Permalink
Find file Copy path
Fetching contributors…
Cannot retrieve contributors at this time
4723 lines (4722 sloc) 282 KB
Namespace(batch_size=50, data_name='MR', dropout=0.5, epochs=60, gpu=0, log_interval=30, lr=0.0001, model_mode='static', save_prefix='sa-model')
Use gpu0
2320
56
Done! Tokenizing Time=0.97s, #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 0.0152094, throughput 5.56208K wps
[Epoch 0 Batch 60/173] avg loss 0.0150597, throughput 13.2439K wps
[Epoch 0 Batch 90/173] avg loss 0.0148112, throughput 13.4026K wps
[Epoch 0 Batch 120/173] avg loss 0.0144108, throughput 13.3786K wps
[Epoch 0 Batch 150/173] avg loss 0.0143735, throughput 13.3697K wps
Begin Testing...
[Epoch 0] train avg loss 0.0147006, test acc 0.5521, test avg loss 0.680266, throughput 9.53156K wps
Observed Improvement.
Begin Testing...
[Epoch 1 Batch 30/173] avg loss 0.0136797, throughput 13.7697K wps
[Epoch 1 Batch 60/173] avg loss 0.0137806, throughput 13.374K wps
[Epoch 1 Batch 90/173] avg loss 0.0138905, throughput 13.4027K wps
[Epoch 1 Batch 120/173] avg loss 0.0134149, throughput 13.3732K wps
[Epoch 1 Batch 150/173] avg loss 0.0135498, throughput 13.2528K wps
Begin Testing...
[Epoch 1] train avg loss 0.0136941, test acc 0.5948, test avg loss 0.660665, throughput 13.4352K wps
Observed Improvement.
Begin Testing...
[Epoch 2 Batch 30/173] avg loss 0.0131915, throughput 13.6748K wps
[Epoch 2 Batch 60/173] avg loss 0.013364, throughput 13.4096K wps
[Epoch 2 Batch 90/173] avg loss 0.0131036, throughput 13.3665K wps
[Epoch 2 Batch 120/173] avg loss 0.0130817, throughput 13.3779K wps
[Epoch 2 Batch 150/173] avg loss 0.0129739, throughput 13.4115K wps
Begin Testing...
[Epoch 2] train avg loss 0.0131109, test acc 0.5854, test avg loss 0.654022, throughput 13.4234K wps
[Epoch 3 Batch 30/173] avg loss 0.0129616, throughput 13.6482K wps
[Epoch 3 Batch 60/173] avg loss 0.0127992, throughput 13.3754K wps
[Epoch 3 Batch 90/173] avg loss 0.0125477, throughput 13.3903K wps
[Epoch 3 Batch 120/173] avg loss 0.012688, throughput 13.2564K wps
[Epoch 3 Batch 150/173] avg loss 0.0