.99703K wps
[Epoch 8 Batch 90/173] avg loss 0.00649157, throughput 5.99805K wps
[Epoch 8 Batch 120/173] avg loss 0.00630905, throughput 5.99683K wps
[Epoch 8 Batch 150/173] avg loss 0.00645199, throughput 6.00118K wps
Begin Testing...
[Epoch 8] train avg loss 0.00654615, test acc 0.8042, test avg loss 0.458718, throughput 6.02369K wps
[Epoch 9 Batch 30/173] avg loss 0.00571046, throughput 6.15551K wps
[Epoch 9 Batch 60/173] avg loss 0.00586214, throughput 6.00324K wps
[Epoch 9 Batch 90/173] avg loss 0.00561143, throughput 5.99431K wps
[Epoch 9 Batch 120/173] avg loss 0.00576037, throughput 5.98534K wps
[Epoch 9 Batch 150/173] avg loss 0.00588594, throughput 5.99272K wps
Begin Testing...
[Epoch 9] train avg loss 0.00578535, test acc 0.8083, test avg loss 0.446859, throughput 6.02237K wps
Observed Improvement.
Begin Testing...
[Epoch 10 Batch 30/173] avg loss 0.00525967, throughput 6.13499K wps
[Epoch 10 Batch 60/173] avg loss 0.00496178, throughput 5.97672K wps
[Epoch 10 Batch 90/173] avg loss 0.00511248, throughput 5.99707K wps
[Epoch 10 Batch 120/173] avg loss 0.00480856, throughput 6.00963K wps
[Epoch 10 Batch 150/173] avg loss 0.00481594, throughput 5.99694K wps
Begin Testing...
[Epoch 10] train avg loss 0.00495467, test acc 0.7885, test avg loss 0.443832, throughput 6.01757K wps
[Epoch 11 Batch 30/173] avg loss 0.00420271, throughput 6.15001K wps
[Epoch 11 Batch 60/173] avg loss 0.00434954, throughput 6.00298K wps
[Epoch 11 Batch 90/173] avg loss 0.00420178, throughput 6.00219K wps
[Epoch 11 Batch 120/173] avg loss 0.00408409, throughput 6.00936K wps
[Epoch 11 Batch 150/173] avg loss 0.00434032, throughput 6.00343K wps
Begin Testing...
[Epoch 11] train avg loss 0.00427771, test acc 0.8135, test avg loss 0.439628, throughput 6.02864K wps
Observed Improvement.
Begin Testing...
[Epoch 12 Batch 30/173] avg loss 0.00361401, throughput 6.15728K wps
[Epoch 12 Batch 60/173] avg loss 0.00350866, throughput 6.00402K wps
[Epoch 12 Batch 90/173] avg loss 0.00368365, throughput 6.00126K wps
[Epoch 12 Batch 120/173] avg loss 0.00355655, throughput 5.99411K wps
[Epoch 12 Batch 150/173] avg loss 0.00392568, throughput 6.00716K wps
Begin Testing...
[Epoch 12] train avg loss 0.00364773, test acc 0.8083, test avg loss 0.440844, throughput 6.02908K wps
[Epoch 13 Batch 30/173] avg loss 0.0028576, throughput 6.14333K wps
[Epoch 13 Batch 60/173] avg loss 0.0030863, throughput 5.99536K wps
[Epoch 13 Batch 90/173] avg loss 0.00327135, throughput 6.00088K wps
[Epoch 13 Batch 120/173] avg loss 0.00302687, throughput 5.98593K wps
[Epoch 13 Batch 150/173] avg loss 0.00307869, throughput 5.99972K wps
Begin Testing...
[Epoch 13] train avg loss 0.0030926, test acc 0.7927, test avg loss 0.448757, throughput 6.02264K wps
[Epoch 14 Batch 30/173] avg loss 0.00273057, throughput 6.13822K wps
[Epoch 14 Batch 60/173] avg loss 0.00268075, throughput 5.98933K wps
[Epoch 14 Batch 90/173] avg loss 0.00275298, throughput 5.98612K wps
[Epoch 14 Batch 120/173] avg loss 0.00268638, throughput 5.99223K wps
[Epoch 14 Batch 150/173] avg loss 0.00257625, throughput 5.98524K wps
Begin Testing...
[Epoch 14] train avg loss 0.00269544, test acc 0.7948, test avg loss 0.458171, throughput 6.01515K wps
[Epoch 15 Batch 30/173] avg loss 0.00218164, throughput 6.14896K wps
[Epoch 15 Batch 60/173] avg loss 0.00225972, throughput 6.00669K wps
[Epoch 15 Batch 90/173] avg loss 0.00211544, throughput 6.00227K wps
[Epoch 15 Batch 120/173] avg loss 0.00219695, throughput 5.98062K wps
[Epoch 15 Batch 150/173] avg loss 0.00211103, throughput 5.99756K wps
Begin Testing...
[Epoch 15] train avg loss 0.00223945, test acc 0.7948, test avg loss 0.468623, throughput 6.02376K wps
[Epoch 16 Batch 30/173] avg loss 0.00211956, throughput 6.14726K wps
[Epoch 16 Batch 60/173] avg loss 0.00174818, throughput 5.99702K wps
[Epoch 16 Batch 90/173] avg loss 0.00197313, throughput 6.01729K wps
[Epoch 16 Batch 120/173] avg loss 0.00187452, throughput 5.9967K wps
[Epoch 16 Batch 150/173] avg loss 0.00188881, throughput 5.99296K wps
Begin Testing...
[Epoch 16] train avg loss 0.00193645, test acc 0.7979, test avg loss 0.479733, throughput 6.02478K wps
[Epoch 17 Batch 30/173] avg loss 0.00170465, throughput 6.15614K wps
[Epoch 17 Batch 60/173] avg loss 0.00154352, throughput 6.00759K wps
[Epoch 17 Batch 90/173] avg loss 0.00170148, throughput 6.00982K wps
[Epoch 17 Batch 120/173] avg loss 0.00167459, throughput 6.01261K wps
[Epoch 17 Batch 150/173] avg loss 0.00153869, throughput 6.01108K wps
Begin Testing...
[Epoch 17] train avg loss 0.00164006, test acc 0.7896, test avg loss 0.490793, throughput 6.0334K wps
[Epoch 18 Batch 30/173] avg loss 0.00124826, throughput 6.13964K wps
[Epoch 18 Batch 60/173] avg loss 0.00141292, throughput 5.99514K wps
[Epoch 18 Batch 90/173] avg loss 0.00137814, throughput 5.98446K wps
[Epoch 18 Batch 120/173] avg loss 0.00136704, throughput 6.00408K wps
[Epoch 18 Batch 150/173] avg loss 0.00144401, throughput 5.98915K wps
Begin Testing...
[Epoch 18] train avg loss 0.00136354, test acc 0.7937, test avg loss 0.507296, throughput 6.01891K wps
[Epoch 19 Batch 30/173] avg loss 0.00126162, throughput 6.13814K wps
[Epoch 19 Batch 60/173] avg loss 0.00112373, throughput 5.99513K wps
[Epoch 19 Batch 90/173] avg loss 0.00111123, throughput 6.00019K wps
[Epoch 19 Batch 120/173] avg loss 0.00112323, throughput 5.99942K wps
[Epoch 19 Batch 150/173] avg loss 0.001147, throughput 6.00152K wps
Begin Testing...
[Epoch 19] train avg loss 0.00116806, test acc 0.7854, test avg loss 0.533093, throughput 6.02405K wps
[Epoch 20 Batch 30/173] avg loss 0.00103603, throughput 6.1554K wps
[Epoch 20 Batch 60/173] avg loss 0.00101429, throughput 5.99956K wps
[Epoch 20 Batch 90/173] avg loss 0.000982855, throughput 6.00943K wps
[Epoch 20 Batch 120/173] avg loss 0.00109029, throughput 5.99736K wps
[Epoch 20 Batch 150/173] avg loss 0.00102961, throughput 5.99608K wps
Begin Testing...
[Epoch 20] train avg loss 0.00102248, test acc 0.7844, test avg loss 0.547993, throughput 6.02572K wps
[Epoch 21 Batch 30/173] avg loss 0.000864184, throughput 6.14826K wps
[Epoch 21 Batch 60/173] avg loss 0.0008394, throughput 5.99189K wps
[Epoch 21 Batch 90/173] avg loss 0.000853243, throughput 6.01103K wps
[Epoch 21 Batch 120/173] avg loss 0.000876986, throughput 5.99791K wps
[Epoch 21 Batch 150/173] avg loss 0.000846855, throughput 5.99484K wps
Begin Testing...
[Epoch 21] train avg loss 0.000862934, test acc 0.7854, test avg loss 0.55638, throughput 6.02472K wps
[Epoch 22 Batch 30/173] avg loss 0.000816606, throughput 6.14511K wps
[Epoch 22 Batch 60/173] avg loss 0.00069719, throughput 5.99768K wps
[Epoch 22 Batch 90/173] avg loss 0.000765764, throughput 5.99885K wps
[Epoch 22 Batch 120/173] avg loss 0.000705963, throughput 5.98892K wps
[Epoch 22 Batch 150/173] avg loss 0.000702838, throughput 5.99486K wps
Begin Testing...
[Epoch 22] train avg loss 0.00072789, test acc 0.7823, test avg loss 0.576576, throughput 6.02218K wps
[Epoch 23 Batch 30/173] avg loss 0.000634512, throughput 6.13105K wps
[Epoch 23 Batch 60/173] avg loss 0.000665213, throughput 5.98583K wps
[Epoch 23 Batch 90/173] avg loss 0.000564733, throughput 5.96476K wps
[Epoch 23 Batch 120/173] avg loss 0.000569888, throughput 5.97407K wps
[Epoch 23 Batch 150/173] avg loss 0.000649226, throughput 5.98988K wps
Begin Testing...
[Epoch 23] train avg loss 0.000618801, test acc 0.7823, test avg loss 0.595572, throughput 6.00759K wps
[Epoch 24 Batch 30/173] avg loss 0.000547422, throughput 6.15027K wps
[Epoch 24 Batch 60/173] avg loss 0.000472362, throughput 5.99515K wps
[Epoch 24 Batch 90/173] avg loss 0.000586555, throughput 5.99095K wps
[Epoch 24 Batch 120/173] avg loss 0.000485386, throughput 5.99251K wps
[Epoch 24 Batch 150/173] avg loss 0.000559992, throughput 5.99141K wps
Begin Testing...
[Epoch 24] train avg loss 0.000530616, test acc 0.7812, test avg loss 0.607208, throughput 6.02075K wps
[Epoch 25 Batch 30/173] avg loss 0.000415101, throughput 6.14621K wps
[Epoch 25 Batch 60/173] avg loss 0.000434595, throughput 5.9864K wps
[Epoch 25 Batch 90/173] avg loss 0.000516154, throughput 5.98656K wps
[Epoch 25 Batch 120/173] avg loss 0.000481321, throughput 5.98929K wps
[Epoch 25 Batch 150/173] avg loss 0.000405948, throughput 5.99457K wps
Begin Testing...
[Epoch 25] train avg loss 0.000453001, test acc 0.7750, test avg loss 0.624017, throughput 6.01675K wps
[Epoch 26 Batch 30/173] avg loss 0.000349517, throughput 6.14643K wps
[Epoch 26 Batch 60/173] avg loss 0.000348748, throughput 5.99934K wps
[Epoch 26 Batch 90/173] avg loss 0.000372584, throughput 5.99598K wps
[Epoch 26 Batch 120/173] avg loss 0.000474922, throughput 5.99847K wps
[Epoch 26 Batch 150/173] avg loss 0.00039487, throughput 6.00259K wps
Begin Testing...
[Epoch 26] train avg loss 0.000392213, test acc 0.7802, test avg loss 0.646391, throughput 6.02417K wps
[Epoch 27 Batch 30/173] avg loss 0.000333621, throughput 6.1332K wps
[Epoch 27 Batch 60/173] avg loss 0.00039358, throughput 5.99015K wps
[Epoch 27 Batch 90/173] avg loss 0.000356638, throughput 5.99337K wps
[Epoch 27 Batch 120/173] avg loss 0.000377422, throughput 5.99819K wps
[Epoch 27 Batch 150/173] avg loss 0.00034602, throughput 6.00505K wps
Begin Testing...
[Epoch 27] train avg loss 0.000360316, test acc 0.7802, test avg loss 0.668504, throughput 6.01982K wps
[Epoch 28 Batch 30/173] avg loss 0.000315345, throughput 6.14396K wps
[Epoch 28 Batch 60/173] avg loss 0.000320589, throughput 5.99236K wps
[Epoch 28 Batch 90/173] avg loss 0.000321843, throughput 6.00456K wps
[Epoch 28 Batch 120/173] avg loss 0.000281583, throughput 6.00497K wps
[Epoch 28 Batch 150/173] avg loss 0.000300034, throughput 5.99621K wps
Begin Testing...
[Epoch 28] train avg loss 0.00031181, test acc 0.7771, test avg loss 0.679277, throughput 6.02481K wps
[Epoch 29 Batch 30/173] avg loss 0.000265364, throughput 6.13626K wps
[Epoch 29 Batch 60/173] avg loss 0.000215339, throughput 6.00579K wps
[Epoch 29 Batch 90/173] avg loss 0.000253526, throughput 5.99767K wps
[Epoch 29 Batch 120/173] avg loss 0.000216906, throughput 5.98477K wps
[Epoch 29 Batch 150/173] avg loss 0.000274736, throughput 5.99322K wps
Begin Testing...
[Epoch 29] train avg loss 0.000251157, test acc 0.7781, test avg loss 0.699119, throughput 6.0205K wps
[Epoch 30 Batch 30/173] avg loss 0.000206887, throughput 6.15038K wps
[Epoch 30 Batch 60/173] avg loss 0.000220955, throughput 5.99691K wps
[Epoch 30 Batch 90/173] avg loss 0.0002444, throughput 5.9996K wps
[Epoch 30 Batch 120/173] avg loss 0.000267063, throughput 5.99676K wps
[Epoch 30 Batch 150/173] avg loss 0.000226267, throughput 5.98291K wps
Begin Testing...
[Epoch 30] train avg loss 0.00023089, test acc 0.7729, test avg loss 0.720483, throughput 6.02092K wps
[Epoch 31 Batch 30/173] avg loss 0.000232384, throughput 6.15334K wps
[Epoch 31 Batch 60/173] avg loss 0.000189562, throughput 6.00387K wps
[Epoch 31 Batch 90/173] avg loss 0.000178279, throughput 5.99083K wps
[Epoch 31 Batch 120/173] avg loss 0.000221569, throughput 5.98828K wps
[Epoch 31 Batch 150/173] avg loss 0.000176382, throughput 5.99433K wps
Begin Testing...
[Epoch 31] train avg loss 0.000196868, test acc 0.7760, test avg loss 0.740467, throughput 6.02338K wps
[Epoch 32 Batch 30/173] avg loss 0.000153428, throughput 6.15501K wps
[Epoch 32 Batch 60/173] avg loss 0.00016901, throughput 5.99215K wps
[Epoch 32 Batch 90/173] avg loss 0.000163119, throughput 5.99019K wps
[Epoch 32 Batch 120/173] avg loss 0.000189205, throughput 5.98479K wps
[Epoch 32 Batch 150/173] avg loss 0.000166805, throughput 5.99111K wps
Begin Testing...
[Epoch 32] train avg loss 0.00017086, test acc 0.7740, test avg loss 0.76739, throughput 6.02021K wps
[Epoch 33 Batch 30/173] avg loss 0.0001319, throughput 6.14333K wps
[Epoch 33 Batch 60/173] avg loss 0.000163171, throughput 5.99322K wps
[Epoch 33 Batch 90/173] avg loss 0.000149056, throughput 6.00373K wps
[Epoch 33 Batch 120/173] avg loss 0.000142219, throughput 5.99973K wps
[Epoch 33 Batch 150/173] avg loss 0.000158867, throughput 6.00758K wps
Begin Testing...
[Epoch 33] train avg loss 0.000152172, test acc 0.7719, test avg loss 0.773191, throughput 6.0251K wps
[Epoch 34 Batch 30/173] avg loss 0.000115465, throughput 6.14907K wps
[Epoch 34 Batch 60/173] avg loss 0.000119139, throughput 5.99213K wps
[Epoch 34 Batch 90/173] avg loss 0.00013649, throughput 6.00443K wps
[Epoch 34 Batch 120/173] avg loss 0.000139629, throughput 6.00264K wps
[Epoch 34 Batch 150/173] avg loss 0.000134851, throughput 5.99406K wps
Begin Testing...
[Epoch 34] train avg loss 0.00012923, test acc 0.7708, test avg loss 0.788221, throughput 6.02479K wps
[Epoch 35 Batch 30/173] avg loss 0.00010959, throughput 6.15834K wps
[Epoch 35 Batch 60/173] avg loss 0.000116958, throughput 6.01353K wps
[Epoch 35 Batch 90/173] avg loss 0.000140023, throughput 6.00248K wps
[Epoch 35 Batch 120/173] avg loss 0.000112242, throughput 5.98845K wps
[Epoch 35 Batch 150/173] avg loss 0.000100259, throughput 5.99071K wps
Begin Testing...
[Epoch 35] train avg loss 0.000117894, test acc 0.7729, test avg loss 0.802967, throughput 6.02658K wps
[Epoch 36 Batch 30/173] avg loss 9.5387e-05, throughput 6.15077K wps
[Epoch 36 Batch 60/173] avg loss 9.26215e-05, throughput 5.99751K wps
[Epoch 36 Batch 90/173] avg loss 9.33792e-05, throughput 5.9906K wps
[Epoch 36 Batch 120/173] avg loss 0.000109681, throughput 5.99351K wps
[Epoch 36 Batch 150/173] avg loss 0.000104903, throughput 5.98776K wps
Begin Testing...
[Epoch 36] train avg loss 9.87095e-05, test acc 0.7708, test avg loss 0.825734, throughput 6.02162K wps
[Epoch 37 Batch 30/173] avg loss 8.04468e-05, throughput 6.15663K wps
[Epoch 37 Batch 60/173] avg loss 8.38545e-05, throughput 6.00995K wps
[Epoch 37 Batch 90/173] avg loss 9.69498e-05, throughput 5.99977K wps
[Epoch 37 Batch 120/173] avg loss 7.64889e-05, throughput 5.99717K wps
[Epoch 37 Batch 150/173] avg loss 8.4258e-05, throughput 5.99348K wps
Begin Testing...
[Epoch 37] train avg loss 8.6986e-05, test acc 0.7708, test avg loss 0.846129, throughput 6.02565K wps
[Epoch 38 Batch 30/173] avg loss 8.9993e-05, throughput 6.13485K wps
[Epoch 38 Batch 60/173] avg loss 7.957e-05, throughput 6.00533K wps
[Epoch 38 Batch 90/173] avg loss 7.49339e-05, throughput 6.0033K wps
[Epoch 38 Batch 120/173] avg loss 8.51524e-05, throughput 5.9943K wps
[Epoch 38 Batch 150/173] avg loss 8.78358e-05, throughput 6.00292K wps
Begin Testing...
[Epoch 38] train avg loss 8.31488e-05, test acc 0.7677, test avg loss 0.862617, throughput 6.02295K wps
[Epoch 39 Batch 30/173] avg loss 6.84844e-05, throughput 6.13429K wps
[Epoch 39 Batch 60/173] avg loss 9.54667e-05, throughput 5.99748K wps
[Epoch 39 Batch 90/173] avg loss 6.62785e-05, throughput 5.99316K wps
[Epoch 39 Batch 120/173] avg loss 6.45086e-05, throughput 5.99954K wps
[Epoch 39 Batch 150/173] avg loss 6.40997e-05, throughput 5.99933K wps
Begin Testing...
[Epoch 39] train avg loss 7.3126e-05, test acc 0.7677, test avg loss 0.887488, throughput 6.02084K wps
[Epoch 40 Batch 30/173] avg loss 6.5201e-05, throughput 6.1234K wps
[Epoch 40 Batch 60/173] avg loss 5.85698e-05, throughput 5.98022K wps
[Epoch 40 Batch 90/173] avg loss 6.50525e-05, throughput 5.98225K wps
[Epoch 40 Batch 120/173] avg loss 8.84722e-05, throughput 5.99096K wps
[Epoch 40 Batch 150/173] avg loss 5.66231e-05, throughput 5.9914K wps
Begin Testing...
[Epoch 40] train avg loss 6.9522e-05, test acc 0.7688, test avg loss 0.911678, throughput 6.01193K wps
[Epoch 41 Batch 30/173] avg loss 5.29031e-05, throughput 6.14646K wps
[Epoch 41 Batch 60/173] avg loss 5.9136e-05, throughput 5.99176K wps
[Epoch 41 Batch 90/173] avg loss 4.88733e-05, throughput 5.99597K wps
[Epoch 41 Batch 120/173] avg loss 5.53615e-05, throughput 5.99176K wps
[Epoch 41 Batch 150/173] avg loss 6.50143e-05, throughput 5.99908K wps
Begin Testing...
[Epoch 41] train avg loss 6.17605e-05, test acc 0.7677, test avg loss 0.92689, throughput 6.02131K wps
[Epoch 42 Batch 30/173] avg loss 5.1538e-05, throughput 6.15552K wps
[Epoch 42 Batch 60/173] avg loss 4.7985e-05, throughput 5.99143K wps
[Epoch 42 Batch 90/173] avg loss 4.87636e-05, throughput 5.9899K wps
[Epoch 42 Batch 120/173] avg loss 5.6609e-05, throughput 5.98106K wps
[Epoch 42 Batch 150/173] avg loss 5.30638e-05, throughput 5.99877K wps
Begin Testing...
[Epoch 42] train avg loss 5.15293e-05, test acc 0.7646, test avg loss 0.942455, throughput 6.02071K wps
[Epoch 43 Batch 30/173] avg loss 3.35511e-05, throughput 6.15158K wps
[Epoch 43 Batch 60/173] avg loss 5.14134e-05, throughput 6.00776K wps
[Epoch 43 Batch 90/173] avg loss 5.96828e-05, throughput 6.00469K wps
[Epoch 43 Batch 120/173] avg loss 4.27907e-05, throughput 5.9986K wps
[Epoch 43 Batch 150/173] avg loss 4.22612e-05, throughput 5.99197K wps
Begin Testing...
[Epoch 43] train avg loss 4.60521e-05, test acc 0.7646, test avg loss 0.952462, throughput 6.02525K wps
[Epoch 44 Batch 30/173] avg loss 5.05674e-05, throughput 6.13928K wps
[Epoch 44 Batch 60/173] avg loss 4.12347e-05, throughput 5.99557K wps
[Epoch 44 Batch 90/173] avg loss 4.27844e-05, throughput 5.99372K wps
[Epoch 44 Batch 120/173] avg loss 3.54433e-05, throughput 5.99266K wps
[Epoch 44 Batch 150/173] avg loss 4.58845e-05, throughput 5.98814K wps
Begin Testing...
[Epoch 44] train avg loss 4.15864e-05, test acc 0.7615, test avg loss 0.970857, throughput 6.02006K wps
[Epoch 45 Batch 30/173] avg loss 2.95334e-05, throughput 6.14414K wps
[Epoch 45 Batch 60/173] avg loss 3.87215e-05, throughput 5.99929K wps
[Epoch 45 Batch 90/173] avg loss 3.45245e-05, throughput 5.99032K wps
[Epoch 45 Batch 120/173] avg loss 4.52517e-05, throughput 5.9822K wps
[Epoch 45 Batch 150/173] avg loss 3.51491e-05, throughput 5.99554K wps
Begin Testing...
[Epoch 45] train avg loss 3.64274e-05, test acc 0.7625, test avg loss 1.00053, throughput 6.01927K wps
[Epoch 46 Batch 30/173] avg loss 3.93814e-05, throughput 6.12627K wps
[Epoch 46 Batch 60/173] avg loss 3.52209e-05, throughput 5.98928K wps
[Epoch 46 Batch 90/173] avg loss 3.11886e-05, throughput 5.98565K wps
[Epoch 46 Batch 120/173] avg loss 3.50888e-05, throughput 5.99696K wps
[Epoch 46 Batch 150/173] avg loss 3.22111e-05, throughput 5.99333K wps
Begin Testing...
[Epoch 46] train avg loss 3.40226e-05, test acc 0.7635, test avg loss 1.01527, throughput 6.01445K wps
[Epoch 47 Batch 30/173] avg loss 3.72629e-05, throughput 6.14259K wps
[Epoch 47 Batch 60/173] avg loss 4.90414e-05, throughput 6.0216K wps
[Epoch 47 Batch 90/173] avg loss 3.31006e-05, throughput 6.01218K wps
[Epoch 47 Batch 120/173] avg loss 3.32845e-05, throughput 6.00407K wps
[Epoch 47 Batch 150/173] avg loss 3.84455e-05, throughput 6.01353K wps
Begin Testing...
[Epoch 47] train avg loss 3.73046e-05, test acc 0.7583, test avg loss 1.03024, throughput 6.03322K wps
[Epoch 48 Batch 30/173] avg loss 2.90113e-05, throughput 6.14578K wps
[Epoch 48 Batch 60/173] avg loss 2.44965e-05, throughput 6.00573K wps
[Epoch 48 Batch 90/173] avg loss 2.99649e-05, throughput 5.99766K wps
[Epoch 48 Batch 120/173] avg loss 2.96413e-05, throughput 5.9951K wps
[Epoch 48 Batch 150/173] avg loss 2.73222e-05, throughput 5.99674K wps
Begin Testing...
[Epoch 48] train avg loss 2.72881e-05, test acc 0.7594, test avg loss 1.04173, throughput 6.02589K wps
[Epoch 49 Batch 30/173] avg loss 1.9001e-05, throughput 6.15209K wps
[Epoch 49 Batch 60/173] avg loss 3.28083e-05, throughput 5.99498K wps
[Epoch 49 Batch 90/173] avg loss 2.82882e-05, throughput 6.00095K wps
[Epoch 49 Batch 120/173] avg loss 2.97296e-05, throughput 6.00717K wps
[Epoch 49 Batch 150/173] avg loss 2.27373e-05, throughput 6.00984K wps
Begin Testing...
[Epoch 49] train avg loss 2.62002e-05, test acc 0.7635, test avg loss 1.07598, throughput 6.0292K wps
[Epoch 50 Batch 30/173] avg loss 2.31087e-05, throughput 6.14495K wps
[Epoch 50 Batch 60/173] avg loss 2.56454e-05, throughput 6.01124K wps
[Epoch 50 Batch 90/173] avg loss 2.03025e-05, throughput 5.99888K wps
[Epoch 50 Batch 120/173] avg loss 2.16864e-05, throughput 5.98091K wps
[Epoch 50 Batch 150/173] avg loss 2.25859e-05, throughput 5.99144K wps
Begin Testing...
[Epoch 50] train avg loss 2.26559e-05, test acc 0.7625, test avg loss 1.0902, throughput 6.02113K wps
[Epoch 51 Batch 30/173] avg loss 1.60458e-05, throughput 6.14698K wps
[Epoch 51 Batch 60/173] avg loss 1.93544e-05, throughput 6.00331K wps
[Epoch 51 Batch 90/173] avg loss 1.61028e-05, throughput 5.99863K wps
[Epoch 51 Batch 120/173] avg loss 1.92517e-05, throughput 5.9961K wps
[Epoch 51 Batch 150/173] avg loss 2.51354e-05, throughput 5.99618K wps
Begin Testing...
[Epoch 51] train avg loss 2.05614e-05, test acc 0.7635, test avg loss 1.0876, throughput 6.02364K wps
[Epoch 52 Batch 30/173] avg loss 2.07358e-05, throughput 6.14147K wps
[Epoch 52 Batch 60/173] avg loss 1.75161e-05, throughput 5.98426K wps
[Epoch 52 Batch 90/173] avg loss 2.24207e-05, throughput 5.99202K wps
[Epoch 52 Batch 120/173] avg loss 2.42257e-05, throughput 6.00434K wps
[Epoch 52 Batch 150/173] avg loss 1.72605e-05, throughput 6.0033K wps
Begin Testing...
[Epoch 52] train avg loss 1.97779e-05, test acc 0.7604, test avg loss 1.11784, throughput 6.02328K wps
[Epoch 53 Batch 30/173] avg loss 2.09493e-05, throughput 6.15275K wps
[Epoch 53 Batch 60/173] avg loss 1.83641e-05, throughput 5.9945K wps
[Epoch 53 Batch 90/173] avg loss 1.50021e-05, throughput 5.99371K wps
[Epoch 53 Batch 120/173] avg loss 1.92227e-05, throughput 5.94817K wps
[Epoch 53 Batch 150/173] avg loss 1.80866e-05, throughput 5.96896K wps
Begin Testing...
[Epoch 53] train avg loss 1.83192e-05, test acc 0.7583, test avg loss 1.12647, throughput 6.01105K wps
[Epoch 54 Batch 30/173] avg loss 1.97651e-05, throughput 6.15122K wps
[Epoch 54 Batch 60/173] avg loss 3.01481e-05, throughput 6.00767K wps
[Epoch 54 Batch 90/173] avg loss 1.76947e-05, throughput 6.00009K wps
[Epoch 54 Batch 120/173] avg loss 1.21645e-05, throughput 5.99831K wps
[Epoch 54 Batch 150/173] avg loss 2.83563e-05, throughput 6.00398K wps
Begin Testing...
[Epoch 54] train avg loss 2.1609e-05, test acc 0.7604, test avg loss 1.14654, throughput 6.02984K wps
[Epoch 55 Batch 30/173] avg loss 1.3342e-05, throughput 6.14917K wps
[Epoch 55 Batch 60/173] avg loss 1.42032e-05, throughput 6.00022K wps
[Epoch 55 Batch 90/173] avg loss 2.32494e-05, throughput 6.00525K wps
[Epoch 55 Batch 120/173] avg loss 1.0576e-05, throughput 5.9943K wps
[Epoch 55 Batch 150/173] avg loss 1.2224e-05, throughput 6.00428K wps
Begin Testing...
[Epoch 55] train avg loss 1.4715e-05, test acc 0.7604, test avg loss 1.16194, throughput 6.02609K wps
[Epoch 56 Batch 30/173] avg loss 1.31894e-05, throughput 6.147K wps
[Epoch 56 Batch 60/173] avg loss 1.15427e-05, throughput 5.99729K wps
[Epoch 56 Batch 90/173] avg loss 1.00948e-05, throughput 5.98989K wps
[Epoch 56 Batch 120/173] avg loss 1.1045e-05, throughput 5.99003K wps
[Epoch 56 Batch 150/173] avg loss 1.39296e-05, throughput 5.99611K wps
Begin Testing...
[Epoch 56] train avg loss 1.22878e-05, test acc 0.7542, test avg loss 1.177, throughput 6.02113K wps
[Epoch 57 Batch 30/173] avg loss 9.97e-06, throughput 6.12955K wps
[Epoch 57 Batch 60/173] avg loss 3.1793e-05, throughput 5.98732K wps
[Epoch 57 Batch 90/173] avg loss 1.15764e-05, throughput 5.99269K wps
[Epoch 57 Batch 120/173] avg loss 1.2008e-05, throughput 6.00442K wps
[Epoch 57 Batch 150/173] avg loss 1.33996e-05, throughput 6.00233K wps
Begin Testing...
[Epoch 57] train avg loss 1.552e-05, test acc 0.7594, test avg loss 1.1922, throughput 6.01932K wps
[Epoch 58 Batch 30/173] avg loss 9.02993e-06, throughput 6.14329K wps
[Epoch 58 Batch 60/173] avg loss 9.13628e-06, throughput 6.00866K wps
[Epoch 58 Batch 90/173] avg loss 1.18706e-05, throughput 6.01826K wps
[Epoch 58 Batch 120/173] avg loss 1.32477e-05, throughput 6.01584K wps
[Epoch 58 Batch 150/173] avg loss 1.18262e-05, throughput 5.98717K wps
Begin Testing...
[Epoch 58] train avg loss 1.21475e-05, test acc 0.7594, test avg loss 1.20074, throughput 6.02881K wps
[Epoch 59 Batch 30/173] avg loss 8.30425e-06, throughput 6.15428K wps
[Epoch 59 Batch 60/173] avg loss 1.18138e-05, throughput 6.0046K wps
[Epoch 59 Batch 90/173] avg loss 1.44365e-05, throughput 6.00923K wps
[Epoch 59 Batch 120/173] avg loss 1.28228e-05, throughput 5.9866K wps
[Epoch 59 Batch 150/173] avg loss 9.23067e-06, throughput 5.99714K wps
Begin Testing...
[Epoch 59] train avg loss 1.13194e-05, test acc 0.7552, test avg loss 1.22628, throughput 6.02609K wps
Test loss 0.440559, test acc 0.7880
Total time cost 358.34s
9596 1066
[Epoch 0 Batch 30/173] avg loss 0.0156688, throughput 5.78975K wps
[Epoch 0 Batch 60/173] avg loss 0.0151272, throughput 6.00152K wps
[Epoch 0 Batch 90/173] avg loss 0.0150184, throughput 5.99237K wps
[Epoch 0 Batch 120/173] avg loss 0.0142408, throughput 5.99222K wps
[Epoch 0 Batch 150/173] avg loss 0.0144613, throughput 5.99197K wps
Begin Testing...
[Epoch 0] train avg loss 0.014826, test acc 0.5948, test avg loss 0.666499, throughput 5.95843K wps
Observed Improvement.
Begin Testing...
[Epoch 1 Batch 30/173] avg loss 0.0138151, throughput 6.15091K wps
[Epoch 1 Batch 60/173] avg loss 0.0138088, throughput 5.99242K wps
[Epoch 1 Batch 90/173] avg loss 0.0135555, throughput 6.00047K wps
[Epoch 1 Batch 120/173] avg loss 0.0134279, throughput 6.00179K wps
[Epoch 1 Batch 150/173] avg loss 0.0129633, throughput 6.01115K wps
Begin Testing...
[Epoch 1] train avg loss 0.0134726, test acc 0.6188, test avg loss 0.64861, throughput 6.02812K wps
Observed Improvement.
Begin Testing...
[Epoch 2 Batch 30/173] avg loss 0.0125935, throughput 6.14876K wps
[Epoch 2 Batch 60/173] avg loss 0.012721, throughput 6.00695K wps
[Epoch 2 Batch 90/173] avg loss 0.0128253, throughput 5.99775K wps
[Epoch 2 Batch 120/173] avg loss 0.0126811, throughput 5.98998K wps
[Epoch 2 Batch 150/173] avg loss 0.0125815, throughput 5.99982K wps
Begin Testing...
[Epoch 2] train avg loss 0.0126508, test acc 0.6323, test avg loss 0.630208, throughput 6.02662K wps
Observed Improvement.
Begin Testing...
[Epoch 3 Batch 30/173] avg loss 0.0121086, throughput 6.15876K wps
[Epoch 3 Batch 60/173] avg loss 0.0116297, throughput 6.01349K wps
[Epoch 3 Batch 90/173] avg loss 0.0118923, throughput 6.00411K wps
[Epoch 3 Batch 120/173] avg loss 0.0116386, throughput 6.01197K wps
[Epoch 3 Batch 150/173] avg loss 0.0116672, throughput 6.00777K wps
Begin Testing...
[Epoch 3] train avg loss 0.0117923, test acc 0.6948, test avg loss 0.600056, throughput 6.03636K wps
Observed Improvement.
Begin Testing...
[Epoch 4 Batch 30/173] avg loss 0.0110265, throughput 6.15308K wps
[Epoch 4 Batch 60/173] avg loss 0.0108627, throughput 6.00085K wps
[Epoch 4 Batch 90/173] avg loss 0.0107949, throughput 5.99498K wps
[Epoch 4 Batch 120/173] avg loss 0.0109161, throughput 6.00301K wps
[Epoch 4 Batch 150/173] avg loss 0.0105955, throughput 5.99752K wps
Begin Testing...
[Epoch 4] train avg loss 0.0108175, test acc 0.7469, test avg loss 0.566387, throughput 6.02596K wps
Observed Improvement.
Begin Testing...
[Epoch 5 Batch 30/173] avg loss 0.010052, throughput 6.15395K wps
[Epoch 5 Batch 60/173] avg loss 0.00994154, throughput 6.0012K wps
[Epoch 5 Batch 90/173] avg loss 0.00988416, throughput 6.00026K wps
[Epoch 5 Batch 120/173] avg loss 0.00972356, throughput 6.01341K wps
[Epoch 5 Batch 150/173] avg loss 0.00952699, throughput 6.00917K wps
Begin Testing...
[Epoch 5] train avg loss 0.00981504, test acc 0.7604, test avg loss 0.531868, throughput 6.03041K wps
Observed Improvement.
Begin Testing...
[Epoch 6 Batch 30/173] avg loss 0.00915765, throughput 6.14541K wps
[Epoch 6 Batch 60/173] avg loss 0.00869193, throughput 6.00474K wps
[Epoch 6 Batch 90/173] avg loss 0.00859567, throughput 6.01176K wps
[Epoch 6 Batch 120/173] avg loss 0.0086173, throughput 6.0077K wps
[Epoch 6 Batch 150/173] avg loss 0.00854164, throughput 6.00922K wps
Begin Testing...
[Epoch 6] train avg loss 0.0087253, test acc 0.7510, test avg loss 0.504461, throughput 6.02987K wps
[Epoch 7 Batch 30/173] avg loss 0.00763278, throughput 6.13821K wps
[Epoch 7 Batch 60/173] avg loss 0.00787646, throughput 6.00141K wps
[Epoch 7 Batch 90/173] avg loss 0.00777735, throughput 5.99865K wps
[Epoch 7 Batch 120/173] avg loss 0.00750333, throughput 6.01228K wps
[Epoch 7 Batch 150/173] avg loss 0.00741139, throughput 6.00758K wps
Begin Testing...
[Epoch 7] train avg loss 0.00764676, test acc 0.7750, test avg loss 0.476194, throughput 6.02716K wps
Observed Improvement.
Begin Testing...
[Epoch 8 Batch 30/173] avg loss 0.00675433, throughput 6.15896K wps
[Epoch 8 Batch 60/173] avg loss 0.00683738, throughput 6.01367K wps
[Epoch 8 Batch 90/173] avg loss 0.0064358, throughput 5.99372K wps
[Epoch 8 Batch 120/173] avg loss 0.0067688, throughput 6.00359K wps
[Epoch 8 Batch 150/173] avg loss 0.00636642, throughput 6.00046K wps
Begin Testing...
[Epoch 8] train avg loss 0.00662274, test acc 0.7865, test avg loss 0.457532, throughput 6.03065K wps
Observed Improvement.
Begin Testing...
[Epoch 9 Batch 30/173] avg loss 0.0058607, throughput 6.15167K wps
[Epoch 9 Batch 60/173] avg loss 0.00608821, throughput 6.00053K wps
[Epoch 9 Batch 90/173] avg loss 0.00581974, throughput 5.9926K wps
[Epoch 9 Batch 120/173] avg loss 0.00578645, throughput 6.01301K wps
[Epoch 9 Batch 150/173] avg loss 0.00556503, throughput 5.98909K wps
Begin Testing...
[Epoch 9] train avg loss 0.00580619, test acc 0.7958, test avg loss 0.445206, throughput 6.02455K wps
Observed Improvement.
Begin Testing...
[Epoch 10 Batch 30/173] avg loss 0.00489959, throughput 6.16348K wps
[Epoch 10 Batch 60/173] avg loss 0.00488901, throughput 6.01049K wps
[Epoch 10 Batch 90/173] avg loss 0.00504869, throughput 6.00281K wps
[Epoch 10 Batch 120/173] avg loss 0.00538608, throughput 5.99029K wps
[Epoch 10 Batch 150/173] avg loss 0.00469047, throughput 6.00869K wps
Begin Testing...
[Epoch 10] train avg loss 0.00496705, test acc 0.7979, test avg loss 0.440101, throughput 6.0301K wps
Observed Improvement.
Begin Testing...
[Epoch 11 Batch 30/173] avg loss 0.00442001, throughput 6.14102K wps
[Epoch 11 Batch 60/173] avg loss 0.00434167, throughput 5.9985K wps
[Epoch 11 Batch 90/173] avg loss 0.00449685, throughput 6.00428K wps
[Epoch 11 Batch 120/173] avg loss 0.00441164, throughput 6.01027K wps
[Epoch 11 Batch 150/173] avg loss 0.00432356, throughput 6.00205K wps
Begin Testing...
[Epoch 11] train avg loss 0.00442028, test acc 0.8021, test avg loss 0.436622, throughput 6.02874K wps
Observed Improvement.
Begin Testing...
[Epoch 12 Batch 30/173] avg loss 0.00379514, throughput 6.14945K wps
[Epoch 12 Batch 60/173] avg loss 0.00378829, throughput 5.99169K wps
[Epoch 12 Batch 90/173] avg loss 0.00380354, throughput 6.0051K wps
[Epoch 12 Batch 120/173] avg loss 0.00365978, throughput 5.9927K wps
[Epoch 12 Batch 150/173] avg loss 0.00368607, throughput 5.98429K wps
Begin Testing...
[Epoch 12] train avg loss 0.00374651, test acc 0.7969, test avg loss 0.43865, throughput 6.02068K wps
[Epoch 13 Batch 30/173] avg loss 0.00307615, throughput 6.13666K wps
[Epoch 13 Batch 60/173] avg loss 0.0031254, throughput 6.00202K wps
[Epoch 13 Batch 90/173] avg loss 0.00318101, throughput 6.00796K wps
[Epoch 13 Batch 120/173] avg loss 0.00331421, throughput 6.00137K wps
[Epoch 13 Batch 150/173] avg loss 0.00326383, throughput 6.01454K wps
Begin Testing...
[Epoch 13] train avg loss 0.00320048, test acc 0.8000, test avg loss 0.444323, throughput 6.02881K wps
[Epoch 14 Batch 30/173] avg loss 0.00269361, throughput 6.15557K wps
[Epoch 14 Batch 60/173] avg loss 0.00269985, throughput 6.01034K wps
[Epoch 14 Batch 90/173] avg loss 0.00270387, throughput 5.99684K wps
[Epoch 14 Batch 120/173] avg loss 0.00260867, throughput 6.00649K wps
[Epoch 14 Batch 150/173] avg loss 0.00268164, throughput 6.00292K wps
Begin Testing...
[Epoch 14] train avg loss 0.00268063, test acc 0.7990, test avg loss 0.459769, throughput 6.03178K wps
[Epoch 15 Batch 30/173] avg loss 0.00220694, throughput 6.15262K wps
[Epoch 15 Batch 60/173] avg loss 0.00223156, throughput 6.00478K wps
[Epoch 15 Batch 90/173] avg loss 0.00236472, throughput 6.00438K wps
[Epoch 15 Batch 120/173] avg loss 0.00234628, throughput 6.00741K wps
[Epoch 15 Batch 150/173] avg loss 0.00239698, throughput 6.00727K wps
Begin Testing...
[Epoch 15] train avg loss 0.00232439, test acc 0.7958, test avg loss 0.457169, throughput 6.03149K wps
[Epoch 16 Batch 30/173] avg loss 0.00187116, throughput 6.15174K wps
[Epoch 16 Batch 60/173] avg loss 0.00214823, throughput 6.00881K wps
[Epoch 16 Batch 90/173] avg loss 0.00195785, throughput 5.99041K wps
[Epoch 16 Batch 120/173] avg loss 0.00204304, throughput 5.99664K wps
[Epoch 16 Batch 150/173] avg loss 0.00207047, throughput 5.99704K wps
Begin Testing...
[Epoch 16] train avg loss 0.00200246, test acc 0.7990, test avg loss 0.466319, throughput 6.02469K wps
[Epoch 17 Batch 30/173] avg loss 0.00163897, throughput 6.14664K wps
[Epoch 17 Batch 60/173] avg loss 0.00176381, throughput 6.00762K wps
[Epoch 17 Batch 90/173] avg loss 0.00173246, throughput 5.99896K wps
[Epoch 17 Batch 120/173] avg loss 0.00166747, throughput 6.00366K wps
[Epoch 17 Batch 150/173] avg loss 0.00164086, throughput 6.00001K wps
Begin Testing...
[Epoch 17] train avg loss 0.00170319, test acc 0.7948, test avg loss 0.479293, throughput 6.0277K wps
[Epoch 18 Batch 30/173] avg loss 0.00136828, throughput 6.14892K wps
[Epoch 18 Batch 60/173] avg loss 0.00154262, throughput 5.99925K wps
[Epoch 18 Batch 90/173] avg loss 0.00152415, throughput 5.99509K wps
[Epoch 18 Batch 120/173] avg loss 0.00136929, throughput 5.99816K wps
[Epoch 18 Batch 150/173] avg loss 0.00126165, throughput 5.99773K wps
Begin Testing...
[Epoch 18] train avg loss 0.00141264, test acc 0.7917, test avg loss 0.49184, throughput 6.02421K wps
[Epoch 19 Batch 30/173] avg loss 0.00121293, throughput 6.14301K wps
[Epoch 19 Batch 60/173] avg loss 0.00124759, throughput 5.99496K wps
[Epoch 19 Batch 90/173] avg loss 0.00122472, throughput 6.00792K wps
[Epoch 19 Batch 120/173] avg loss 0.00119194, throughput 6.00146K wps
[Epoch 19 Batch 150/173] avg loss 0.00122391, throughput 5.99532K wps
Begin Testing...
[Epoch 19] train avg loss 0.00124573, test acc 0.7917, test avg loss 0.505086, throughput 6.02406K wps
[Epoch 20 Batch 30/173] avg loss 0.00105223, throughput 6.15002K wps
[Epoch 20 Batch 60/173] avg loss 0.00105593, throughput 6.0011K wps
[Epoch 20 Batch 90/173] avg loss 0.000982382, throughput 5.99547K wps
[Epoch 20 Batch 120/173] avg loss 0.00107456, throughput 5.98641K wps
[Epoch 20 Batch 150/173] avg loss 0.00095852, throughput 5.98971K wps
Begin Testing...
[Epoch 20] train avg loss 0.00103103, test acc 0.7927, test avg loss 0.523127, throughput 6.02148K wps
[Epoch 21 Batch 30/173] avg loss 0.000917559, throughput 6.14917K wps
[Epoch 21 Batch 60/173] avg loss 0.000894971, throughput 5.99335K wps
[Epoch 21 Batch 90/173] avg loss 0.000930288, throughput 5.99537K wps
[Epoch 21 Batch 120/173] avg loss 0.000887271, throughput 5.99908K wps
[Epoch 21 Batch 150/173] avg loss 0.000762609, throughput 5.99972K wps
Begin Testing...
[Epoch 21] train avg loss 0.000862619, test acc 0.7885, test avg loss 0.538205, throughput 6.02393K wps
[Epoch 22 Batch 30/173] avg loss 0.000813974, throughput 6.14604K wps
[Epoch 22 Batch 60/173] avg loss 0.000727964, throughput 6.01016K wps
[Epoch 22 Batch 90/173] avg loss 0.000758545, throughput 5.99761K wps
[Epoch 22 Batch 120/173] avg loss 0.000717, throughput 5.99959K wps
[Epoch 22 Batch 150/173] avg loss 0.000680599, throughput 6.00081K wps
Begin Testing...
[Epoch 22] train avg loss 0.000747697, test acc 0.7937, test avg loss 0.564974, throughput 6.02646K wps
[Epoch 23 Batch 30/173] avg loss 0.000661211, throughput 6.13989K wps
[Epoch 23 Batch 60/173] avg loss 0.000652073, throughput 5.98354K wps
[Epoch 23 Batch 90/173] avg loss 0.000640652, throughput 5.99211K wps
[Epoch 23 Batch 120/173] avg loss 0.000658878, throughput 5.96692K wps
[Epoch 23 Batch 150/173] avg loss 0.000678387, throughput 5.96055K wps
Begin Testing...
[Epoch 23] train avg loss 0.000648739, test acc 0.7958, test avg loss 0.576048, throughput 6.00671K wps
[Epoch 24 Batch 30/173] avg loss 0.000566414, throughput 6.13855K wps
[Epoch 24 Batch 60/173] avg loss 0.000568878, throughput 6.0007K wps
[Epoch 24 Batch 90/173] avg loss 0.00056934, throughput 6.00624K wps
[Epoch 24 Batch 120/173] avg loss 0.000538314, throughput 5.97728K wps
[Epoch 24 Batch 150/173] avg loss 0.000539458, throughput 6.00176K wps
Begin Testing...
[Epoch 24] train avg loss 0.000551542, test acc 0.7885, test avg loss 0.590735, throughput 6.02158K wps
[Epoch 25 Batch 30/173] avg loss 0.000443393, throughput 6.15631K wps
[Epoch 25 Batch 60/173] avg loss 0.000449203, throughput 5.99243K wps
[Epoch 25 Batch 90/173] avg loss 0.000445037, throughput 5.99009K wps
[Epoch 25 Batch 120/173] avg loss 0.000453557, throughput 5.99611K wps
[Epoch 25 Batch 150/173] avg loss 0.000461968, throughput 6.00192K wps
Begin Testing...
[Epoch 25] train avg loss 0.000450159, test acc 0.7917, test avg loss 0.612235, throughput 6.02481K wps
[Epoch 26 Batch 30/173] avg loss 0.000387355, throughput 6.15807K wps
[Epoch 26 Batch 60/173] avg loss 0.000415614, throughput 5.99081K wps
[Epoch 26 Batch 90/173] avg loss 0.000378623, throughput 6.01249K wps
[Epoch 26 Batch 120/173] avg loss 0.000445796, throughput 6.00247K wps
[Epoch 26 Batch 150/173] avg loss 0.000397629, throughput 6.00879K wps
Begin Testing...
[Epoch 26] train avg loss 0.000401218, test acc 0.7885, test avg loss 0.636116, throughput 6.02756K wps
[Epoch 27 Batch 30/173] avg loss 0.00033496, throughput 6.15418K wps
[Epoch 27 Batch 60/173] avg loss 0.000386535, throughput 5.99366K wps
[Epoch 27 Batch 90/173] avg loss 0.000387835, throughput 6.00525K wps
[Epoch 27 Batch 120/173] avg loss 0.000326451, throughput 5.99694K wps
[Epoch 27 Batch 150/173] avg loss 0.000339179, throughput 6.00313K wps
Begin Testing...
[Epoch 27] train avg loss 0.000354621, test acc 0.7865, test avg loss 0.653204, throughput 6.02911K wps
[Epoch 28 Batch 30/173] avg loss 0.000307951, throughput 6.14642K wps
[Epoch 28 Batch 60/173] avg loss 0.000329748, throughput 6.00696K wps
[Epoch 28 Batch 90/173] avg loss 0.000264059, throughput 5.99596K wps
[Epoch 28 Batch 120/173] avg loss 0.00028369, throughput 6.01179K wps
[Epoch 28 Batch 150/173] avg loss 0.00032884, throughput 5.99805K wps
Begin Testing...
[Epoch 28] train avg loss 0.000299969, test acc 0.7812, test avg loss 0.667511, throughput 6.02653K wps
[Epoch 29 Batch 30/173] avg loss 0.000284818, throughput 6.15083K wps
[Epoch 29 Batch 60/173] avg loss 0.000259256, throughput 6.00165K wps
[Epoch 29 Batch 90/173] avg loss 0.000257676, throughput 6.00509K wps
[Epoch 29 Batch 120/173] avg loss 0.000228206, throughput 6.01262K wps
[Epoch 29 Batch 150/173] avg loss 0.000245599, throughput 5.99706K wps
Begin Testing...
[Epoch 29] train avg loss 0.000258391, test acc 0.7854, test avg loss 0.686215, throughput 6.02733K wps
[Epoch 30 Batch 30/173] avg loss 0.000251323, throughput 6.15628K wps
[Epoch 30 Batch 60/173] avg loss 0.000248667, throughput 5.9993K wps
[Epoch 30 Batch 90/173] avg loss 0.000206624, throughput 6.00304K wps
[Epoch 30 Batch 120/173] avg loss 0.000215062, throughput 5.99116K wps
[Epoch 30 Batch 150/173] avg loss 0.000211053, throughput 6.00282K wps
Begin Testing...
[Epoch 30] train avg loss 0.000223811, test acc 0.7823, test avg loss 0.704699, throughput 6.02739K wps
[Epoch 31 Batch 30/173] avg loss 0.000213796, throughput 6.15257K wps
[Epoch 31 Batch 60/173] avg loss 0.000217085, throughput 5.99137K wps
[Epoch 31 Batch 90/173] avg loss 0.000180647, throughput 5.99169K wps
[Epoch 31 Batch 120/173] avg loss 0.000199536, throughput 6.01442K wps
[Epoch 31 Batch 150/173] avg loss 0.000181738, throughput 6.0025K wps
Begin Testing...
[Epoch 31] train avg loss 0.00020126, test acc 0.7812, test avg loss 0.732118, throughput 6.02884K wps
[Epoch 32 Batch 30/173] avg loss 0.000154968, throughput 6.15549K wps
[Epoch 32 Batch 60/173] avg loss 0.000168462, throughput 6.00459K wps
[Epoch 32 Batch 90/173] avg loss 0.00021753, throughput 5.9945K wps
[Epoch 32 Batch 120/173] avg loss 0.000170977, throughput 6.01688K wps
[Epoch 32 Batch 150/173] avg loss 0.00014334, throughput 6.00008K wps
Begin Testing...
[Epoch 32] train avg loss 0.000171332, test acc 0.7792, test avg loss 0.745216, throughput 6.02975K wps
[Epoch 33 Batch 30/173] avg loss 0.000120213, throughput 6.13308K wps
[Epoch 33 Batch 60/173] avg loss 0.000146178, throughput 5.97966K wps
[Epoch 33 Batch 90/173] avg loss 0.000155387, throughput 5.99433K wps
[Epoch 33 Batch 120/173] avg loss 0.000160249, throughput 5.99952K wps
[Epoch 33 Batch 150/173] avg loss 0.000148346, throughput 5.99785K wps
Begin Testing...
[Epoch 33] train avg loss 0.000149544, test acc 0.7833, test avg loss 0.770435, throughput 6.0181K wps
[Epoch 34 Batch 30/173] avg loss 0.000123428, throughput 6.14148K wps
[Epoch 34 Batch 60/173] avg loss 0.000128328, throughput 5.99203K wps
[Epoch 34 Batch 90/173] avg loss 0.000124256, throughput 5.99843K wps
[Epoch 34 Batch 120/173] avg loss 0.000137816, throughput 5.99399K wps
[Epoch 34 Batch 150/173] avg loss 0.000123363, throughput 5.98989K wps
Begin Testing...
[Epoch 34] train avg loss 0.000127893, test acc 0.7823, test avg loss 0.786965, throughput 6.02K wps
[Epoch 35 Batch 30/173] avg loss 0.000136469, throughput 6.13689K wps
[Epoch 35 Batch 60/173] avg loss 0.000104998, throughput 5.99971K wps
[Epoch 35 Batch 90/173] avg loss 0.000133406, throughput 5.99579K wps
[Epoch 35 Batch 120/173] avg loss 0.000136753, throughput 5.99953K wps
[Epoch 35 Batch 150/173] avg loss 0.000105653, throughput 6.00152K wps
Begin Testing...
[Epoch 35] train avg loss 0.000121859, test acc 0.7823, test avg loss 0.807985, throughput 6.02247K wps
[Epoch 36 Batch 30/173] avg loss 9.50393e-05, throughput 6.15915K wps
[Epoch 36 Batch 60/173] avg loss 0.000100799, throughput 6.00407K wps
[Epoch 36 Batch 90/173] avg loss 0.000114365, throughput 5.99495K wps
[Epoch 36 Batch 120/173] avg loss 0.000117044, throughput 5.99934K wps
[Epoch 36 Batch 150/173] avg loss 8.89675e-05, throughput 5.99306K wps
Begin Testing...
[Epoch 36] train avg loss 0.000101625, test acc 0.7823, test avg loss 0.836043, throughput 6.02534K wps
[Epoch 37 Batch 30/173] avg loss 7.80919e-05, throughput 6.14221K wps
[Epoch 37 Batch 60/173] avg loss 9.96378e-05, throughput 5.99815K wps
[Epoch 37 Batch 90/173] avg loss 8.07089e-05, throughput 6.00813K wps
[Epoch 37 Batch 120/173] avg loss 9.44392e-05, throughput 5.99054K wps
[Epoch 37 Batch 150/173] avg loss 0.000106792, throughput 5.99641K wps
Begin Testing...
[Epoch 37] train avg loss 9.48745e-05, test acc 0.7760, test avg loss 0.847043, throughput 6.02325K wps
[Epoch 38 Batch 30/173] avg loss 8.26537e-05, throughput 6.15087K wps
[Epoch 38 Batch 60/173] avg loss 8.42828e-05, throughput 5.99657K wps
[Epoch 38 Batch 90/173] avg loss 0.000107781, throughput 5.99461K wps
[Epoch 38 Batch 120/173] avg loss 0.000109799, throughput 5.9917K wps
[Epoch 38 Batch 150/173] avg loss 8.53098e-05, throughput 5.98848K wps
Begin Testing...
[Epoch 38] train avg loss 9.23679e-05, test acc 0.7760, test avg loss 0.867209, throughput 6.02064K wps
[Epoch 39 Batch 30/173] avg loss 8.25183e-05, throughput 6.1535K wps
[Epoch 39 Batch 60/173] avg loss 8.0763e-05, throughput 5.98228K wps
[Epoch 39 Batch 90/173] avg loss 8.8131e-05, throughput 5.98584K wps
[Epoch 39 Batch 120/173] avg loss 6.35864e-05, throughput 6.00099K wps
[Epoch 39 Batch 150/173] avg loss 7.56968e-05, throughput 6.00568K wps
Begin Testing...
[Epoch 39] train avg loss 7.99615e-05, test acc 0.7760, test avg loss 0.884518, throughput 6.02257K wps
[Epoch 40 Batch 30/173] avg loss 6.61435e-05, throughput 6.15117K wps
[Epoch 40 Batch 60/173] avg loss 7.31848e-05, throughput 5.99754K wps
[Epoch 40 Batch 90/173] avg loss 6.40721e-05, throughput 6.00205K wps
[Epoch 40 Batch 120/173] avg loss 6.29811e-05, throughput 5.98792K wps
[Epoch 40 Batch 150/173] avg loss 7.80841e-05, throughput 5.98792K wps
Begin Testing...
[Epoch 40] train avg loss 6.84369e-05, test acc 0.7760, test avg loss 0.910644, throughput 6.02083K wps
[Epoch 41 Batch 30/173] avg loss 6.19534e-05, throughput 6.14825K wps
[Epoch 41 Batch 60/173] avg loss 5.40264e-05, throughput 5.98631K wps
[Epoch 41 Batch 90/173] avg loss 4.9516e-05, throughput 5.99564K wps
[Epoch 41 Batch 120/173] avg loss 7.13552e-05, throughput 5.99199K wps
[Epoch 41 Batch 150/173] avg loss 5.78523e-05, throughput 6.00014K wps
Begin Testing...
[Epoch 41] train avg loss 6.04684e-05, test acc 0.7771, test avg loss 0.920384, throughput 6.02099K wps
[Epoch 42 Batch 30/173] avg loss 4.75421e-05, throughput 6.13748K wps
[Epoch 42 Batch 60/173] avg loss 4.19893e-05, throughput 5.99376K wps
[Epoch 42 Batch 90/173] avg loss 4.96171e-05, throughput 6.00289K wps
[Epoch 42 Batch 120/173] avg loss 6.01857e-05, throughput 6.00183K wps
[Epoch 42 Batch 150/173] avg loss 5.05996e-05, throughput 5.99611K wps
Begin Testing...
[Epoch 42] train avg loss 5.34498e-05, test acc 0.7802, test avg loss 0.935025, throughput 6.02173K wps
[Epoch 43 Batch 30/173] avg loss 4.92776e-05, throughput 6.14542K wps
[Epoch 43 Batch 60/173] avg loss 5.78478e-05, throughput 5.9924K wps
[Epoch 43 Batch 90/173] avg loss 4.50646e-05, throughput 5.99255K wps
[Epoch 43 Batch 120/173] avg loss 3.45419e-05, throughput 5.99813K wps
[Epoch 43 Batch 150/173] avg loss 7.65694e-05, throughput 5.99883K wps
Begin Testing...
[Epoch 43] train avg loss 5.48494e-05, test acc 0.7719, test avg loss 0.956034, throughput 6.0205K wps
[Epoch 44 Batch 30/173] avg loss 3.51754e-05, throughput 6.1568K wps
[Epoch 44 Batch 60/173] avg loss 3.12813e-05, throughput 6.00512K wps
[Epoch 44 Batch 90/173] avg loss 4.36128e-05, throughput 5.99543K wps
[Epoch 44 Batch 120/173] avg loss 4.27045e-05, throughput 5.99395K wps
[Epoch 44 Batch 150/173] avg loss 5.44197e-05, throughput 5.99603K wps
Begin Testing...
[Epoch 44] train avg loss 4.20875e-05, test acc 0.7740, test avg loss 0.976469, throughput 6.02563K wps
[Epoch 45 Batch 30/173] avg loss 3.49565e-05, throughput 6.14151K wps
[Epoch 45 Batch 60/173] avg loss 3.9889e-05, throughput 5.99204K wps
[Epoch 45 Batch 90/173] avg loss 4.76585e-05, throughput 5.99097K wps
[Epoch 45 Batch 120/173] avg loss 3.19675e-05, throughput 5.99155K wps
[Epoch 45 Batch 150/173] avg loss 3.54789e-05, throughput 6.00535K wps
Begin Testing...
[Epoch 45] train avg loss 4.01902e-05, test acc 0.7740, test avg loss 0.996498, throughput 6.02121K wps
[Epoch 46 Batch 30/173] avg loss 3.15679e-05, throughput 6.14305K wps
[Epoch 46 Batch 60/173] avg loss 3.72083e-05, throughput 5.98931K wps
[Epoch 46 Batch 90/173] avg loss 3.45597e-05, throughput 5.99312K wps
[Epoch 46 Batch 120/173] avg loss 3.17446e-05, throughput 5.99397K wps
[Epoch 46 Batch 150/173] avg loss 5.80335e-05, throughput 5.99076K wps
Begin Testing...
[Epoch 46] train avg loss 3.73901e-05, test acc 0.7719, test avg loss 1.02009, throughput 6.01752K wps
[Epoch 47 Batch 30/173] avg loss 3.77729e-05, throughput 6.14154K wps
[Epoch 47 Batch 60/173] avg loss 3.49321e-05, throughput 5.98451K wps
[Epoch 47 Batch 90/173] avg loss 3.27343e-05, throughput 5.99327K wps
[Epoch 47 Batch 120/173] avg loss 3.75879e-05, throughput 6.00355K wps
[Epoch 47 Batch 150/173] avg loss 2.84759e-05, throughput 5.98768K wps
Begin Testing...
[Epoch 47] train avg loss 3.30655e-05, test acc 0.7729, test avg loss 1.02884, throughput 6.01782K wps
[Epoch 48 Batch 30/173] avg loss 3.31922e-05, throughput 6.13375K wps
[Epoch 48 Batch 60/173] avg loss 2.8128e-05, throughput 5.99942K wps
[Epoch 48 Batch 90/173] avg loss 2.58881e-05, throughput 5.98865K wps
[Epoch 48 Batch 120/173] avg loss 4.10257e-05, throughput 5.98977K wps
[Epoch 48 Batch 150/173] avg loss 2.79598e-05, throughput 6.00497K wps
Begin Testing...
[Epoch 48] train avg loss 3.03305e-05, test acc 0.7708, test avg loss 1.04769, throughput 6.01934K wps
[Epoch 49 Batch 30/173] avg loss 2.857e-05, throughput 6.15414K wps
[Epoch 49 Batch 60/173] avg loss 2.66884e-05, throughput 6.00367K wps
[Epoch 49 Batch 90/173] avg loss 2.81847e-05, throughput 6.00827K wps
[Epoch 49 Batch 120/173] avg loss 2.18353e-05, throughput 5.98393K wps
[Epoch 49 Batch 150/173] avg loss 2.95034e-05, throughput 5.99345K wps
Begin Testing...
[Epoch 49] train avg loss 2.78015e-05, test acc 0.7677, test avg loss 1.07654, throughput 6.02535K wps
[Epoch 50 Batch 30/173] avg loss 3.33329e-05, throughput 6.14055K wps
[Epoch 50 Batch 60/173] avg loss 3.30785e-05, throughput 5.99244K wps
[Epoch 50 Batch 90/173] avg loss 3.27173e-05, throughput 5.99985K wps
[Epoch 50 Batch 120/173] avg loss 2.32039e-05, throughput 5.99798K wps
[Epoch 50 Batch 150/173] avg loss 2.4531e-05, throughput 6.0088K wps
Begin Testing...
[Epoch 50] train avg loss 2.83992e-05, test acc 0.7635, test avg loss 1.09563, throughput 6.02227K wps
[Epoch 51 Batch 30/173] avg loss 2.19547e-05, throughput 6.1464K wps
[Epoch 51 Batch 60/173] avg loss 1.53994e-05, throughput 5.99303K wps
[Epoch 51 Batch 90/173] avg loss 1.70994e-05, throughput 5.99068K wps
[Epoch 51 Batch 120/173] avg loss 2.0124e-05, throughput 6.00066K wps
[Epoch 51 Batch 150/173] avg loss 2.87547e-05, throughput 6.00025K wps
Begin Testing...
[Epoch 51] train avg loss 1.9747e-05, test acc 0.7698, test avg loss 1.11051, throughput 6.02162K wps
[Epoch 52 Batch 30/173] avg loss 1.62408e-05, throughput 6.16394K wps
[Epoch 52 Batch 60/173] avg loss 2.04869e-05, throughput 6.02105K wps
[Epoch 52 Batch 90/173] avg loss 1.74591e-05, throughput 6.00504K wps
[Epoch 52 Batch 120/173] avg loss 1.61603e-05, throughput 6.00962K wps
[Epoch 52 Batch 150/173] avg loss 2.05389e-05, throughput 6.00471K wps
Begin Testing...
[Epoch 52] train avg loss 1.94514e-05, test acc 0.7667, test avg loss 1.13091, throughput 6.03451K wps
[Epoch 53 Batch 30/173] avg loss 2.10032e-05, throughput 6.1468K wps
[Epoch 53 Batch 60/173] avg loss 2.03863e-05, throughput 5.98994K wps
[Epoch 53 Batch 90/173] avg loss 1.43258e-05, throughput 5.98616K wps
[Epoch 53 Batch 120/173] avg loss 1.46581e-05, throughput 5.98871K wps
[Epoch 53 Batch 150/173] avg loss 1.42068e-05, throughput 5.9616K wps
Begin Testing...
[Epoch 53] train avg loss 1.82729e-05, test acc 0.7656, test avg loss 1.15871, throughput 6.00465K wps
[Epoch 54 Batch 30/173] avg loss 1.69866e-05, throughput 6.12708K wps
[Epoch 54 Batch 60/173] avg loss 1.26413e-05, throughput 6.01435K wps
[Epoch 54 Batch 90/173] avg loss 1.90203e-05, throughput 6.00103K wps
[Epoch 54 Batch 120/173] avg loss 1.45073e-05, throughput 6.00788K wps
[Epoch 54 Batch 150/173] avg loss 1.56377e-05, throughput 5.99465K wps
Begin Testing...
[Epoch 54] train avg loss 1.59037e-05, test acc 0.7688, test avg loss 1.17322, throughput 6.02223K wps
[Epoch 55 Batch 30/173] avg loss 1.37651e-05, throughput 6.14954K wps
[Epoch 55 Batch 60/173] avg loss 1.21685e-05, throughput 6.00645K wps
[Epoch 55 Batch 90/173] avg loss 1.35425e-05, throughput 6.00593K wps
[Epoch 55 Batch 120/173] avg loss 1.30701e-05, throughput 5.99084K wps
[Epoch 55 Batch 150/173] avg loss 2.12925e-05, throughput 5.99595K wps
Begin Testing...
[Epoch 55] train avg loss 1.45259e-05, test acc 0.7677, test avg loss 1.18781, throughput 6.02707K wps
[Epoch 56 Batch 30/173] avg loss 1.07773e-05, throughput 6.15243K wps
[Epoch 56 Batch 60/173] avg loss 1.10481e-05, throughput 5.99051K wps
[Epoch 56 Batch 90/173] avg loss 1.48214e-05, throughput 6.01159K wps
[Epoch 56 Batch 120/173] avg loss 1.77082e-05, throughput 5.99347K wps
[Epoch 56 Batch 150/173] avg loss 1.34583e-05, throughput 5.99281K wps
Begin Testing...
[Epoch 56] train avg loss 1.3541e-05, test acc 0.7698, test avg loss 1.2118, throughput 6.0224K wps
[Epoch 57 Batch 30/173] avg loss 1.35929e-05, throughput 6.15198K wps
[Epoch 57 Batch 60/173] avg loss 8.32685e-06, throughput 6.00365K wps
[Epoch 57 Batch 90/173] avg loss 1.06345e-05, throughput 5.99145K wps
[Epoch 57 Batch 120/173] avg loss 1.7225e-05, throughput 6.0084K wps
[Epoch 57 Batch 150/173] avg loss 1.56554e-05, throughput 6.00573K wps
Begin Testing...
[Epoch 57] train avg loss 1.33228e-05, test acc 0.7646, test avg loss 1.22134, throughput 6.02793K wps
[Epoch 58 Batch 30/173] avg loss 1.12848e-05, throughput 6.13583K wps
[Epoch 58 Batch 60/173] avg loss 1.84565e-05, throughput 6.00179K wps
[Epoch 58 Batch 90/173] avg loss 6.45819e-06, throughput 6.00558K wps
[Epoch 58 Batch 120/173] avg loss 9.35174e-06, throughput 6.01031K wps
[Epoch 58 Batch 150/173] avg loss 1.03537e-05, throughput 6.0036K wps
Begin Testing...
[Epoch 58] train avg loss 1.12576e-05, test acc 0.7646, test avg loss 1.25059, throughput 6.02911K wps
[Epoch 59 Batch 30/173] avg loss 1.00398e-05, throughput 6.15662K wps
[Epoch 59 Batch 60/173] avg loss 1.15765e-05, throughput 5.99133K wps
[Epoch 59 Batch 90/173] avg loss 8.77916e-06, throughput 6.00095K wps
[Epoch 59 Batch 120/173] avg loss 1.55742e-05, throughput 6.00775K wps
[Epoch 59 Batch 150/173] avg loss 1.00894e-05, throughput 5.98823K wps
Begin Testing...
[Epoch 59] train avg loss 1.11062e-05, test acc 0.7667, test avg loss 1.27407, throughput 6.02421K wps
Test loss 0.421029, test acc 0.8086
Total time cost 358.42s
9596 1066
[Epoch 0 Batch 30/173] avg loss 0.015438, throughput 5.79137K wps
[Epoch 0 Batch 60/173] avg loss 0.01505, throughput 6.01008K wps
[Epoch 0 Batch 90/173] avg loss 0.0148122, throughput 5.99555K wps
[Epoch 0 Batch 120/173] avg loss 0.0145189, throughput 5.99737K wps
[Epoch 0 Batch 150/173] avg loss 0.0143428, throughput 6.00645K wps
Begin Testing...
[Epoch 0] train avg loss 0.0148157, test acc 0.6042, test avg loss 0.664189, throughput 5.96455K wps
Observed Improvement.
Begin Testing...
[Epoch 1 Batch 30/173] avg loss 0.0137525, throughput 6.13721K wps
[Epoch 1 Batch 60/173] avg loss 0.0134763, throughput 5.99636K wps
[Epoch 1 Batch 90/173] avg loss 0.0132434, throughput 5.99181K wps
[Epoch 1 Batch 120/173] avg loss 0.0133819, throughput 5.99101K wps
[Epoch 1 Batch 150/173] avg loss 0.0134191, throughput 6.00153K wps
Begin Testing...
[Epoch 1] train avg loss 0.0134811, test acc 0.6260, test avg loss 0.645767, throughput 6.01888K wps
Observed Improvement.
Begin Testing...
[Epoch 2 Batch 30/173] avg loss 0.0128142, throughput 6.14681K wps
[Epoch 2 Batch 60/173] avg loss 0.0127127, throughput 5.999K wps
[Epoch 2 Batch 90/173] avg loss 0.0125136, throughput 6.00182K wps
[Epoch 2 Batch 120/173] avg loss 0.0126046, throughput 6.00013K wps
[Epoch 2 Batch 150/173] avg loss 0.0123222, throughput 5.99328K wps
Begin Testing...
[Epoch 2] train avg loss 0.0125479, test acc 0.6562, test avg loss 0.625197, throughput 6.02482K wps
Observed Improvement.
Begin Testing...
[Epoch 3 Batch 30/173] avg loss 0.011889, throughput 6.14464K wps
[Epoch 3 Batch 60/173] avg loss 0.0118464, throughput 5.98776K wps
[Epoch 3 Batch 90/173] avg loss 0.0115105, throughput 6.00864K wps
[Epoch 3 Batch 120/173] avg loss 0.011603, throughput 6.00321K wps
[Epoch 3 Batch 150/173] avg loss 0.0115156, throughput 5.9938K wps
Begin Testing...
[Epoch 3] train avg loss 0.0116542, test acc 0.7052, test avg loss 0.600792, throughput 6.02192K wps
Observed Improvement.
Begin Testing...
[Epoch 4 Batch 30/173] avg loss 0.0109189, throughput 6.15519K wps
[Epoch 4 Batch 60/173] avg loss 0.0108698, throughput 6.00547K wps
[Epoch 4 Batch 90/173] avg loss 0.0107241, throughput 5.98824K wps
[Epoch 4 Batch 120/173] avg loss 0.0105995, throughput 5.99584K wps
[Epoch 4 Batch 150/173] avg loss 0.0105533, throughput 6.00142K wps
Begin Testing...
[Epoch 4] train avg loss 0.0106985, test acc 0.7375, test avg loss 0.570432, throughput 6.02526K wps
Observed Improvement.
Begin Testing...
[Epoch 5 Batch 30/173] avg loss 0.00963978, throughput 6.13659K wps
[Epoch 5 Batch 60/173] avg loss 0.00988564, throughput 6.0019K wps
[Epoch 5 Batch 90/173] avg loss 0.00949809, throughput 5.99252K wps
[Epoch 5 Batch 120/173] avg loss 0.00955926, throughput 5.99788K wps
[Epoch 5 Batch 150/173] avg loss 0.0095623, throughput 5.9873K wps
Begin Testing...
[Epoch 5] train avg loss 0.00961769, test acc 0.7667, test avg loss 0.531268, throughput 6.02153K wps
Observed Improvement.
Begin Testing...
[Epoch 6 Batch 30/173] avg loss 0.00907422, throughput 6.14671K wps
[Epoch 6 Batch 60/173] avg loss 0.00908854, throughput 5.99263K wps
[Epoch 6 Batch 90/173] avg loss 0.00861592, throughput 6.01574K wps
[Epoch 6 Batch 120/173] avg loss 0.00811923, throughput 6.01056K wps
[Epoch 6 Batch 150/173] avg loss 0.00840462, throughput 5.99842K wps
Begin Testing...
[Epoch 6] train avg loss 0.00864989, test acc 0.7771, test avg loss 0.500314, throughput 6.0274K wps
Observed Improvement.
Begin Testing...
[Epoch 7 Batch 30/173] avg loss 0.00784245, throughput 6.15833K wps
[Epoch 7 Batch 60/173] avg loss 0.00748681, throughput 6.01027K wps
[Epoch 7 Batch 90/173] avg loss 0.00736389, throughput 6.00714K wps
[Epoch 7 Batch 120/173] avg loss 0.00731014, throughput 6.00908K wps
[Epoch 7 Batch 150/173] avg loss 0.00745271, throughput 6.01243K wps
Begin Testing...
[Epoch 7] train avg loss 0.00745944, test acc 0.7802, test avg loss 0.475877, throughput 6.03604K wps
Observed Improvement.
Begin Testing...
[Epoch 8 Batch 30/173] avg loss 0.00671752, throughput 6.14932K wps
[Epoch 8 Batch 60/173] avg loss 0.0065014, throughput 5.99215K wps
[Epoch 8 Batch 90/173] avg loss 0.00643796, throughput 6.00123K wps
[Epoch 8 Batch 120/173] avg loss 0.00664138, throughput 5.98404K wps
[Epoch 8 Batch 150/173] avg loss 0.00626001, throughput 5.99264K wps
Begin Testing...
[Epoch 8] train avg loss 0.00648529, test acc 0.8000, test avg loss 0.45403, throughput 6.01959K wps
Observed Improvement.
Begin Testing...
[Epoch 9 Batch 30/173] avg loss 0.00544402, throughput 6.15574K wps
[Epoch 9 Batch 60/173] avg loss 0.00582831, throughput 6.00317K wps
[Epoch 9 Batch 90/173] avg loss 0.00561562, throughput 5.99296K wps
[Epoch 9 Batch 120/173] avg loss 0.00548294, throughput 5.99767K wps
[Epoch 9 Batch 150/173] avg loss 0.00563259, throughput 5.98575K wps
Begin Testing...
[Epoch 9] train avg loss 0.00556996, test acc 0.8052, test avg loss 0.442095, throughput 6.02194K wps
Observed Improvement.
Begin Testing...
[Epoch 10 Batch 30/173] avg loss 0.00481701, throughput 6.14653K wps
[Epoch 10 Batch 60/173] avg loss 0.0051847, throughput 5.99967K wps
[Epoch 10 Batch 90/173] avg loss 0.00474289, throughput 6.00589K wps
[Epoch 10 Batch 120/173] avg loss 0.00476867, throughput 5.98892K wps
[Epoch 10 Batch 150/173] avg loss 0.00473373, throughput 5.99896K wps
Begin Testing...
[Epoch 10] train avg loss 0.00482648, test acc 0.8010, test avg loss 0.438738, throughput 6.02467K wps
[Epoch 11 Batch 30/173] avg loss 0.00425395, throughput 6.15537K wps
[Epoch 11 Batch 60/173] avg loss 0.00425143, throughput 5.99944K wps
[Epoch 11 Batch 90/173] avg loss 0.00440106, throughput 6.00712K wps
[Epoch 11 Batch 120/173] avg loss 0.00415379, throughput 5.99395K wps
[Epoch 11 Batch 150/173] avg loss 0.00425022, throughput 5.9902K wps
Begin Testing...
[Epoch 11] train avg loss 0.00423461, test acc 0.7979, test avg loss 0.435948, throughput 6.02461K wps
[Epoch 12 Batch 30/173] avg loss 0.00350517, throughput 6.14261K wps
[Epoch 12 Batch 60/173] avg loss 0.00371578, throughput 6.00088K wps
[Epoch 12 Batch 90/173] avg loss 0.00370083, throughput 5.97945K wps
[Epoch 12 Batch 120/173] avg loss 0.00361556, throughput 5.98017K wps
[Epoch 12 Batch 150/173] avg loss 0.00339437, throughput 5.99325K wps
Begin Testing...
[Epoch 12] train avg loss 0.00358205, test acc 0.7927, test avg loss 0.438511, throughput 6.01571K wps
[Epoch 13 Batch 30/173] avg loss 0.00319799, throughput 6.13639K wps
[Epoch 13 Batch 60/173] avg loss 0.00302433, throughput 6.00444K wps
[Epoch 13 Batch 90/173] avg loss 0.00304823, throughput 6.0078K wps
[Epoch 13 Batch 120/173] avg loss 0.00320322, throughput 5.99885K wps
[Epoch 13 Batch 150/173] avg loss 0.00292272, throughput 6.00725K wps
Begin Testing...
[Epoch 13] train avg loss 0.00304784, test acc 0.7865, test avg loss 0.446777, throughput 6.02878K wps
[Epoch 14 Batch 30/173] avg loss 0.00246737, throughput 6.1649K wps
[Epoch 14 Batch 60/173] avg loss 0.00241521, throughput 6.0081K wps
[Epoch 14 Batch 90/173] avg loss 0.00253433, throughput 6.01465K wps
[Epoch 14 Batch 120/173] avg loss 0.00264567, throughput 6.01428K wps
[Epoch 14 Batch 150/173] avg loss 0.00257993, throughput 5.9945K wps
Begin Testing...
[Epoch 14] train avg loss 0.0025391, test acc 0.7896, test avg loss 0.459172, throughput 6.03251K wps
[Epoch 15 Batch 30/173] avg loss 0.00207201, throughput 6.13953K wps
[Epoch 15 Batch 60/173] avg loss 0.00216262, throughput 5.9942K wps
[Epoch 15 Batch 90/173] avg loss 0.00216028, throughput 6.0189K wps
[Epoch 15 Batch 120/173] avg loss 0.00214303, throughput 6.00211K wps
[Epoch 15 Batch 150/173] avg loss 0.00216751, throughput 5.98769K wps
Begin Testing...
[Epoch 15] train avg loss 0.00217584, test acc 0.7906, test avg loss 0.475394, throughput 6.0236K wps
[Epoch 16 Batch 30/173] avg loss 0.00189141, throughput 6.16371K wps
[Epoch 16 Batch 60/173] avg loss 0.00183918, throughput 6.00047K wps
[Epoch 16 Batch 90/173] avg loss 0.00189206, throughput 6.00427K wps
[Epoch 16 Batch 120/173] avg loss 0.00183878, throughput 5.99113K wps
[Epoch 16 Batch 150/173] avg loss 0.00177005, throughput 5.99077K wps
Begin Testing...
[Epoch 16] train avg loss 0.00183182, test acc 0.7812, test avg loss 0.499359, throughput 6.02469K wps
[Epoch 17 Batch 30/173] avg loss 0.00155453, throughput 6.13707K wps
[Epoch 17 Batch 60/173] avg loss 0.0014953, throughput 5.98656K wps
[Epoch 17 Batch 90/173] avg loss 0.00148307, throughput 5.99411K wps
[Epoch 17 Batch 120/173] avg loss 0.0017079, throughput 5.99107K wps
[Epoch 17 Batch 150/173] avg loss 0.00167937, throughput 5.99769K wps
Begin Testing...
[Epoch 17] train avg loss 0.00157782, test acc 0.7792, test avg loss 0.512287, throughput 6.01757K wps
[Epoch 18 Batch 30/173] avg loss 0.00137484, throughput 6.1436K wps
[Epoch 18 Batch 60/173] avg loss 0.00133943, throughput 5.99694K wps
[Epoch 18 Batch 90/173] avg loss 0.00135419, throughput 6.00131K wps
[Epoch 18 Batch 120/173] avg loss 0.00127577, throughput 5.99553K wps
[Epoch 18 Batch 150/173] avg loss 0.00133922, throughput 5.9996K wps
Begin Testing...
[Epoch 18] train avg loss 0.00133533, test acc 0.7760, test avg loss 0.530143, throughput 6.02252K wps
[Epoch 19 Batch 30/173] avg loss 0.00123441, throughput 6.14518K wps
[Epoch 19 Batch 60/173] avg loss 0.00114998, throughput 5.99574K wps
[Epoch 19 Batch 90/173] avg loss 0.00110121, throughput 5.99686K wps
[Epoch 19 Batch 120/173] avg loss 0.00105408, throughput 5.99043K wps
[Epoch 19 Batch 150/173] avg loss 0.00107723, throughput 5.99426K wps
Begin Testing...
[Epoch 19] train avg loss 0.00111995, test acc 0.7771, test avg loss 0.558443, throughput 6.02058K wps
[Epoch 20 Batch 30/173] avg loss 0.00101703, throughput 6.14913K wps
[Epoch 20 Batch 60/173] avg loss 0.000924095, throughput 6.00272K wps
[Epoch 20 Batch 90/173] avg loss 0.000953164, throughput 5.98884K wps
[Epoch 20 Batch 120/173] avg loss 0.00100377, throughput 5.99878K wps
[Epoch 20 Batch 150/173] avg loss 0.000856042, throughput 5.95994K wps
Begin Testing...
[Epoch 20] train avg loss 0.000937677, test acc 0.7667, test avg loss 0.57579, throughput 6.01841K wps
[Epoch 21 Batch 30/173] avg loss 0.00073112, throughput 6.14973K wps
[Epoch 21 Batch 60/173] avg loss 0.000757765, throughput 5.99573K wps
[Epoch 21 Batch 90/173] avg loss 0.000767385, throughput 5.99765K wps
[Epoch 21 Batch 120/173] avg loss 0.000838128, throughput 5.99365K wps
[Epoch 21 Batch 150/173] avg loss 0.000829266, throughput 5.99713K wps
Begin Testing...
[Epoch 21] train avg loss 0.000804602, test acc 0.7708, test avg loss 0.598076, throughput 6.02394K wps
[Epoch 22 Batch 30/173] avg loss 0.000648545, throughput 6.14458K wps
[Epoch 22 Batch 60/173] avg loss 0.000711993, throughput 5.99987K wps
[Epoch 22 Batch 90/173] avg loss 0.000696275, throughput 6.00521K wps
[Epoch 22 Batch 120/173] avg loss 0.000709213, throughput 5.9931K wps
[Epoch 22 Batch 150/173] avg loss 0.000738988, throughput 5.99217K wps
Begin Testing...
[Epoch 22] train avg loss 0.000698732, test acc 0.7698, test avg loss 0.62198, throughput 6.02372K wps
[Epoch 23 Batch 30/173] avg loss 0.000568418, throughput 6.14066K wps
[Epoch 23 Batch 60/173] avg loss 0.000541259, throughput 5.99323K wps
[Epoch 23 Batch 90/173] avg loss 0.000570232, throughput 6.00334K wps
[Epoch 23 Batch 120/173] avg loss 0.000558305, throughput 6.00277K wps
[Epoch 23 Batch 150/173] avg loss 0.00064205, throughput 5.99277K wps
Begin Testing...
[Epoch 23] train avg loss 0.000580646, test acc 0.7688, test avg loss 0.647637, throughput 6.01344K wps
[Epoch 24 Batch 30/173] avg loss 0.000467536, throughput 6.1281K wps
[Epoch 24 Batch 60/173] avg loss 0.000456377, throughput 5.99815K wps
[Epoch 24 Batch 90/173] avg loss 0.000504124, throughput 5.99664K wps
[Epoch 24 Batch 120/173] avg loss 0.000485323, throughput 6.00914K wps
[Epoch 24 Batch 150/173] avg loss 0.000472967, throughput 5.98893K wps
Begin Testing...
[Epoch 24] train avg loss 0.000475472, test acc 0.7656, test avg loss 0.67317, throughput 6.01954K wps
[Epoch 25 Batch 30/173] avg loss 0.000424914, throughput 6.14203K wps
[Epoch 25 Batch 60/173] avg loss 0.000393486, throughput 5.99755K wps
[Epoch 25 Batch 90/173] avg loss 0.000425048, throughput 5.99881K wps
[Epoch 25 Batch 120/173] avg loss 0.000467585, throughput 5.99527K wps
[Epoch 25 Batch 150/173] avg loss 0.000473203, throughput 5.99489K wps
Begin Testing...
[Epoch 25] train avg loss 0.000435913, test acc 0.7667, test avg loss 0.701485, throughput 6.02394K wps
[Epoch 26 Batch 30/173] avg loss 0.000342649, throughput 6.14785K wps
[Epoch 26 Batch 60/173] avg loss 0.00040675, throughput 5.98957K wps
[Epoch 26 Batch 90/173] avg loss 0.000401586, throughput 6.00518K wps
[Epoch 26 Batch 120/173] avg loss 0.000333401, throughput 6.01472K wps
[Epoch 26 Batch 150/173] avg loss 0.000368564, throughput 6.00732K wps
Begin Testing...
[Epoch 26] train avg loss 0.000372626, test acc 0.7646, test avg loss 0.720964, throughput 6.02984K wps
[Epoch 27 Batch 30/173] avg loss 0.000277766, throughput 6.1575K wps
[Epoch 27 Batch 60/173] avg loss 0.00031502, throughput 6.00055K wps
[Epoch 27 Batch 90/173] avg loss 0.000304948, throughput 5.99006K wps
[Epoch 27 Batch 120/173] avg loss 0.000307769, throughput 5.99611K wps
[Epoch 27 Batch 150/173] avg loss 0.000268989, throughput 6.01346K wps
Begin Testing...
[Epoch 27] train avg loss 0.000307038, test acc 0.7635, test avg loss 0.747356, throughput 6.02943K wps
[Epoch 28 Batch 30/173] avg loss 0.000256192, throughput 6.15695K wps
[Epoch 28 Batch 60/173] avg loss 0.000293943, throughput 5.99035K wps
[Epoch 28 Batch 90/173] avg loss 0.000284109, throughput 6.00739K wps
[Epoch 28 Batch 120/173] avg loss 0.000326083, throughput 6.00804K wps
[Epoch 28 Batch 150/173] avg loss 0.000278801, throughput 6.01537K wps
Begin Testing...
[Epoch 28] train avg loss 0.000284622, test acc 0.7583, test avg loss 0.77518, throughput 6.03324K wps
[Epoch 29 Batch 30/173] avg loss 0.000237342, throughput 6.1578K wps
[Epoch 29 Batch 60/173] avg loss 0.000230008, throughput 6.01155K wps
[Epoch 29 Batch 90/173] avg loss 0.000227523, throughput 5.99312K wps
[Epoch 29 Batch 120/173] avg loss 0.000244104, throughput 5.9981K wps
[Epoch 29 Batch 150/173] avg loss 0.000240873, throughput 6.00872K wps
Begin Testing...
[Epoch 29] train avg loss 0.000237892, test acc 0.7635, test avg loss 0.792301, throughput 6.02991K wps
[Epoch 30 Batch 30/173] avg loss 0.000199458, throughput 6.15245K wps
[Epoch 30 Batch 60/173] avg loss 0.000200891, throughput 5.99847K wps
[Epoch 30 Batch 90/173] avg loss 0.000194882, throughput 5.99776K wps
[Epoch 30 Batch 120/173] avg loss 0.000202514, throughput 6.01096K wps
[Epoch 30 Batch 150/173] avg loss 0.000231556, throughput 5.99859K wps
Begin Testing...
[Epoch 30] train avg loss 0.000203105, test acc 0.7615, test avg loss 0.818228, throughput 6.02817K wps
[Epoch 31 Batch 30/173] avg loss 0.000166931, throughput 6.1351K wps
[Epoch 31 Batch 60/173] avg loss 0.000197265, throughput 6.0042K wps
[Epoch 31 Batch 90/173] avg loss 0.000167647, throughput 6.01369K wps
[Epoch 31 Batch 120/173] avg loss 0.000178576, throughput 6.00677K wps
[Epoch 31 Batch 150/173] avg loss 0.000177414, throughput 5.99291K wps
Begin Testing...
[Epoch 31] train avg loss 0.000178085, test acc 0.7656, test avg loss 0.843861, throughput 6.02771K wps
[Epoch 32 Batch 30/173] avg loss 0.000207923, throughput 6.14161K wps
[Epoch 32 Batch 60/173] avg loss 0.000141209, throughput 5.99227K wps
[Epoch 32 Batch 90/173] avg loss 0.000142203, throughput 6.00623K wps
[Epoch 32 Batch 120/173] avg loss 0.000153807, throughput 6.00775K wps
[Epoch 32 Batch 150/173] avg loss 0.000153273, throughput 5.99816K wps
Begin Testing...
[Epoch 32] train avg loss 0.000165153, test acc 0.7625, test avg loss 0.865583, throughput 6.02585K wps
[Epoch 33 Batch 30/173] avg loss 0.000121794, throughput 6.1614K wps
[Epoch 33 Batch 60/173] avg loss 0.000129073, throughput 5.99819K wps
[Epoch 33 Batch 90/173] avg loss 0.000132426, throughput 6.01146K wps
[Epoch 33 Batch 120/173] avg loss 0.000138708, throughput 6.00781K wps
[Epoch 33 Batch 150/173] avg loss 0.000176989, throughput 6.00378K wps
Begin Testing...
[Epoch 33] train avg loss 0.000141025, test acc 0.7615, test avg loss 0.886543, throughput 6.03346K wps
[Epoch 34 Batch 30/173] avg loss 0.00013067, throughput 6.15236K wps
[Epoch 34 Batch 60/173] avg loss 0.000124527, throughput 5.99316K wps
[Epoch 34 Batch 90/173] avg loss 0.000116581, throughput 5.99577K wps
[Epoch 34 Batch 120/173] avg loss 0.000118708, throughput 5.98759K wps
[Epoch 34 Batch 150/173] avg loss 0.00012865, throughput 5.99118K wps
Begin Testing...
[Epoch 34] train avg loss 0.000125613, test acc 0.7583, test avg loss 0.919681, throughput 6.02007K wps
[Epoch 35 Batch 30/173] avg loss 0.000117511, throughput 6.15276K wps
[Epoch 35 Batch 60/173] avg loss 0.00011107, throughput 6.00205K wps
[Epoch 35 Batch 90/173] avg loss 0.000124356, throughput 6.00717K wps
[Epoch 35 Batch 120/173] avg loss 9.66867e-05, throughput 5.99875K wps
[Epoch 35 Batch 150/173] avg loss 0.000107052, throughput 5.99363K wps
Begin Testing...
[Epoch 35] train avg loss 0.000112316, test acc 0.7510, test avg loss 0.944074, throughput 6.02631K wps
[Epoch 36 Batch 30/173] avg loss 8.88411e-05, throughput 6.13765K wps
[Epoch 36 Batch 60/173] avg loss 9.55341e-05, throughput 5.99608K wps
[Epoch 36 Batch 90/173] avg loss 8.0728e-05, throughput 5.99606K wps
[Epoch 36 Batch 120/173] avg loss 8.37496e-05, throughput 5.99928K wps
[Epoch 36 Batch 150/173] avg loss 0.000103538, throughput 5.99739K wps
Begin Testing...
[Epoch 36] train avg loss 9.11172e-05, test acc 0.7583, test avg loss 0.956822, throughput 6.02298K wps
[Epoch 37 Batch 30/173] avg loss 8.48709e-05, throughput 6.14072K wps
[Epoch 37 Batch 60/173] avg loss 7.41357e-05, throughput 5.99661K wps
[Epoch 37 Batch 90/173] avg loss 9.17508e-05, throughput 6.00664K wps
[Epoch 37 Batch 120/173] avg loss 9.2015e-05, throughput 5.99696K wps
[Epoch 37 Batch 150/173] avg loss 7.60027e-05, throughput 5.98692K wps
Begin Testing...
[Epoch 37] train avg loss 8.39967e-05, test acc 0.7562, test avg loss 0.978009, throughput 6.0215K wps
[Epoch 38 Batch 30/173] avg loss 7.46901e-05, throughput 6.14329K wps
[Epoch 38 Batch 60/173] avg loss 6.95122e-05, throughput 5.98898K wps
[Epoch 38 Batch 90/173] avg loss 7.00128e-05, throughput 5.99379K wps
[Epoch 38 Batch 120/173] avg loss 8.32694e-05, throughput 5.99489K wps
[Epoch 38 Batch 150/173] avg loss 7.55622e-05, throughput 5.99138K wps
Begin Testing...
[Epoch 38] train avg loss 7.31208e-05, test acc 0.7583, test avg loss 0.998918, throughput 6.01846K wps
[Epoch 39 Batch 30/173] avg loss 6.48153e-05, throughput 6.14708K wps
[Epoch 39 Batch 60/173] avg loss 5.98667e-05, throughput 6.00339K wps
[Epoch 39 Batch 90/173] avg loss 8.14819e-05, throughput 6.00086K wps
[Epoch 39 Batch 120/173] avg loss 6.66736e-05, throughput 5.98836K wps
[Epoch 39 Batch 150/173] avg loss 6.16431e-05, throughput 6.00037K wps
Begin Testing...
[Epoch 39] train avg loss 6.58179e-05, test acc 0.7552, test avg loss 1.02359, throughput 6.0249K wps
[Epoch 40 Batch 30/173] avg loss 5.65128e-05, throughput 6.13388K wps
[Epoch 40 Batch 60/173] avg loss 5.15384e-05, throughput 6.00262K wps
[Epoch 40 Batch 90/173] avg loss 5.44461e-05, throughput 5.98794K wps
[Epoch 40 Batch 120/173] avg loss 6.37677e-05, throughput 5.99478K wps
[Epoch 40 Batch 150/173] avg loss 5.10974e-05, throughput 5.9915K wps
Begin Testing...
[Epoch 40] train avg loss 5.61617e-05, test acc 0.7552, test avg loss 1.03924, throughput 6.01833K wps
[Epoch 41 Batch 30/173] avg loss 5.53097e-05, throughput 6.15342K wps
[Epoch 41 Batch 60/173] avg loss 6.99756e-05, throughput 5.9962K wps
[Epoch 41 Batch 90/173] avg loss 5.52421e-05, throughput 5.97946K wps
[Epoch 41 Batch 120/173] avg loss 5.83413e-05, throughput 5.9942K wps
[Epoch 41 Batch 150/173] avg loss 4.97787e-05, throughput 6.01016K wps
Begin Testing...
[Epoch 41] train avg loss 6.05409e-05, test acc 0.7510, test avg loss 1.06542, throughput 6.02373K wps
[Epoch 42 Batch 30/173] avg loss 5.39554e-05, throughput 6.14523K wps
[Epoch 42 Batch 60/173] avg loss 5.66215e-05, throughput 6.01069K wps
[Epoch 42 Batch 90/173] avg loss 4.18528e-05, throughput 5.99507K wps
[Epoch 42 Batch 120/173] avg loss 6.34516e-05, throughput 5.99743K wps
[Epoch 42 Batch 150/173] avg loss 5.96261e-05, throughput 5.98303K wps
Begin Testing...
[Epoch 42] train avg loss 5.67472e-05, test acc 0.7490, test avg loss 1.07504, throughput 6.02094K wps
[Epoch 43 Batch 30/173] avg loss 5.13162e-05, throughput 6.13297K wps
[Epoch 43 Batch 60/173] avg loss 4.18759e-05, throughput 5.99492K wps
[Epoch 43 Batch 90/173] avg loss 3.77806e-05, throughput 5.99105K wps
[Epoch 43 Batch 120/173] avg loss 5.35571e-05, throughput 6.00402K wps
[Epoch 43 Batch 150/173] avg loss 4.16365e-05, throughput 6.00297K wps
Begin Testing...
[Epoch 43] train avg loss 4.52563e-05, test acc 0.7479, test avg loss 1.08781, throughput 6.02047K wps
[Epoch 44 Batch 30/173] avg loss 4.91222e-05, throughput 6.14565K wps
[Epoch 44 Batch 60/173] avg loss 3.40944e-05, throughput 5.98825K wps
[Epoch 44 Batch 90/173] avg loss 4.22108e-05, throughput 6.00447K wps
[Epoch 44 Batch 120/173] avg loss 5.3959e-05, throughput 6.00529K wps
[Epoch 44 Batch 150/173] avg loss 4.56968e-05, throughput 5.99641K wps
Begin Testing...
[Epoch 44] train avg loss 4.3364e-05, test acc 0.7531, test avg loss 1.11416, throughput 6.02324K wps
[Epoch 45 Batch 30/173] avg loss 3.4262e-05, throughput 6.14548K wps
[Epoch 45 Batch 60/173] avg loss 2.98771e-05, throughput 5.99162K wps
[Epoch 45 Batch 90/173] avg loss 2.93336e-05, throughput 6.00303K wps
[Epoch 45 Batch 120/173] avg loss 3.70914e-05, throughput 6.00089K wps
[Epoch 45 Batch 150/173] avg loss 2.85728e-05, throughput 5.99762K wps
Begin Testing...
[Epoch 45] train avg loss 3.22479e-05, test acc 0.7583, test avg loss 1.13807, throughput 6.02337K wps
[Epoch 46 Batch 30/173] avg loss 2.93478e-05, throughput 6.14624K wps
[Epoch 46 Batch 60/173] avg loss 2.57499e-05, throughput 5.99255K wps
[Epoch 46 Batch 90/173] avg loss 3.86407e-05, throughput 5.99777K wps
[Epoch 46 Batch 120/173] avg loss 3.31944e-05, throughput 5.99581K wps
[Epoch 46 Batch 150/173] avg loss 2.8731e-05, throughput 6.0083K wps
Begin Testing...
[Epoch 46] train avg loss 3.12505e-05, test acc 0.7500, test avg loss 1.15765, throughput 6.023K wps
[Epoch 47 Batch 30/173] avg loss 2.50323e-05, throughput 6.14852K wps
[Epoch 47 Batch 60/173] avg loss 2.88475e-05, throughput 5.99764K wps
[Epoch 47 Batch 90/173] avg loss 2.47692e-05, throughput 6.00015K wps
[Epoch 47 Batch 120/173] avg loss 2.37944e-05, throughput 6.00906K wps
[Epoch 47 Batch 150/173] avg loss 2.66182e-05, throughput 6.00174K wps
Begin Testing...
[Epoch 47] train avg loss 2.57701e-05, test acc 0.7500, test avg loss 1.18736, throughput 6.02747K wps
[Epoch 48 Batch 30/173] avg loss 2.48493e-05, throughput 6.14462K wps
[Epoch 48 Batch 60/173] avg loss 2.05705e-05, throughput 5.9908K wps
[Epoch 48 Batch 90/173] avg loss 4.12901e-05, throughput 5.99356K wps
[Epoch 48 Batch 120/173] avg loss 2.87959e-05, throughput 6.00421K wps
[Epoch 48 Batch 150/173] avg loss 3.1123e-05, throughput 5.9903K wps
Begin Testing...
[Epoch 48] train avg loss 2.87047e-05, test acc 0.7552, test avg loss 1.20342, throughput 6.02062K wps
[Epoch 49 Batch 30/173] avg loss 1.98073e-05, throughput 6.14807K wps
[Epoch 49 Batch 60/173] avg loss 2.12328e-05, throughput 5.996K wps
[Epoch 49 Batch 90/173] avg loss 2.86492e-05, throughput 5.99665K wps
[Epoch 49 Batch 120/173] avg loss 2.29091e-05, throughput 6.00341K wps
[Epoch 49 Batch 150/173] avg loss 2.36766e-05, throughput 6.00091K wps
Begin Testing...
[Epoch 49] train avg loss 2.28362e-05, test acc 0.7469, test avg loss 1.22372, throughput 6.02424K wps
[Epoch 50 Batch 30/173] avg loss 1.87766e-05, throughput 6.15748K wps
[Epoch 50 Batch 60/173] avg loss 1.70967e-05, throughput 6.00989K wps
[Epoch 50 Batch 90/173] avg loss 2.2675e-05, throughput 6.0004K wps
[Epoch 50 Batch 120/173] avg loss 1.65067e-05, throughput 6.00089K wps
[Epoch 50 Batch 150/173] avg loss 2.03885e-05, throughput 5.99621K wps
Begin Testing...
[Epoch 50] train avg loss 1.90152e-05, test acc 0.7458, test avg loss 1.25464, throughput 6.02764K wps
[Epoch 51 Batch 30/173] avg loss 1.82274e-05, throughput 6.1368K wps
[Epoch 51 Batch 60/173] avg loss 1.50478e-05, throughput 5.98476K wps
[Epoch 51 Batch 90/173] avg loss 2.33769e-05, throughput 6.00187K wps
[Epoch 51 Batch 120/173] avg loss 2.10004e-05, throughput 5.99417K wps
[Epoch 51 Batch 150/173] avg loss 1.8401e-05, throughput 6.01026K wps
Begin Testing...
[Epoch 51] train avg loss 1.97076e-05, test acc 0.7469, test avg loss 1.26019, throughput 6.02288K wps
[Epoch 52 Batch 30/173] avg loss 1.53671e-05, throughput 6.1549K wps
[Epoch 52 Batch 60/173] avg loss 1.61796e-05, throughput 6.00201K wps
[Epoch 52 Batch 90/173] avg loss 1.29096e-05, throughput 6.01076K wps
[Epoch 52 Batch 120/173] avg loss 1.81184e-05, throughput 6.00151K wps
[Epoch 52 Batch 150/173] avg loss 2.55399e-05, throughput 6.00803K wps
Begin Testing...
[Epoch 52] train avg loss 1.74321e-05, test acc 0.7448, test avg loss 1.28634, throughput 6.03037K wps
[Epoch 53 Batch 30/173] avg loss 1.77664e-05, throughput 6.13191K wps
[Epoch 53 Batch 60/173] avg loss 1.74871e-05, throughput 6.00407K wps
[Epoch 53 Batch 90/173] avg loss 1.69941e-05, throughput 5.99745K wps
[Epoch 53 Batch 120/173] avg loss 2.50372e-05, throughput 6.01074K wps
[Epoch 53 Batch 150/173] avg loss 1.28824e-05, throughput 5.99465K wps
Begin Testing...
[Epoch 53] train avg loss 1.76731e-05, test acc 0.7427, test avg loss 1.31548, throughput 6.0237K wps
[Epoch 54 Batch 30/173] avg loss 1.32124e-05, throughput 6.14147K wps
[Epoch 54 Batch 60/173] avg loss 2.58072e-05, throughput 5.97421K wps
[Epoch 54 Batch 90/173] avg loss 1.80651e-05, throughput 5.98753K wps
[Epoch 54 Batch 120/173] avg loss 1.25057e-05, throughput 5.99893K wps
[Epoch 54 Batch 150/173] avg loss 1.75493e-05, throughput 6.00801K wps
Begin Testing...
[Epoch 54] train avg loss 1.68935e-05, test acc 0.7427, test avg loss 1.34177, throughput 6.02102K wps
[Epoch 55 Batch 30/173] avg loss 1.31346e-05, throughput 6.15341K wps
[Epoch 55 Batch 60/173] avg loss 1.14894e-05, throughput 6.00602K wps
[Epoch 55 Batch 90/173] avg loss 1.18659e-05, throughput 6.00909K wps
[Epoch 55 Batch 120/173] avg loss 1.66762e-05, throughput 6.00526K wps
[Epoch 55 Batch 150/173] avg loss 2.41274e-05, throughput 6.01174K wps
Begin Testing...
[Epoch 55] train avg loss 1.52805e-05, test acc 0.7438, test avg loss 1.35574, throughput 6.03196K wps
[Epoch 56 Batch 30/173] avg loss 1.2089e-05, throughput 6.1441K wps
[Epoch 56 Batch 60/173] avg loss 1.38252e-05, throughput 6.00556K wps
[Epoch 56 Batch 90/173] avg loss 2.08065e-05, throughput 6.00787K wps
[Epoch 56 Batch 120/173] avg loss 1.56018e-05, throughput 6.00984K wps
[Epoch 56 Batch 150/173] avg loss 1.26487e-05, throughput 5.99374K wps
Begin Testing...
[Epoch 56] train avg loss 1.4472e-05, test acc 0.7448, test avg loss 1.37602, throughput 6.0296K wps
[Epoch 57 Batch 30/173] avg loss 9.98869e-06, throughput 6.15015K wps
[Epoch 57 Batch 60/173] avg loss 7.69881e-06, throughput 5.99584K wps
[Epoch 57 Batch 90/173] avg loss 1.11711e-05, throughput 6.00366K wps
[Epoch 57 Batch 120/173] avg loss 1.13544e-05, throughput 6.00702K wps
[Epoch 57 Batch 150/173] avg loss 1.29087e-05, throughput 6.01218K wps
Begin Testing...
[Epoch 57] train avg loss 1.08858e-05, test acc 0.7448, test avg loss 1.40312, throughput 6.03132K wps
[Epoch 58 Batch 30/173] avg loss 7.9552e-06, throughput 6.1457K wps
[Epoch 58 Batch 60/173] avg loss 1.02845e-05, throughput 5.99571K wps
[Epoch 58 Batch 90/173] avg loss 1.31831e-05, throughput 6.0096K wps
[Epoch 58 Batch 120/173] avg loss 1.09326e-05, throughput 6.01616K wps
[Epoch 58 Batch 150/173] avg loss 9.47659e-06, throughput 6.00579K wps
Begin Testing...
[Epoch 58] train avg loss 1.01797e-05, test acc 0.7458, test avg loss 1.41714, throughput 6.03029K wps
[Epoch 59 Batch 30/173] avg loss 1.02076e-05, throughput 6.1619K wps
[Epoch 59 Batch 60/173] avg loss 1.0662e-05, throughput 5.99039K wps
[Epoch 59 Batch 90/173] avg loss 8.84234e-06, throughput 5.99669K wps
[Epoch 59 Batch 120/173] avg loss 1.46742e-05, throughput 5.98591K wps
[Epoch 59 Batch 150/173] avg loss 9.02695e-06, throughput 5.99798K wps
Begin Testing...
[Epoch 59] train avg loss 1.05172e-05, test acc 0.7406, test avg loss 1.43195, throughput 6.02517K wps
Test loss 0.453572, test acc 0.7824
Total time cost 358.25s
9596 1066
[Epoch 0 Batch 30/173] avg loss 0.0153354, throughput 5.79679K wps
[Epoch 0 Batch 60/173] avg loss 0.0153627, throughput 6.00517K wps
[Epoch 0 Batch 90/173] avg loss 0.0147762, throughput 5.99168K wps
[Epoch 0 Batch 120/173] avg loss 0.014077, throughput 5.99978K wps
[Epoch 0 Batch 150/173] avg loss 0.0142338, throughput 6.00248K wps
Begin Testing...
[Epoch 0] train avg loss 0.0147347, test acc 0.5729, test avg loss 0.670592, throughput 5.96639K wps
Observed Improvement.
Begin Testing...
[Epoch 1 Batch 30/173] avg loss 0.0137784, throughput 6.12991K wps
[Epoch 1 Batch 60/173] avg loss 0.0135589, throughput 5.99917K wps
[Epoch 1 Batch 90/173] avg loss 0.0133536, throughput 5.99218K wps
[Epoch 1 Batch 120/173] avg loss 0.0134308, throughput 5.99437K wps
[Epoch 1 Batch 150/173] avg loss 0.0131315, throughput 6.00117K wps
Begin Testing...
[Epoch 1] train avg loss 0.0134141, test acc 0.6365, test avg loss 0.647911, throughput 6.01953K wps
Observed Improvement.
Begin Testing...
[Epoch 2 Batch 30/173] avg loss 0.0127082, throughput 6.16128K wps
[Epoch 2 Batch 60/173] avg loss 0.0126256, throughput 6.00959K wps
[Epoch 2 Batch 90/173] avg loss 0.0124526, throughput 6.01167K wps
[Epoch 2 Batch 120/173] avg loss 0.0124632, throughput 6.00512K wps
[Epoch 2 Batch 150/173] avg loss 0.0124181, throughput 5.9938K wps
Begin Testing...
[Epoch 2] train avg loss 0.0125487, test acc 0.6844, test avg loss 0.620711, throughput 6.03106K wps
Observed Improvement.
Begin Testing...
[Epoch 3 Batch 30/173] avg loss 0.0117835, throughput 6.14552K wps
[Epoch 3 Batch 60/173] avg loss 0.0117168, throughput 6.00731K wps
[Epoch 3 Batch 90/173] avg loss 0.0119264, throughput 6.00711K wps
[Epoch 3 Batch 120/173] avg loss 0.0115274, throughput 6.00559K wps
[Epoch 3 Batch 150/173] avg loss 0.0114979, throughput 5.99726K wps
Begin Testing...
[Epoch 3] train avg loss 0.011674, test acc 0.7167, test avg loss 0.592583, throughput 6.02806K wps
Observed Improvement.
Begin Testing...
[Epoch 4 Batch 30/173] avg loss 0.0108444, throughput 6.1554K wps
[Epoch 4 Batch 60/173] avg loss 0.0108674, throughput 6.00698K wps
[Epoch 4 Batch 90/173] avg loss 0.0107917, throughput 6.01374K wps
[Epoch 4 Batch 120/173] avg loss 0.0106167, throughput 6.00508K wps
[Epoch 4 Batch 150/173] avg loss 0.0105926, throughput 5.98784K wps
Begin Testing...
[Epoch 4] train avg loss 0.0107201, test acc 0.7406, test avg loss 0.563922, throughput 6.02918K wps
Observed Improvement.
Begin Testing...
[Epoch 5 Batch 30/173] avg loss 0.00999394, throughput 6.13923K wps
[Epoch 5 Batch 60/173] avg loss 0.00947725, throughput 6.00459K wps
[Epoch 5 Batch 90/173] avg loss 0.00968434, throughput 5.99069K wps
[Epoch 5 Batch 120/173] avg loss 0.0093867, throughput 6.00062K wps
[Epoch 5 Batch 150/173] avg loss 0.0095758, throughput 5.99846K wps
Begin Testing...
[Epoch 5] train avg loss 0.00957547, test acc 0.7594, test avg loss 0.527989, throughput 6.02196K wps
Observed Improvement.
Begin Testing...
[Epoch 6 Batch 30/173] avg loss 0.00885535, throughput 6.15544K wps
[Epoch 6 Batch 60/173] avg loss 0.00867238, throughput 6.00975K wps
[Epoch 6 Batch 90/173] avg loss 0.00844149, throughput 6.00141K wps
[Epoch 6 Batch 120/173] avg loss 0.00859286, throughput 6.01159K wps
[Epoch 6 Batch 150/173] avg loss 0.00837842, throughput 6.00611K wps
Begin Testing...
[Epoch 6] train avg loss 0.00853772, test acc 0.7740, test avg loss 0.49873, throughput 6.03339K wps
Observed Improvement.
Begin Testing...
[Epoch 7 Batch 30/173] avg loss 0.00734434, throughput 6.1411K wps
[Epoch 7 Batch 60/173] avg loss 0.00789566, throughput 6.00657K wps
[Epoch 7 Batch 90/173] avg loss 0.00721993, throughput 5.99447K wps
[Epoch 7 Batch 120/173] avg loss 0.00760884, throughput 5.99985K wps
[Epoch 7 Batch 150/173] avg loss 0.00729428, throughput 6.01538K wps
Begin Testing...
[Epoch 7] train avg loss 0.00748334, test acc 0.7833, test avg loss 0.473819, throughput 6.02898K wps
Observed Improvement.
Begin Testing...
[Epoch 8 Batch 30/173] avg loss 0.00649516, throughput 6.13652K wps
[Epoch 8 Batch 60/173] avg loss 0.00675793, throughput 6.00513K wps
[Epoch 8 Batch 90/173] avg loss 0.00649066, throughput 5.99342K wps
[Epoch 8 Batch 120/173] avg loss 0.00650672, throughput 5.98649K wps
[Epoch 8 Batch 150/173] avg loss 0.00643657, throughput 5.99795K wps
Begin Testing...
[Epoch 8] train avg loss 0.00654285, test acc 0.7906, test avg loss 0.458402, throughput 6.01955K wps
Observed Improvement.
Begin Testing...
[Epoch 9 Batch 30/173] avg loss 0.00571866, throughput 6.1371K wps
[Epoch 9 Batch 60/173] avg loss 0.00576849, throughput 5.98679K wps
[Epoch 9 Batch 90/173] avg loss 0.00576456, throughput 5.98706K wps
[Epoch 9 Batch 120/173] avg loss 0.0054831, throughput 6.01011K wps
[Epoch 9 Batch 150/173] avg loss 0.00573349, throughput 6.0153K wps
Begin Testing...
[Epoch 9] train avg loss 0.00566799, test acc 0.7896, test avg loss 0.448197, throughput 6.02434K wps
[Epoch 10 Batch 30/173] avg loss 0.00487499, throughput 6.14946K wps
[Epoch 10 Batch 60/173] avg loss 0.00502285, throughput 5.98841K wps
[Epoch 10 Batch 90/173] avg loss 0.00505027, throughput 6.01098K wps
[Epoch 10 Batch 120/173] avg loss 0.00473001, throughput 6.00225K wps
[Epoch 10 Batch 150/173] avg loss 0.00492387, throughput 6.00877K wps
Begin Testing...
[Epoch 10] train avg loss 0.00497974, test acc 0.7833, test avg loss 0.445282, throughput 6.02946K wps
[Epoch 11 Batch 30/173] avg loss 0.00429122, throughput 6.15551K wps
[Epoch 11 Batch 60/173] avg loss 0.00432358, throughput 6.00776K wps
[Epoch 11 Batch 90/173] avg loss 0.00437404, throughput 6.00588K wps
[Epoch 11 Batch 120/173] avg loss 0.00438799, throughput 5.99718K wps
[Epoch 11 Batch 150/173] avg loss 0.00391067, throughput 5.98993K wps
Begin Testing...
[Epoch 11] train avg loss 0.00425563, test acc 0.7844, test avg loss 0.44115, throughput 6.02757K wps
[Epoch 12 Batch 30/173] avg loss 0.00354329, throughput 6.13495K wps
[Epoch 12 Batch 60/173] avg loss 0.00373838, throughput 5.99798K wps
[Epoch 12 Batch 90/173] avg loss 0.00378298, throughput 5.99675K wps
[Epoch 12 Batch 120/173] avg loss 0.00346787, throughput 5.99726K wps
[Epoch 12 Batch 150/173] avg loss 0.0035041, throughput 5.98694K wps
Begin Testing...
[Epoch 12] train avg loss 0.00361211, test acc 0.7823, test avg loss 0.444485, throughput 6.01932K wps
[Epoch 13 Batch 30/173] avg loss 0.00318579, throughput 6.14057K wps
[Epoch 13 Batch 60/173] avg loss 0.00324921, throughput 6.00166K wps
[Epoch 13 Batch 90/173] avg loss 0.00300724, throughput 6.00248K wps
[Epoch 13 Batch 120/173] avg loss 0.00311519, throughput 5.98595K wps
[Epoch 13 Batch 150/173] avg loss 0.00296176, throughput 5.99031K wps
Begin Testing...
[Epoch 13] train avg loss 0.0031152, test acc 0.7812, test avg loss 0.449888, throughput 6.02109K wps
[Epoch 14 Batch 30/173] avg loss 0.00277029, throughput 6.13369K wps
[Epoch 14 Batch 60/173] avg loss 0.00257431, throughput 5.98291K wps
[Epoch 14 Batch 90/173] avg loss 0.00270211, throughput 5.99741K wps
[Epoch 14 Batch 120/173] avg loss 0.00266539, throughput 5.99999K wps
[Epoch 14 Batch 150/173] avg loss 0.00259509, throughput 5.99556K wps
Begin Testing...
[Epoch 14] train avg loss 0.00268362, test acc 0.7844, test avg loss 0.456972, throughput 6.01859K wps
[Epoch 15 Batch 30/173] avg loss 0.00231025, throughput 6.16201K wps
[Epoch 15 Batch 60/173] avg loss 0.00233661, throughput 6.00981K wps
[Epoch 15 Batch 90/173] avg loss 0.00224868, throughput 5.99406K wps
[Epoch 15 Batch 120/173] avg loss 0.00231932, throughput 6.00021K wps
[Epoch 15 Batch 150/173] avg loss 0.00219641, throughput 5.9972K wps
Begin Testing...
[Epoch 15] train avg loss 0.0022754, test acc 0.7760, test avg loss 0.469474, throughput 6.02837K wps
[Epoch 16 Batch 30/173] avg loss 0.00197625, throughput 6.13738K wps
[Epoch 16 Batch 60/173] avg loss 0.00189983, throughput 5.97907K wps
[Epoch 16 Batch 90/173] avg loss 0.00169353, throughput 5.99599K wps
[Epoch 16 Batch 120/173] avg loss 0.00180437, throughput 5.99258K wps
[Epoch 16 Batch 150/173] avg loss 0.00196209, throughput 5.98586K wps
Begin Testing...
[Epoch 16] train avg loss 0.00189799, test acc 0.7760, test avg loss 0.48144, throughput 6.01723K wps
[Epoch 17 Batch 30/173] avg loss 0.00170029, throughput 6.15318K wps
[Epoch 17 Batch 60/173] avg loss 0.00169703, throughput 6.01408K wps
[Epoch 17 Batch 90/173] avg loss 0.00147073, throughput 5.99951K wps
[Epoch 17 Batch 120/173] avg loss 0.00165913, throughput 6.00684K wps
[Epoch 17 Batch 150/173] avg loss 0.00165625, throughput 5.98983K wps
Begin Testing...
[Epoch 17] train avg loss 0.00164497, test acc 0.7729, test avg loss 0.50008, throughput 6.02887K wps
[Epoch 18 Batch 30/173] avg loss 0.001361, throughput 6.15583K wps
[Epoch 18 Batch 60/173] avg loss 0.00125839, throughput 5.97835K wps
[Epoch 18 Batch 90/173] avg loss 0.00137964, throughput 6.0069K wps
[Epoch 18 Batch 120/173] avg loss 0.00134894, throughput 6.00487K wps
[Epoch 18 Batch 150/173] avg loss 0.00145257, throughput 6.00097K wps
Begin Testing...
[Epoch 18] train avg loss 0.0013794, test acc 0.7802, test avg loss 0.51523, throughput 6.02779K wps
[Epoch 19 Batch 30/173] avg loss 0.00123064, throughput 6.14402K wps
[Epoch 19 Batch 60/173] avg loss 0.00127724, throughput 5.99429K wps
[Epoch 19 Batch 90/173] avg loss 0.00113733, throughput 5.99563K wps
[Epoch 19 Batch 120/173] avg loss 0.00108897, throughput 5.99105K wps
[Epoch 19 Batch 150/173] avg loss 0.00131147, throughput 5.98254K wps
Begin Testing...
[Epoch 19] train avg loss 0.00117827, test acc 0.7760, test avg loss 0.532569, throughput 6.01724K wps
[Epoch 20 Batch 30/173] avg loss 0.00091357, throughput 6.12603K wps
[Epoch 20 Batch 60/173] avg loss 0.000970681, throughput 5.98707K wps
[Epoch 20 Batch 90/173] avg loss 0.000909167, throughput 5.99375K wps
[Epoch 20 Batch 120/173] avg loss 0.00110664, throughput 5.99291K wps
[Epoch 20 Batch 150/173] avg loss 0.000928708, throughput 5.99064K wps
Begin Testing...
[Epoch 20] train avg loss 0.000987078, test acc 0.7781, test avg loss 0.551006, throughput 6.01476K wps
[Epoch 21 Batch 30/173] avg loss 0.000812313, throughput 6.15637K wps
[Epoch 21 Batch 60/173] avg loss 0.000904368, throughput 5.99199K wps
[Epoch 21 Batch 90/173] avg loss 0.000810926, throughput 6.0137K wps
[Epoch 21 Batch 120/173] avg loss 0.000775011, throughput 6.01072K wps
[Epoch 21 Batch 150/173] avg loss 0.000934814, throughput 6.0119K wps
Begin Testing...
[Epoch 21] train avg loss 0.000858731, test acc 0.7719, test avg loss 0.574232, throughput 6.03432K wps
[Epoch 22 Batch 30/173] avg loss 0.000667503, throughput 6.14888K wps
[Epoch 22 Batch 60/173] avg loss 0.000766116, throughput 6.00936K wps
[Epoch 22 Batch 90/173] avg loss 0.000749451, throughput 6.00622K wps
[Epoch 22 Batch 120/173] avg loss 0.000844039, throughput 6.00615K wps
[Epoch 22 Batch 150/173] avg loss 0.000650112, throughput 5.99303K wps
Begin Testing...
[Epoch 22] train avg loss 0.000730772, test acc 0.7771, test avg loss 0.594758, throughput 6.02739K wps
[Epoch 23 Batch 30/173] avg loss 0.000568533, throughput 6.14517K wps
[Epoch 23 Batch 60/173] avg loss 0.000524652, throughput 5.99187K wps
[Epoch 23 Batch 90/173] avg loss 0.000593951, throughput 6.0001K wps
[Epoch 23 Batch 120/173] avg loss 0.000719124, throughput 5.99671K wps
[Epoch 23 Batch 150/173] avg loss 0.000529851, throughput 5.99549K wps
Begin Testing...
[Epoch 23] train avg loss 0.000597132, test acc 0.7771, test avg loss 0.611306, throughput 6.0222K wps
[Epoch 24 Batch 30/173] avg loss 0.000539165, throughput 6.1157K wps
[Epoch 24 Batch 60/173] avg loss 0.00049234, throughput 5.9842K wps
[Epoch 24 Batch 90/173] avg loss 0.000540745, throughput 5.98117K wps
[Epoch 24 Batch 120/173] avg loss 0.000554371, throughput 6.00822K wps
[Epoch 24 Batch 150/173] avg loss 0.000526534, throughput 6.00683K wps
Begin Testing...
[Epoch 24] train avg loss 0.000541461, test acc 0.7781, test avg loss 0.634054, throughput 6.0183K wps
[Epoch 25 Batch 30/173] avg loss 0.000442452, throughput 6.14126K wps
[Epoch 25 Batch 60/173] avg loss 0.000426305, throughput 5.99438K wps
[Epoch 25 Batch 90/173] avg loss 0.000405533, throughput 6.00347K wps
[Epoch 25 Batch 120/173] avg loss 0.000441914, throughput 6.00235K wps
[Epoch 25 Batch 150/173] avg loss 0.000450515, throughput 6.00778K wps
Begin Testing...
[Epoch 25] train avg loss 0.000443074, test acc 0.7802, test avg loss 0.653955, throughput 6.02535K wps
[Epoch 26 Batch 30/173] avg loss 0.000374402, throughput 6.14608K wps
[Epoch 26 Batch 60/173] avg loss 0.000360532, throughput 5.9962K wps
[Epoch 26 Batch 90/173] avg loss 0.000373941, throughput 5.99166K wps
[Epoch 26 Batch 120/173] avg loss 0.000408633, throughput 6.00052K wps
[Epoch 26 Batch 150/173] avg loss 0.000362522, throughput 6.0023K wps
Begin Testing...
[Epoch 26] train avg loss 0.000373358, test acc 0.7760, test avg loss 0.680603, throughput 6.02179K wps
[Epoch 27 Batch 30/173] avg loss 0.000339728, throughput 6.13501K wps
[Epoch 27 Batch 60/173] avg loss 0.000316303, throughput 5.99972K wps
[Epoch 27 Batch 90/173] avg loss 0.000342605, throughput 6.0106K wps
[Epoch 27 Batch 120/173] avg loss 0.000280833, throughput 5.99634K wps
[Epoch 27 Batch 150/173] avg loss 0.000355566, throughput 6.00296K wps
Begin Testing...
[Epoch 27] train avg loss 0.000329486, test acc 0.7760, test avg loss 0.70008, throughput 6.02663K wps
[Epoch 28 Batch 30/173] avg loss 0.000312599, throughput 6.15475K wps
[Epoch 28 Batch 60/173] avg loss 0.000305696, throughput 6.00287K wps
[Epoch 28 Batch 90/173] avg loss 0.0002905, throughput 5.99945K wps
[Epoch 28 Batch 120/173] avg loss 0.000291698, throughput 5.99224K wps
[Epoch 28 Batch 150/173] avg loss 0.000291015, throughput 6.00176K wps
Begin Testing...
[Epoch 28] train avg loss 0.000297941, test acc 0.7740, test avg loss 0.72677, throughput 6.02514K wps
[Epoch 29 Batch 30/173] avg loss 0.000231422, throughput 6.13549K wps
[Epoch 29 Batch 60/173] avg loss 0.000250389, throughput 5.9915K wps
[Epoch 29 Batch 90/173] avg loss 0.00029658, throughput 5.99697K wps
[Epoch 29 Batch 120/173] avg loss 0.000223956, throughput 6.00715K wps
[Epoch 29 Batch 150/173] avg loss 0.000268604, throughput 6.00268K wps
Begin Testing...
[Epoch 29] train avg loss 0.000250062, test acc 0.7771, test avg loss 0.746326, throughput 6.0234K wps
[Epoch 30 Batch 30/173] avg loss 0.000208687, throughput 6.14514K wps
[Epoch 30 Batch 60/173] avg loss 0.000228411, throughput 6.00148K wps
[Epoch 30 Batch 90/173] avg loss 0.000205959, throughput 5.9981K wps
[Epoch 30 Batch 120/173] avg loss 0.000218115, throughput 5.98675K wps
[Epoch 30 Batch 150/173] avg loss 0.000216322, throughput 5.99732K wps
Begin Testing...
[Epoch 30] train avg loss 0.000216964, test acc 0.7760, test avg loss 0.764175, throughput 6.02076K wps
[Epoch 31 Batch 30/173] avg loss 0.000200079, throughput 6.14409K wps
[Epoch 31 Batch 60/173] avg loss 0.000178921, throughput 5.99928K wps
[Epoch 31 Batch 90/173] avg loss 0.000183703, throughput 6.00128K wps
[Epoch 31 Batch 120/173] avg loss 0.000178991, throughput 5.99631K wps
[Epoch 31 Batch 150/173] avg loss 0.000184931, throughput 5.99753K wps
Begin Testing...
[Epoch 31] train avg loss 0.000191247, test acc 0.7760, test avg loss 0.78972, throughput 6.02368K wps
[Epoch 32 Batch 30/173] avg loss 0.000168218, throughput 6.14439K wps
[Epoch 32 Batch 60/173] avg loss 0.000160997, throughput 5.99638K wps
[Epoch 32 Batch 90/173] avg loss 0.000150053, throughput 5.99553K wps
[Epoch 32 Batch 120/173] avg loss 0.000162464, throughput 6.00086K wps
[Epoch 32 Batch 150/173] avg loss 0.000175377, throughput 5.99824K wps
Begin Testing...
[Epoch 32] train avg loss 0.000166649, test acc 0.7740, test avg loss 0.803073, throughput 6.02268K wps
[Epoch 33 Batch 30/173] avg loss 0.000136079, throughput 6.14761K wps
[Epoch 33 Batch 60/173] avg loss 0.000165513, throughput 6.00069K wps
[Epoch 33 Batch 90/173] avg loss 0.00014218, throughput 5.9973K wps
[Epoch 33 Batch 120/173] avg loss 0.000134565, throughput 5.99073K wps
[Epoch 33 Batch 150/173] avg loss 0.000138451, throughput 5.99221K wps
Begin Testing...
[Epoch 33] train avg loss 0.000145316, test acc 0.7781, test avg loss 0.821789, throughput 6.02304K wps
[Epoch 34 Batch 30/173] avg loss 0.000128263, throughput 6.15329K wps
[Epoch 34 Batch 60/173] avg loss 0.000116763, throughput 5.99384K wps
[Epoch 34 Batch 90/173] avg loss 0.000140993, throughput 5.98924K wps
[Epoch 34 Batch 120/173] avg loss 0.000158273, throughput 5.98568K wps
[Epoch 34 Batch 150/173] avg loss 0.000124259, throughput 5.99378K wps
Begin Testing...
[Epoch 34] train avg loss 0.000135508, test acc 0.7792, test avg loss 0.846811, throughput 6.0217K wps
[Epoch 35 Batch 30/173] avg loss 0.000112418, throughput 6.13795K wps
[Epoch 35 Batch 60/173] avg loss 0.000116394, throughput 6.00147K wps
[Epoch 35 Batch 90/173] avg loss 0.000132629, throughput 6.00229K wps
[Epoch 35 Batch 120/173] avg loss 0.000119271, throughput 6.00601K wps
[Epoch 35 Batch 150/173] avg loss 0.000136432, throughput 5.99249K wps
Begin Testing...
[Epoch 35] train avg loss 0.00012616, test acc 0.7740, test avg loss 0.868944, throughput 6.02452K wps
[Epoch 36 Batch 30/173] avg loss 9.2104e-05, throughput 6.14698K wps
[Epoch 36 Batch 60/173] avg loss 0.000101166, throughput 5.98787K wps
[Epoch 36 Batch 90/173] avg loss 9.50903e-05, throughput 5.99441K wps
[Epoch 36 Batch 120/173] avg loss 0.000104174, throughput 5.99142K wps
[Epoch 36 Batch 150/173] avg loss 0.000132694, throughput 6.00729K wps
Begin Testing...
[Epoch 36] train avg loss 0.000104849, test acc 0.7729, test avg loss 0.880425, throughput 6.02235K wps
[Epoch 37 Batch 30/173] avg loss 8.94239e-05, throughput 6.13261K wps
[Epoch 37 Batch 60/173] avg loss 0.00010936, throughput 5.99784K wps
[Epoch 37 Batch 90/173] avg loss 9.28438e-05, throughput 6.00067K wps
[Epoch 37 Batch 120/173] avg loss 9.01257e-05, throughput 6.00245K wps
[Epoch 37 Batch 150/173] avg loss 9.75046e-05, throughput 5.99335K wps
Begin Testing...
[Epoch 37] train avg loss 9.45114e-05, test acc 0.7729, test avg loss 0.905819, throughput 6.02154K wps
[Epoch 38 Batch 30/173] avg loss 7.80478e-05, throughput 6.14199K wps
[Epoch 38 Batch 60/173] avg loss 8.60538e-05, throughput 5.99951K wps
[Epoch 38 Batch 90/173] avg loss 7.13798e-05, throughput 6.00293K wps
[Epoch 38 Batch 120/173] avg loss 7.40495e-05, throughput 5.99525K wps
[Epoch 38 Batch 150/173] avg loss 7.27456e-05, throughput 5.99077K wps
Begin Testing...
[Epoch 38] train avg loss 7.79026e-05, test acc 0.7708, test avg loss 0.928196, throughput 6.02208K wps
[Epoch 39 Batch 30/173] avg loss 7.34166e-05, throughput 6.13931K wps
[Epoch 39 Batch 60/173] avg loss 8.06294e-05, throughput 5.99589K wps
[Epoch 39 Batch 90/173] avg loss 6.63203e-05, throughput 5.99784K wps
[Epoch 39 Batch 120/173] avg loss 7.1551e-05, throughput 5.99227K wps
[Epoch 39 Batch 150/173] avg loss 7.61199e-05, throughput 6.0058K wps
Begin Testing...
[Epoch 39] train avg loss 7.63363e-05, test acc 0.7708, test avg loss 0.949292, throughput 6.02282K wps
[Epoch 40 Batch 30/173] avg loss 5.57588e-05, throughput 6.15218K wps
[Epoch 40 Batch 60/173] avg loss 5.3027e-05, throughput 6.00763K wps
[Epoch 40 Batch 90/173] avg loss 6.3667e-05, throughput 5.99828K wps
[Epoch 40 Batch 120/173] avg loss 7.03673e-05, throughput 5.99038K wps
[Epoch 40 Batch 150/173] avg loss 7.0139e-05, throughput 5.98472K wps
Begin Testing...
[Epoch 40] train avg loss 6.40728e-05, test acc 0.7760, test avg loss 0.968992, throughput 6.02366K wps
[Epoch 41 Batch 30/173] avg loss 5.39992e-05, throughput 6.16209K wps
[Epoch 41 Batch 60/173] avg loss 5.88942e-05, throughput 5.99731K wps
[Epoch 41 Batch 90/173] avg loss 4.69739e-05, throughput 6.00832K wps
[Epoch 41 Batch 120/173] avg loss 7.64736e-05, throughput 6.01213K wps
[Epoch 41 Batch 150/173] avg loss 5.58491e-05, throughput 6.01443K wps
Begin Testing...
[Epoch 41] train avg loss 5.80483e-05, test acc 0.7729, test avg loss 0.985062, throughput 6.03573K wps
[Epoch 42 Batch 30/173] avg loss 5.93431e-05, throughput 6.13826K wps
[Epoch 42 Batch 60/173] avg loss 4.13986e-05, throughput 5.99333K wps
[Epoch 42 Batch 90/173] avg loss 4.89707e-05, throughput 6.01844K wps
[Epoch 42 Batch 120/173] avg loss 7.36797e-05, throughput 6.00818K wps
[Epoch 42 Batch 150/173] avg loss 6.44282e-05, throughput 6.00445K wps
Begin Testing...
[Epoch 42] train avg loss 5.75265e-05, test acc 0.7729, test avg loss 1.00926, throughput 6.02996K wps
[Epoch 43 Batch 30/173] avg loss 4.15867e-05, throughput 6.12989K wps
[Epoch 43 Batch 60/173] avg loss 3.79854e-05, throughput 5.99624K wps
[Epoch 43 Batch 90/173] avg loss 4.00248e-05, throughput 6.00734K wps
[Epoch 43 Batch 120/173] avg loss 5.76911e-05, throughput 6.00632K wps
[Epoch 43 Batch 150/173] avg loss 4.7222e-05, throughput 5.98853K wps
Begin Testing...
[Epoch 43] train avg loss 4.55571e-05, test acc 0.7708, test avg loss 1.02566, throughput 6.0243K wps
[Epoch 44 Batch 30/173] avg loss 3.9704e-05, throughput 6.15148K wps
[Epoch 44 Batch 60/173] avg loss 4.04825e-05, throughput 5.99538K wps
[Epoch 44 Batch 90/173] avg loss 4.61163e-05, throughput 5.99978K wps
[Epoch 44 Batch 120/173] avg loss 2.74158e-05, throughput 5.99587K wps
[Epoch 44 Batch 150/173] avg loss 4.60456e-05, throughput 5.9936K wps
Begin Testing...
[Epoch 44] train avg loss 4.16041e-05, test acc 0.7729, test avg loss 1.04835, throughput 6.02486K wps
[Epoch 45 Batch 30/173] avg loss 3.80104e-05, throughput 6.15583K wps
[Epoch 45 Batch 60/173] avg loss 3.39774e-05, throughput 6.0143K wps
[Epoch 45 Batch 90/173] avg loss 4.97982e-05, throughput 5.99612K wps
[Epoch 45 Batch 120/173] avg loss 3.08535e-05, throughput 5.99849K wps
[Epoch 45 Batch 150/173] avg loss 3.73079e-05, throughput 5.99964K wps
Begin Testing...
[Epoch 45] train avg loss 3.77774e-05, test acc 0.7719, test avg loss 1.0748, throughput 6.03002K wps
[Epoch 46 Batch 30/173] avg loss 4.10064e-05, throughput 6.15535K wps
[Epoch 46 Batch 60/173] avg loss 4.4349e-05, throughput 5.99926K wps
[Epoch 46 Batch 90/173] avg loss 3.63932e-05, throughput 6.01169K wps
[Epoch 46 Batch 120/173] avg loss 4.29965e-05, throughput 6.00894K wps
[Epoch 46 Batch 150/173] avg loss 2.84082e-05, throughput 5.98683K wps
Begin Testing...
[Epoch 46] train avg loss 3.77081e-05, test acc 0.7656, test avg loss 1.09316, throughput 6.02748K wps
[Epoch 47 Batch 30/173] avg loss 3.21021e-05, throughput 6.14701K wps
[Epoch 47 Batch 60/173] avg loss 2.55587e-05, throughput 6.00555K wps
[Epoch 47 Batch 90/173] avg loss 3.12334e-05, throughput 6.01119K wps
[Epoch 47 Batch 120/173] avg loss 3.62674e-05, throughput 6.00293K wps
[Epoch 47 Batch 150/173] avg loss 2.96879e-05, throughput 6.00669K wps
Begin Testing...
[Epoch 47] train avg loss 3.13818e-05, test acc 0.7667, test avg loss 1.10736, throughput 6.02836K wps
[Epoch 48 Batch 30/173] avg loss 2.71806e-05, throughput 6.1443K wps
[Epoch 48 Batch 60/173] avg loss 4.49106e-05, throughput 6.01087K wps
[Epoch 48 Batch 90/173] avg loss 3.31586e-05, throughput 6.00552K wps
[Epoch 48 Batch 120/173] avg loss 2.68204e-05, throughput 5.99152K wps
[Epoch 48 Batch 150/173] avg loss 2.86323e-05, throughput 5.99108K wps
Begin Testing...
[Epoch 48] train avg loss 3.29852e-05, test acc 0.7667, test avg loss 1.13024, throughput 6.02472K wps
[Epoch 49 Batch 30/173] avg loss 4.03868e-05, throughput 6.15747K wps
[Epoch 49 Batch 60/173] avg loss 3.69481e-05, throughput 5.99171K wps
[Epoch 49 Batch 90/173] avg loss 2.70976e-05, throughput 6.01147K wps
[Epoch 49 Batch 120/173] avg loss 2.37939e-05, throughput 6.008K wps
[Epoch 49 Batch 150/173] avg loss 2.82344e-05, throughput 5.99591K wps
Begin Testing...
[Epoch 49] train avg loss 3.12246e-05, test acc 0.7719, test avg loss 1.14272, throughput 6.02983K wps
[Epoch 50 Batch 30/173] avg loss 2.80098e-05, throughput 6.15456K wps
[Epoch 50 Batch 60/173] avg loss 2.06576e-05, throughput 5.99164K wps
[Epoch 50 Batch 90/173] avg loss 2.06661e-05, throughput 5.99097K wps
[Epoch 50 Batch 120/173] avg loss 2.03541e-05, throughput 5.98469K wps
[Epoch 50 Batch 150/173] avg loss 3.0701e-05, throughput 5.97971K wps
Begin Testing...
[Epoch 50] train avg loss 2.37462e-05, test acc 0.7719, test avg loss 1.16063, throughput 6.01587K wps
[Epoch 51 Batch 30/173] avg loss 2.10278e-05, throughput 6.15557K wps
[Epoch 51 Batch 60/173] avg loss 1.92615e-05, throughput 5.99032K wps
[Epoch 51 Batch 90/173] avg loss 3.56033e-05, throughput 5.99629K wps
[Epoch 51 Batch 120/173] avg loss 2.22551e-05, throughput 5.99553K wps
[Epoch 51 Batch 150/173] avg loss 2.73428e-05, throughput 5.99554K wps
Begin Testing...
[Epoch 51] train avg loss 2.59241e-05, test acc 0.7635, test avg loss 1.18698, throughput 6.02285K wps
[Epoch 52 Batch 30/173] avg loss 1.59957e-05, throughput 6.14173K wps
[Epoch 52 Batch 60/173] avg loss 1.83607e-05, throughput 5.99561K wps
[Epoch 52 Batch 90/173] avg loss 3.26148e-05, throughput 5.99822K wps
[Epoch 52 Batch 120/173] avg loss 1.73158e-05, throughput 6.00541K wps
[Epoch 52 Batch 150/173] avg loss 2.45239e-05, throughput 6.00177K wps
Begin Testing...
[Epoch 52] train avg loss 2.14598e-05, test acc 0.7615, test avg loss 1.2075, throughput 6.02668K wps
[Epoch 53 Batch 30/173] avg loss 1.68352e-05, throughput 6.1457K wps
[Epoch 53 Batch 60/173] avg loss 1.63646e-05, throughput 6.01132K wps
[Epoch 53 Batch 90/173] avg loss 1.97895e-05, throughput 5.99968K wps
[Epoch 53 Batch 120/173] avg loss 1.8436e-05, throughput 6.01321K wps
[Epoch 53 Batch 150/173] avg loss 1.45676e-05, throughput 6.00141K wps
Begin Testing...
[Epoch 53] train avg loss 1.7984e-05, test acc 0.7667, test avg loss 1.22407, throughput 6.02901K wps
[Epoch 54 Batch 30/173] avg loss 1.78696e-05, throughput 6.14139K wps
[Epoch 54 Batch 60/173] avg loss 1.70987e-05, throughput 5.91595K wps
[Epoch 54 Batch 90/173] avg loss 1.7905e-05, throughput 5.95439K wps
[Epoch 54 Batch 120/173] avg loss 1.67358e-05, throughput 5.86675K wps
[Epoch 54 Batch 150/173] avg loss 1.21335e-05, throughput 5.85921K wps
Begin Testing...
[Epoch 54] train avg loss 1.62448e-05, test acc 0.7667, test avg loss 1.24285, throughput 5.94143K wps
[Epoch 55 Batch 30/173] avg loss 1.46743e-05, throughput 6.0511K wps
[Epoch 55 Batch 60/173] avg loss 1.76306e-05, throughput 5.85383K wps
[Epoch 55 Batch 90/173] avg loss 1.39597e-05, throughput 5.92918K wps
[Epoch 55 Batch 120/173] avg loss 1.75818e-05, throughput 6.00073K wps
[Epoch 55 Batch 150/173] avg loss 1.70747e-05, throughput 6.00173K wps
Begin Testing...
[Epoch 55] train avg loss 1.5519e-05, test acc 0.7625, test avg loss 1.26315, throughput 5.97261K wps
[Epoch 56 Batch 30/173] avg loss 1.27986e-05, throughput 6.14825K wps
[Epoch 56 Batch 60/173] avg loss 1.32292e-05, throughput 5.99312K wps
[Epoch 56 Batch 90/173] avg loss 1.12687e-05, throughput 5.99748K wps
[Epoch 56 Batch 120/173] avg loss 1.59436e-05, throughput 6.00331K wps
[Epoch 56 Batch 150/173] avg loss 1.34174e-05, throughput 6.01064K wps
Begin Testing...
[Epoch 56] train avg loss 1.33104e-05, test acc 0.7656, test avg loss 1.28566, throughput 6.02588K wps
[Epoch 57 Batch 30/173] avg loss 2.21088e-05, throughput 6.14464K wps
[Epoch 57 Batch 60/173] avg loss 1.33819e-05, throughput 5.98826K wps
[Epoch 57 Batch 90/173] avg loss 1.68957e-05, throughput 5.9954K wps
[Epoch 57 Batch 120/173] avg loss 1.29116e-05, throughput 6.00363K wps
[Epoch 57 Batch 150/173] avg loss 1.79819e-05, throughput 5.99558K wps
Begin Testing...
[Epoch 57] train avg loss 1.63389e-05, test acc 0.7635, test avg loss 1.29917, throughput 6.02283K wps
[Epoch 58 Batch 30/173] avg loss 1.18739e-05, throughput 6.15707K wps
[Epoch 58 Batch 60/173] avg loss 1.00753e-05, throughput 5.9996K wps
[Epoch 58 Batch 90/173] avg loss 1.29113e-05, throughput 6.00705K wps
[Epoch 58 Batch 120/173] avg loss 1.235e-05, throughput 6.00022K wps
[Epoch 58 Batch 150/173] avg loss 9.53516e-06, throughput 5.99934K wps
Begin Testing...
[Epoch 58] train avg loss 1.11906e-05, test acc 0.7583, test avg loss 1.32345, throughput 6.02848K wps
[Epoch 59 Batch 30/173] avg loss 9.59881e-06, throughput 6.13847K wps
[Epoch 59 Batch 60/173] avg loss 9.05304e-06, throughput 5.9959K wps
[Epoch 59 Batch 90/173] avg loss 8.59794e-06, throughput 5.99269K wps
[Epoch 59 Batch 120/173] avg loss 1.11228e-05, throughput 6.01012K wps
[Epoch 59 Batch 150/173] avg loss 1.96344e-05, throughput 6.00174K wps
Begin Testing...
[Epoch 59] train avg loss 1.19871e-05, test acc 0.7583, test avg loss 1.32916, throughput 6.02626K wps
Test loss 0.462828, test acc 0.7795
Total time cost 358.16s
9596 1066
[Epoch 0 Batch 30/173] avg loss 0.015169, throughput 5.78134K wps
[Epoch 0 Batch 60/173] avg loss 0.014842, throughput 6.01115K wps
[Epoch 0 Batch 90/173] avg loss 0.0144681, throughput 6.00819K wps
[Epoch 0 Batch 120/173] avg loss 0.0140495, throughput 6.00909K wps
[Epoch 0 Batch 150/173] avg loss 0.0142618, throughput 6.01172K wps
Begin Testing...
[Epoch 0] train avg loss 0.0145531, test acc 0.6094, test avg loss 0.659639, throughput 5.96931K wps
Observed Improvement.
Begin Testing...
[Epoch 1 Batch 30/173] avg loss 0.0136357, throughput 6.16105K wps
[Epoch 1 Batch 60/173] avg loss 0.0135039, throughput 6.00707K wps
[Epoch 1 Batch 90/173] avg loss 0.0134939, throughput 5.98724K wps
[Epoch 1 Batch 120/173] avg loss 0.0133165, throughput 5.99666K wps
[Epoch 1 Batch 150/173] avg loss 0.0134407, throughput 6.00191K wps
Begin Testing...
[Epoch 1] train avg loss 0.0134743, test acc 0.6521, test avg loss 0.642323, throughput 6.02813K wps
Observed Improvement.
Begin Testing...
[Epoch 2 Batch 30/173] avg loss 0.0126954, throughput 6.14807K wps
[Epoch 2 Batch 60/173] avg loss 0.0127181, throughput 5.99061K wps
[Epoch 2 Batch 90/173] avg loss 0.0128176, throughput 5.99565K wps
[Epoch 2 Batch 120/173] avg loss 0.0124859, throughput 5.9964K wps
[Epoch 2 Batch 150/173] avg loss 0.012474, throughput 6.00038K wps
Begin Testing...
[Epoch 2] train avg loss 0.01263, test acc 0.6885, test avg loss 0.625127, throughput 6.02258K wps
Observed Improvement.
Begin Testing...
[Epoch 3 Batch 30/173] avg loss 0.0121653, throughput 6.15378K wps
[Epoch 3 Batch 60/173] avg loss 0.0118527, throughput 5.99865K wps
[Epoch 3 Batch 90/173] avg loss 0.0119504, throughput 6.00619K wps
[Epoch 3 Batch 120/173] avg loss 0.0117814, throughput 6.00141K wps
[Epoch 3 Batch 150/173] avg loss 0.0115323, throughput 6.00739K wps
Begin Testing...
[Epoch 3] train avg loss 0.0118437, test acc 0.7250, test avg loss 0.59582, throughput 6.02933K wps
Observed Improvement.
Begin Testing...
[Epoch 4 Batch 30/173] avg loss 0.0111908, throughput 6.15703K wps
[Epoch 4 Batch 60/173] avg loss 0.0109574, throughput 6.00916K wps
[Epoch 4 Batch 90/173] avg loss 0.0108588, throughput 6.0059K wps
[Epoch 4 Batch 120/173] avg loss 0.0109861, throughput 5.99023K wps
[Epoch 4 Batch 150/173] avg loss 0.0109844, throughput 5.99043K wps
Begin Testing...
[Epoch 4] train avg loss 0.0109581, test acc 0.7583, test avg loss 0.566139, throughput 6.02556K wps
Observed Improvement.
Begin Testing...
[Epoch 5 Batch 30/173] avg loss 0.0101583, throughput 6.13965K wps
[Epoch 5 Batch 60/173] avg loss 0.0101574, throughput 6.00164K wps
[Epoch 5 Batch 90/173] avg loss 0.00990431, throughput 6.00619K wps
[Epoch 5 Batch 120/173] avg loss 0.00974881, throughput 6.00228K wps
[Epoch 5 Batch 150/173] avg loss 0.00961542, throughput 5.9932K wps
Begin Testing...
[Epoch 5] train avg loss 0.00987285, test acc 0.7583, test avg loss 0.533498, throughput 6.02377K wps
Observed Improvement.
Begin Testing...
[Epoch 6 Batch 30/173] avg loss 0.00908685, throughput 6.1436K wps
[Epoch 6 Batch 60/173] avg loss 0.00884394, throughput 5.99733K wps
[Epoch 6 Batch 90/173] avg loss 0.00882944, throughput 6.00246K wps
[Epoch 6 Batch 120/173] avg loss 0.00866544, throughput 6.00026K wps
[Epoch 6 Batch 150/173] avg loss 0.00858896, throughput 5.99238K wps
Begin Testing...
[Epoch 6] train avg loss 0.00878798, test acc 0.7646, test avg loss 0.503047, throughput 6.02323K wps
Observed Improvement.
Begin Testing...
[Epoch 7 Batch 30/173] avg loss 0.00814874, throughput 6.13522K wps
[Epoch 7 Batch 60/173] avg loss 0.00781073, throughput 5.99159K wps
[Epoch 7 Batch 90/173] avg loss 0.00802428, throughput 5.98764K wps
[Epoch 7 Batch 120/173] avg loss 0.00776327, throughput 5.99426K wps
[Epoch 7 Batch 150/173] avg loss 0.00732063, throughput 5.98306K wps
Begin Testing...
[Epoch 7] train avg loss 0.00772045, test acc 0.7865, test avg loss 0.473857, throughput 6.01458K wps
Observed Improvement.
Begin Testing...
[Epoch 8 Batch 30/173] avg loss 0.00697275, throughput 6.14612K wps
[Epoch 8 Batch 60/173] avg loss 0.00681342, throughput 5.99986K wps
[Epoch 8 Batch 90/173] avg loss 0.00679546, throughput 6.00297K wps
[Epoch 8 Batch 120/173] avg loss 0.00689323, throughput 5.99339K wps
[Epoch 8 Batch 150/173] avg loss 0.00654358, throughput 5.98997K wps
Begin Testing...
[Epoch 8] train avg loss 0.00674551, test acc 0.7927, test avg loss 0.45175, throughput 6.02394K wps
Observed Improvement.
Begin Testing...
[Epoch 9 Batch 30/173] avg loss 0.00570047, throughput 6.14559K wps
[Epoch 9 Batch 60/173] avg loss 0.00573984, throughput 5.9895K wps
[Epoch 9 Batch 90/173] avg loss 0.00576402, throughput 5.9928K wps
[Epoch 9 Batch 120/173] avg loss 0.00576919, throughput 6.00017K wps
[Epoch 9 Batch 150/173] avg loss 0.00616438, throughput 6.00357K wps
Begin Testing...
[Epoch 9] train avg loss 0.00586389, test acc 0.8031, test avg loss 0.437272, throughput 6.02393K wps
Observed Improvement.
Begin Testing...
[Epoch 10 Batch 30/173] avg loss 0.00509939, throughput 6.15354K wps
[Epoch 10 Batch 60/173] avg loss 0.00516839, throughput 6.00462K wps
[Epoch 10 Batch 90/173] avg loss 0.00492777, throughput 6.00625K wps
[Epoch 10 Batch 120/173] avg loss 0.004828, throughput 6.00707K wps
[Epoch 10 Batch 150/173] avg loss 0.00519371, throughput 6.0067K wps
Begin Testing...
[Epoch 10] train avg loss 0.00509517, test acc 0.7958, test avg loss 0.436826, throughput 6.03151K wps
[Epoch 11 Batch 30/173] avg loss 0.0047047, throughput 6.16279K wps
[Epoch 11 Batch 60/173] avg loss 0.00418993, throughput 6.01385K wps
[Epoch 11 Batch 90/173] avg loss 0.00473038, throughput 6.01108K wps
[Epoch 11 Batch 120/173] avg loss 0.00438295, throughput 5.99675K wps
[Epoch 11 Batch 150/173] avg loss 0.00437284, throughput 5.99909K wps
Begin Testing...
[Epoch 11] train avg loss 0.0044535, test acc 0.8073, test avg loss 0.420914, throughput 6.033K wps
Observed Improvement.
Begin Testing...
[Epoch 12 Batch 30/173] avg loss 0.00362047, throughput 6.14244K wps
[Epoch 12 Batch 60/173] avg loss 0.00376634, throughput 6.00687K wps
[Epoch 12 Batch 90/173] avg loss 0.00372827, throughput 5.99329K wps
[Epoch 12 Batch 120/173] avg loss 0.00395105, throughput 6.0169K wps
[Epoch 12 Batch 150/173] avg loss 0.00369102, throughput 6.01531K wps
Begin Testing...
[Epoch 12] train avg loss 0.00374848, test acc 0.8042, test avg loss 0.427188, throughput 6.03006K wps
[Epoch 13 Batch 30/173] avg loss 0.00326723, throughput 6.1527K wps
[Epoch 13 Batch 60/173] avg loss 0.00344213, throughput 6.0008K wps
[Epoch 13 Batch 90/173] avg loss 0.00316624, throughput 6.00986K wps
[Epoch 13 Batch 120/173] avg loss 0.00309078, throughput 6.00937K wps
[Epoch 13 Batch 150/173] avg loss 0.00313505, throughput 6.01423K wps
Begin Testing...
[Epoch 13] train avg loss 0.00324498, test acc 0.8042, test avg loss 0.427347, throughput 6.03179K wps
[Epoch 14 Batch 30/173] avg loss 0.00282503, throughput 6.13197K wps
[Epoch 14 Batch 60/173] avg loss 0.00274889, throughput 5.98749K wps
[Epoch 14 Batch 90/173] avg loss 0.00278764, throughput 5.99221K wps
[Epoch 14 Batch 120/173] avg loss 0.00274626, throughput 6.01294K wps
[Epoch 14 Batch 150/173] avg loss 0.00287263, throughput 6.00218K wps
Begin Testing...
[Epoch 14] train avg loss 0.0027965, test acc 0.7958, test avg loss 0.42731, throughput 6.02264K wps
[Epoch 15 Batch 30/173] avg loss 0.00256759, throughput 6.14208K wps
[Epoch 15 Batch 60/173] avg loss 0.00256909, throughput 6.00739K wps
[Epoch 15 Batch 90/173] avg loss 0.00231214, throughput 6.009K wps
[Epoch 15 Batch 120/173] avg loss 0.00243076, throughput 5.98764K wps
[Epoch 15 Batch 150/173] avg loss 0.00255575, throughput 5.99581K wps
Begin Testing...
[Epoch 15] train avg loss 0.00243827, test acc 0.7896, test avg loss 0.434646, throughput 6.02685K wps
[Epoch 16 Batch 30/173] avg loss 0.00191605, throughput 6.14785K wps
[Epoch 16 Batch 60/173] avg loss 0.00211161, throughput 5.99486K wps
[Epoch 16 Batch 90/173] avg loss 0.0020576, throughput 5.99769K wps
[Epoch 16 Batch 120/173] avg loss 0.0020451, throughput 5.99552K wps
[Epoch 16 Batch 150/173] avg loss 0.00206802, throughput 5.99583K wps
Begin Testing...
[Epoch 16] train avg loss 0.00203538, test acc 0.7937, test avg loss 0.444687, throughput 6.02324K wps
[Epoch 17 Batch 30/173] avg loss 0.00177648, throughput 6.13882K wps
[Epoch 17 Batch 60/173] avg loss 0.00173627, throughput 6.00125K wps
[Epoch 17 Batch 90/173] avg loss 0.0016201, throughput 6.01269K wps
[Epoch 17 Batch 120/173] avg loss 0.00172173, throughput 6.00332K wps
[Epoch 17 Batch 150/173] avg loss 0.00174117, throughput 6.01115K wps
Begin Testing...
[Epoch 17] train avg loss 0.00172709, test acc 0.7948, test avg loss 0.457112, throughput 6.02822K wps
[Epoch 18 Batch 30/173] avg loss 0.00145907, throughput 6.14442K wps
[Epoch 18 Batch 60/173] avg loss 0.00142904, throughput 6.01065K wps
[Epoch 18 Batch 90/173] avg loss 0.00141331, throughput 6.00543K wps
[Epoch 18 Batch 120/173] avg loss 0.00148162, throughput 6.00572K wps
[Epoch 18 Batch 150/173] avg loss 0.00138535, throughput 5.98759K wps
Begin Testing...
[Epoch 18] train avg loss 0.00145966, test acc 0.7969, test avg loss 0.472575, throughput 6.02604K wps
[Epoch 19 Batch 30/173] avg loss 0.00125627, throughput 6.15327K wps
[Epoch 19 Batch 60/173] avg loss 0.00139444, throughput 6.00257K wps
[Epoch 19 Batch 90/173] avg loss 0.00129248, throughput 6.00102K wps
[Epoch 19 Batch 120/173] avg loss 0.00125057, throughput 5.99755K wps
[Epoch 19 Batch 150/173] avg loss 0.00122064, throughput 6.01317K wps
Begin Testing...
[Epoch 19] train avg loss 0.00128334, test acc 0.7896, test avg loss 0.484616, throughput 6.03202K wps
[Epoch 20 Batch 30/173] avg loss 0.0011189, throughput 6.15283K wps
[Epoch 20 Batch 60/173] avg loss 0.00107881, throughput 6.01925K wps
[Epoch 20 Batch 90/173] avg loss 0.00108504, throughput 6.00606K wps
[Epoch 20 Batch 120/173] avg loss 0.00103944, throughput 5.98822K wps
[Epoch 20 Batch 150/173] avg loss 0.00119734, throughput 6.00868K wps
Begin Testing...
[Epoch 20] train avg loss 0.00112025, test acc 0.7865, test avg loss 0.501293, throughput 6.03005K wps
[Epoch 21 Batch 30/173] avg loss 0.000791621, throughput 6.14821K wps
[Epoch 21 Batch 60/173] avg loss 0.000962991, throughput 6.00257K wps
[Epoch 21 Batch 90/173] avg loss 0.000953481, throughput 6.00052K wps
[Epoch 21 Batch 120/173] avg loss 0.000897111, throughput 5.99839K wps
[Epoch 21 Batch 150/173] avg loss 0.00101737, throughput 6.00412K wps
Begin Testing...
[Epoch 21] train avg loss 0.000916331, test acc 0.7885, test avg loss 0.516077, throughput 6.02468K wps
[Epoch 22 Batch 30/173] avg loss 0.000827556, throughput 6.139K wps
[Epoch 22 Batch 60/173] avg loss 0.000709911, throughput 6.01062K wps
[Epoch 22 Batch 90/173] avg loss 0.000712523, throughput 6.00924K wps
[Epoch 22 Batch 120/173] avg loss 0.00080589, throughput 6.01754K wps
[Epoch 22 Batch 150/173] avg loss 0.00079419, throughput 6.0076K wps
Begin Testing...
[Epoch 22] train avg loss 0.000794133, test acc 0.7875, test avg loss 0.530679, throughput 6.03043K wps
[Epoch 23 Batch 30/173] avg loss 0.000695815, throughput 6.14977K wps
[Epoch 23 Batch 60/173] avg loss 0.000613045, throughput 5.99623K wps
[Epoch 23 Batch 90/173] avg loss 0.000657239, throughput 6.00213K wps
[Epoch 23 Batch 120/173] avg loss 0.000744669, throughput 5.99687K wps
[Epoch 23 Batch 150/173] avg loss 0.000688491, throughput 6.00187K wps
Begin Testing...
[Epoch 23] train avg loss 0.000684847, test acc 0.7854, test avg loss 0.549299, throughput 6.0266K wps
[Epoch 24 Batch 30/173] avg loss 0.000542829, throughput 6.133K wps
[Epoch 24 Batch 60/173] avg loss 0.00064139, throughput 5.99907K wps
[Epoch 24 Batch 90/173] avg loss 0.000587257, throughput 5.97087K wps
[Epoch 24 Batch 120/173] avg loss 0.000523314, throughput 5.97706K wps
[Epoch 24 Batch 150/173] avg loss 0.000628894, throughput 6.00365K wps
Begin Testing...
[Epoch 24] train avg loss 0.000577845, test acc 0.7812, test avg loss 0.567442, throughput 6.01443K wps
[Epoch 25 Batch 30/173] avg loss 0.000466899, throughput 6.14477K wps
[Epoch 25 Batch 60/173] avg loss 0.000501638, throughput 6.01294K wps
[Epoch 25 Batch 90/173] avg loss 0.000495771, throughput 6.00585K wps
[Epoch 25 Batch 120/173] avg loss 0.000476385, throughput 6.01571K wps
[Epoch 25 Batch 150/173] avg loss 0.000459747, throughput 5.99988K wps
Begin Testing...
[Epoch 25] train avg loss 0.000487833, test acc 0.7833, test avg loss 0.583345, throughput 6.03237K wps
[Epoch 26 Batch 30/173] avg loss 0.000462726, throughput 6.15144K wps
[Epoch 26 Batch 60/173] avg loss 0.000424616, throughput 6.00736K wps
[Epoch 26 Batch 90/173] avg loss 0.000397545, throughput 6.00193K wps
[Epoch 26 Batch 120/173] avg loss 0.000447897, throughput 6.01664K wps
[Epoch 26 Batch 150/173] avg loss 0.000431319, throughput 6.01356K wps
Begin Testing...
[Epoch 26] train avg loss 0.000429742, test acc 0.7833, test avg loss 0.601987, throughput 6.03268K wps
[Epoch 27 Batch 30/173] avg loss 0.000334879, throughput 6.14981K wps
[Epoch 27 Batch 60/173] avg loss 0.000351679, throughput 6.00792K wps
[Epoch 27 Batch 90/173] avg loss 0.000371519, throughput 6.00305K wps
[Epoch 27 Batch 120/173] avg loss 0.000353132, throughput 5.98249K wps
[Epoch 27 Batch 150/173] avg loss 0.000442859, throughput 6.00889K wps
Begin Testing...
[Epoch 27] train avg loss 0.000374317, test acc 0.7823, test avg loss 0.622108, throughput 6.02671K wps
[Epoch 28 Batch 30/173] avg loss 0.000338463, throughput 6.15615K wps
[Epoch 28 Batch 60/173] avg loss 0.000305052, throughput 5.98449K wps
[Epoch 28 Batch 90/173] avg loss 0.000324718, throughput 6.00814K wps
[Epoch 28 Batch 120/173] avg loss 0.000367342, throughput 6.0102K wps
[Epoch 28 Batch 150/173] avg loss 0.000333206, throughput 6.01641K wps
Begin Testing...
[Epoch 28] train avg loss 0.000334712, test acc 0.7812, test avg loss 0.639182, throughput 6.03229K wps
[Epoch 29 Batch 30/173] avg loss 0.000295942, throughput 6.15858K wps
[Epoch 29 Batch 60/173] avg loss 0.000278299, throughput 6.00468K wps
[Epoch 29 Batch 90/173] avg loss 0.000261502, throughput 6.01113K wps
[Epoch 29 Batch 120/173] avg loss 0.000290397, throughput 5.99824K wps
[Epoch 29 Batch 150/173] avg loss 0.000303835, throughput 5.98449K wps
Begin Testing...
[Epoch 29] train avg loss 0.000289893, test acc 0.7740, test avg loss 0.657084, throughput 6.02422K wps
[Epoch 30 Batch 30/173] avg loss 0.000266553, throughput 6.13623K wps
[Epoch 30 Batch 60/173] avg loss 0.000249897, throughput 5.99506K wps
[Epoch 30 Batch 90/173] avg loss 0.000243489, throughput 6.00141K wps
[Epoch 30 Batch 120/173] avg loss 0.000235786, throughput 5.99744K wps
[Epoch 30 Batch 150/173] avg loss 0.000249923, throughput 5.98175K wps
Begin Testing...
[Epoch 30] train avg loss 0.000244431, test acc 0.7729, test avg loss 0.675238, throughput 6.01834K wps
[Epoch 31 Batch 30/173] avg loss 0.000178664, throughput 6.14936K wps
[Epoch 31 Batch 60/173] avg loss 0.000200781, throughput 5.99688K wps
[Epoch 31 Batch 90/173] avg loss 0.000214134, throughput 6.00265K wps
[Epoch 31 Batch 120/173] avg loss 0.000222012, throughput 6.01269K wps
[Epoch 31 Batch 150/173] avg loss 0.000207043, throughput 6.00321K wps
Begin Testing...
[Epoch 31] train avg loss 0.000214366, test acc 0.7771, test avg loss 0.695397, throughput 6.02895K wps
[Epoch 32 Batch 30/173] avg loss 0.000198058, throughput 6.14278K wps
[Epoch 32 Batch 60/173] avg loss 0.000219523, throughput 5.99214K wps
[Epoch 32 Batch 90/173] avg loss 0.000212979, throughput 5.99403K wps
[Epoch 32 Batch 120/173] avg loss 0.000186436, throughput 5.9897K wps
[Epoch 32 Batch 150/173] avg loss 0.000195213, throughput 5.99205K wps
Begin Testing...
[Epoch 32] train avg loss 0.000201556, test acc 0.7740, test avg loss 0.712705, throughput 6.01851K wps
[Epoch 33 Batch 30/173] avg loss 0.00015287, throughput 6.14783K wps
[Epoch 33 Batch 60/173] avg loss 0.000136974, throughput 5.99755K wps
[Epoch 33 Batch 90/173] avg loss 0.000159466, throughput 5.9922K wps
[Epoch 33 Batch 120/173] avg loss 0.000153129, throughput 6.00028K wps
[Epoch 33 Batch 150/173] avg loss 0.00016036, throughput 5.992K wps
Begin Testing...
[Epoch 33] train avg loss 0.000154532, test acc 0.7677, test avg loss 0.73581, throughput 6.02232K wps
[Epoch 34 Batch 30/173] avg loss 0.000128704, throughput 6.1438K wps
[Epoch 34 Batch 60/173] avg loss 0.00014611, throughput 5.99699K wps
[Epoch 34 Batch 90/173] avg loss 0.00014464, throughput 5.98831K wps
[Epoch 34 Batch 120/173] avg loss 0.000132015, throughput 5.99451K wps
[Epoch 34 Batch 150/173] avg loss 0.000133715, throughput 5.98735K wps
Begin Testing...
[Epoch 34] train avg loss 0.000139178, test acc 0.7729, test avg loss 0.751733, throughput 6.0183K wps
[Epoch 35 Batch 30/173] avg loss 0.000113868, throughput 6.13353K wps
[Epoch 35 Batch 60/173] avg loss 0.000125035, throughput 5.99377K wps
[Epoch 35 Batch 90/173] avg loss 0.000127877, throughput 6.00003K wps
[Epoch 35 Batch 120/173] avg loss 0.000127546, throughput 6.007K wps
[Epoch 35 Batch 150/173] avg loss 0.000125392, throughput 6.00939K wps
Begin Testing...
[Epoch 35] train avg loss 0.000125358, test acc 0.7708, test avg loss 0.775359, throughput 6.02376K wps
[Epoch 36 Batch 30/173] avg loss 0.000124916, throughput 6.14066K wps
[Epoch 36 Batch 60/173] avg loss 9.64821e-05, throughput 5.99343K wps
[Epoch 36 Batch 90/173] avg loss 0.000130043, throughput 6.00166K wps
[Epoch 36 Batch 120/173] avg loss 0.000136346, throughput 5.99623K wps
[Epoch 36 Batch 150/173] avg loss 0.000105209, throughput 6.00426K wps
Begin Testing...
[Epoch 36] train avg loss 0.000115809, test acc 0.7698, test avg loss 0.791425, throughput 6.02402K wps
[Epoch 37 Batch 30/173] avg loss 8.92276e-05, throughput 6.14874K wps
[Epoch 37 Batch 60/173] avg loss 8.79161e-05, throughput 5.99509K wps
[Epoch 37 Batch 90/173] avg loss 0.000101588, throughput 5.99214K wps
[Epoch 37 Batch 120/173] avg loss 0.000103989, throughput 5.97987K wps
[Epoch 37 Batch 150/173] avg loss 0.000102997, throughput 5.99422K wps
Begin Testing...
[Epoch 37] train avg loss 9.92596e-05, test acc 0.7688, test avg loss 0.803628, throughput 6.01886K wps
[Epoch 38 Batch 30/173] avg loss 7.557e-05, throughput 6.15927K wps
[Epoch 38 Batch 60/173] avg loss 7.56606e-05, throughput 6.00293K wps
[Epoch 38 Batch 90/173] avg loss 7.82834e-05, throughput 6.0009K wps
[Epoch 38 Batch 120/173] avg loss 8.0591e-05, throughput 6.00012K wps
[Epoch 38 Batch 150/173] avg loss 0.000102764, throughput 5.99187K wps
Begin Testing...
[Epoch 38] train avg loss 8.6728e-05, test acc 0.7667, test avg loss 0.822949, throughput 6.02766K wps
[Epoch 39 Batch 30/173] avg loss 7.87873e-05, throughput 6.15138K wps
[Epoch 39 Batch 60/173] avg loss 9.09094e-05, throughput 5.99196K wps
[Epoch 39 Batch 90/173] avg loss 7.99336e-05, throughput 5.99661K wps
[Epoch 39 Batch 120/173] avg loss 8.12586e-05, throughput 6.00638K wps
[Epoch 39 Batch 150/173] avg loss 7.44409e-05, throughput 6.00432K wps
Begin Testing...
[Epoch 39] train avg loss 8.12985e-05, test acc 0.7677, test avg loss 0.84832, throughput 6.02559K wps
[Epoch 40 Batch 30/173] avg loss 7.37104e-05, throughput 6.13697K wps
[Epoch 40 Batch 60/173] avg loss 7.1046e-05, throughput 5.9983K wps
[Epoch 40 Batch 90/173] avg loss 7.03554e-05, throughput 6.00016K wps
[Epoch 40 Batch 120/173] avg loss 7.1874e-05, throughput 6.00142K wps
[Epoch 40 Batch 150/173] avg loss 8.58056e-05, throughput 5.98577K wps
Begin Testing...
[Epoch 40] train avg loss 7.44901e-05, test acc 0.7635, test avg loss 0.870662, throughput 6.02005K wps
[Epoch 41 Batch 30/173] avg loss 6.10645e-05, throughput 6.14879K wps
[Epoch 41 Batch 60/173] avg loss 5.75775e-05, throughput 5.99178K wps
[Epoch 41 Batch 90/173] avg loss 5.37708e-05, throughput 5.98089K wps
[Epoch 41 Batch 120/173] avg loss 7.16451e-05, throughput 6.00307K wps
[Epoch 41 Batch 150/173] avg loss 7.8795e-05, throughput 6.00806K wps
Begin Testing...
[Epoch 41] train avg loss 6.31833e-05, test acc 0.7594, test avg loss 0.891692, throughput 6.02336K wps
[Epoch 42 Batch 30/173] avg loss 4.4854e-05, throughput 6.13743K wps
[Epoch 42 Batch 60/173] avg loss 6.0564e-05, throughput 5.99553K wps
[Epoch 42 Batch 90/173] avg loss 6.17373e-05, throughput 5.99547K wps
[Epoch 42 Batch 120/173] avg loss 5.5056e-05, throughput 5.99181K wps
[Epoch 42 Batch 150/173] avg loss 6.04537e-05, throughput 5.99897K wps
Begin Testing...
[Epoch 42] train avg loss 5.6295e-05, test acc 0.7646, test avg loss 0.903555, throughput 6.01998K wps
[Epoch 43 Batch 30/173] avg loss 4.33731e-05, throughput 6.14503K wps
[Epoch 43 Batch 60/173] avg loss 4.88365e-05, throughput 6.00468K wps
[Epoch 43 Batch 90/173] avg loss 6.21263e-05, throughput 6.01312K wps
[Epoch 43 Batch 120/173] avg loss 5.94481e-05, throughput 5.99093K wps
[Epoch 43 Batch 150/173] avg loss 6.2841e-05, throughput 6.00005K wps
Begin Testing...
[Epoch 43] train avg loss 5.5458e-05, test acc 0.7646, test avg loss 0.921972, throughput 6.02596K wps
[Epoch 44 Batch 30/173] avg loss 6.26188e-05, throughput 6.13047K wps
[Epoch 44 Batch 60/173] avg loss 5.28762e-05, throughput 5.98762K wps
[Epoch 44 Batch 90/173] avg loss 4.62064e-05, throughput 5.99977K wps
[Epoch 44 Batch 120/173] avg loss 6.64998e-05, throughput 5.99504K wps
[Epoch 44 Batch 150/173] avg loss 6.07311e-05, throughput 5.99732K wps
Begin Testing...
[Epoch 44] train avg loss 5.8319e-05, test acc 0.7583, test avg loss 0.942143, throughput 6.02066K wps
[Epoch 45 Batch 30/173] avg loss 3.89616e-05, throughput 6.14655K wps
[Epoch 45 Batch 60/173] avg loss 3.46584e-05, throughput 6.00771K wps
[Epoch 45 Batch 90/173] avg loss 4.46462e-05, throughput 6.00895K wps
[Epoch 45 Batch 120/173] avg loss 3.74461e-05, throughput 6.0015K wps
[Epoch 45 Batch 150/173] avg loss 3.70475e-05, throughput 5.98604K wps
Begin Testing...
[Epoch 45] train avg loss 3.92558e-05, test acc 0.7615, test avg loss 0.962354, throughput 6.02815K wps
[Epoch 46 Batch 30/173] avg loss 3.50525e-05, throughput 6.15541K wps
[Epoch 46 Batch 60/173] avg loss 3.88139e-05, throughput 6.00387K wps
[Epoch 46 Batch 90/173] avg loss 3.82511e-05, throughput 6.00192K wps
[Epoch 46 Batch 120/173] avg loss 2.83142e-05, throughput 6.00239K wps
[Epoch 46 Batch 150/173] avg loss 3.66159e-05, throughput 6.00371K wps
Begin Testing...
[Epoch 46] train avg loss 3.57574e-05, test acc 0.7583, test avg loss 0.971699, throughput 6.02793K wps
[Epoch 47 Batch 30/173] avg loss 2.70219e-05, throughput 6.14406K wps
[Epoch 47 Batch 60/173] avg loss 2.8424e-05, throughput 5.99546K wps
[Epoch 47 Batch 90/173] avg loss 3.24333e-05, throughput 5.99213K wps
[Epoch 47 Batch 120/173] avg loss 5.6839e-05, throughput 5.99664K wps
[Epoch 47 Batch 150/173] avg loss 3.9456e-05, throughput 5.99788K wps
Begin Testing...
[Epoch 47] train avg loss 3.70364e-05, test acc 0.7594, test avg loss 0.988536, throughput 6.0212K wps
[Epoch 48 Batch 30/173] avg loss 3.40627e-05, throughput 6.15166K wps
[Epoch 48 Batch 60/173] avg loss 4.23952e-05, throughput 5.99372K wps
[Epoch 48 Batch 90/173] avg loss 3.31026e-05, throughput 5.99131K wps
[Epoch 48 Batch 120/173] avg loss 3.95316e-05, throughput 6.00258K wps
[Epoch 48 Batch 150/173] avg loss 2.65834e-05, throughput 5.99485K wps
Begin Testing...
[Epoch 48] train avg loss 3.4957e-05, test acc 0.7552, test avg loss 1.00675, throughput 6.02342K wps
[Epoch 49 Batch 30/173] avg loss 2.37868e-05, throughput 6.14602K wps
[Epoch 49 Batch 60/173] avg loss 2.65569e-05, throughput 5.98889K wps
[Epoch 49 Batch 90/173] avg loss 2.44471e-05, throughput 5.97976K wps
[Epoch 49 Batch 120/173] avg loss 2.68041e-05, throughput 5.99825K wps
[Epoch 49 Batch 150/173] avg loss 2.78581e-05, throughput 6.00153K wps
Begin Testing...
[Epoch 49] train avg loss 2.63597e-05, test acc 0.7615, test avg loss 1.02306, throughput 6.01851K wps
[Epoch 50 Batch 30/173] avg loss 2.12135e-05, throughput 6.14447K wps
[Epoch 50 Batch 60/173] avg loss 2.91746e-05, throughput 5.99815K wps
[Epoch 50 Batch 90/173] avg loss 2.1732e-05, throughput 5.99801K wps
[Epoch 50 Batch 120/173] avg loss 2.71689e-05, throughput 5.98386K wps
[Epoch 50 Batch 150/173] avg loss 3.07876e-05, throughput 5.99334K wps
Begin Testing...
[Epoch 50] train avg loss 2.53805e-05, test acc 0.7604, test avg loss 1.03427, throughput 6.01842K wps
[Epoch 51 Batch 30/173] avg loss 1.94197e-05, throughput 6.1466K wps
[Epoch 51 Batch 60/173] avg loss 2.05474e-05, throughput 5.99794K wps
[Epoch 51 Batch 90/173] avg loss 2.4408e-05, throughput 6.00586K wps
[Epoch 51 Batch 120/173] avg loss 2.80302e-05, throughput 5.98742K wps
[Epoch 51 Batch 150/173] avg loss 2.24098e-05, throughput 6.00858K wps
Begin Testing...
[Epoch 51] train avg loss 2.29023e-05, test acc 0.7594, test avg loss 1.04865, throughput 6.02524K wps
[Epoch 52 Batch 30/173] avg loss 2.36766e-05, throughput 6.14308K wps
[Epoch 52 Batch 60/173] avg loss 1.75682e-05, throughput 5.99942K wps
[Epoch 52 Batch 90/173] avg loss 1.91653e-05, throughput 6.01319K wps
[Epoch 52 Batch 120/173] avg loss 2.45446e-05, throughput 6.0051K wps
[Epoch 52 Batch 150/173] avg loss 2.27446e-05, throughput 6.01359K wps
Begin Testing...
[Epoch 52] train avg loss 2.16142e-05, test acc 0.7542, test avg loss 1.06976, throughput 6.0292K wps
[Epoch 53 Batch 30/173] avg loss 1.5806e-05, throughput 6.14816K wps
[Epoch 53 Batch 60/173] avg loss 1.84748e-05, throughput 5.99275K wps
[Epoch 53 Batch 90/173] avg loss 1.4359e-05, throughput 6.01061K wps
[Epoch 53 Batch 120/173] avg loss 1.89819e-05, throughput 6.01409K wps
[Epoch 53 Batch 150/173] avg loss 1.55046e-05, throughput 6.00452K wps
Begin Testing...
[Epoch 53] train avg loss 1.70927e-05, test acc 0.7531, test avg loss 1.09219, throughput 6.02941K wps
[Epoch 54 Batch 30/173] avg loss 1.7004e-05, throughput 6.13673K wps
[Epoch 54 Batch 60/173] avg loss 2.13173e-05, throughput 6.00616K wps
[Epoch 54 Batch 90/173] avg loss 1.82333e-05, throughput 5.99208K wps
[Epoch 54 Batch 120/173] avg loss 1.62296e-05, throughput 5.98869K wps
[Epoch 54 Batch 150/173] avg loss 1.68652e-05, throughput 5.93056K wps
Begin Testing...
[Epoch 54] train avg loss 1.76786e-05, test acc 0.7542, test avg loss 1.11254, throughput 6.00156K wps
[Epoch 55 Batch 30/173] avg loss 1.65262e-05, throughput 6.14312K wps
[Epoch 55 Batch 60/173] avg loss 1.35838e-05, throughput 6.00126K wps
[Epoch 55 Batch 90/173] avg loss 2.00612e-05, throughput 5.99623K wps
[Epoch 55 Batch 120/173] avg loss 1.97041e-05, throughput 5.9798K wps
[Epoch 55 Batch 150/173] avg loss 1.36864e-05, throughput 5.98391K wps
Begin Testing...
[Epoch 55] train avg loss 1.59945e-05, test acc 0.7542, test avg loss 1.12424, throughput 6.01806K wps
[Epoch 56 Batch 30/173] avg loss 1.05274e-05, throughput 6.15043K wps
[Epoch 56 Batch 60/173] avg loss 2.4524e-05, throughput 5.99126K wps
[Epoch 56 Batch 90/173] avg loss 1.49509e-05, throughput 5.98681K wps
[Epoch 56 Batch 120/173] avg loss 1.52456e-05, throughput 5.99012K wps
[Epoch 56 Batch 150/173] avg loss 1.79102e-05, throughput 5.99445K wps
Begin Testing...
[Epoch 56] train avg loss 1.6269e-05, test acc 0.7562, test avg loss 1.14029, throughput 6.01853K wps
[Epoch 57 Batch 30/173] avg loss 1.19815e-05, throughput 6.13866K wps
[Epoch 57 Batch 60/173] avg loss 1.15447e-05, throughput 5.99453K wps
[Epoch 57 Batch 90/173] avg loss 1.48721e-05, throughput 5.98945K wps
[Epoch 57 Batch 120/173] avg loss 1.28869e-05, throughput 5.98028K wps
[Epoch 57 Batch 150/173] avg loss 1.98438e-05, throughput 5.99418K wps
Begin Testing...
[Epoch 57] train avg loss 1.46945e-05, test acc 0.7479, test avg loss 1.16235, throughput 6.01582K wps
[Epoch 58 Batch 30/173] avg loss 9.85775e-06, throughput 6.13752K wps
[Epoch 58 Batch 60/173] avg loss 1.14713e-05, throughput 5.98567K wps
[Epoch 58 Batch 90/173] avg loss 8.75987e-06, throughput 5.99061K wps
[Epoch 58 Batch 120/173] avg loss 1.2503e-05, throughput 6.00622K wps
[Epoch 58 Batch 150/173] avg loss 1.06013e-05, throughput 6.00193K wps
Begin Testing...
[Epoch 58] train avg loss 1.037e-05, test acc 0.7469, test avg loss 1.18493, throughput 6.02137K wps
[Epoch 59 Batch 30/173] avg loss 1.07862e-05, throughput 6.1433K wps
[Epoch 59 Batch 60/173] avg loss 8.28569e-06, throughput 5.99547K wps
[Epoch 59 Batch 90/173] avg loss 1.14595e-05, throughput 6.00542K wps
[Epoch 59 Batch 120/173] avg loss 8.01149e-06, throughput 6.00154K wps
[Epoch 59 Batch 150/173] avg loss 2.04571e-05, throughput 5.99651K wps
Begin Testing...
[Epoch 59] train avg loss 1.15004e-05, test acc 0.7427, test avg loss 1.18707, throughput 6.02525K wps
Test loss 0.42132, test acc 0.7983
Total time cost 358.44s
9596 1066
[Epoch 0 Batch 30/173] avg loss 0.0153216, throughput 5.787K wps
[Epoch 0 Batch 60/173] avg loss 0.0150399, throughput 6.00126K wps
[Epoch 0 Batch 90/173] avg loss 0.0147301, throughput 5.994K wps
[Epoch 0 Batch 120/173] avg loss 0.0146596, throughput 6.00757K wps
[Epoch 0 Batch 150/173] avg loss 0.0143724, throughput 5.99119K wps
Begin Testing...
[Epoch 0] train avg loss 0.0147636, test acc 0.6156, test avg loss 0.663676, throughput 5.96175K wps
Observed Improvement.
Begin Testing...
[Epoch 1 Batch 30/173] avg loss 0.0136168, throughput 6.13721K wps
[Epoch 1 Batch 60/173] avg loss 0.0135029, throughput 5.99816K wps
[Epoch 1 Batch 90/173] avg loss 0.0134187, throughput 5.99237K wps
[Epoch 1 Batch 120/173] avg loss 0.0133511, throughput 6.01755K wps
[Epoch 1 Batch 150/173] avg loss 0.0134438, throughput 5.9986K wps
Begin Testing...
[Epoch 1] train avg loss 0.0134434, test acc 0.6396, test avg loss 0.644786, throughput 6.02389K wps
Observed Improvement.
Begin Testing...
[Epoch 2 Batch 30/173] avg loss 0.0127681, throughput 6.14926K wps
[Epoch 2 Batch 60/173] avg loss 0.0126367, throughput 5.98673K wps
[Epoch 2 Batch 90/173] avg loss 0.0127626, throughput 5.98438K wps
[Epoch 2 Batch 120/173] avg loss 0.0125621, throughput 5.99279K wps
[Epoch 2 Batch 150/173] avg loss 0.0123899, throughput 5.99647K wps
Begin Testing...
[Epoch 2] train avg loss 0.0126051, test acc 0.6771, test avg loss 0.623582, throughput 6.01892K wps
Observed Improvement.
Begin Testing...
[Epoch 3 Batch 30/173] avg loss 0.0118615, throughput 6.14114K wps
[Epoch 3 Batch 60/173] avg loss 0.0117669, throughput 5.99533K wps
[Epoch 3 Batch 90/173] avg loss 0.0118656, throughput 5.98201K wps
[Epoch 3 Batch 120/173] avg loss 0.0116004, throughput 5.99427K wps
[Epoch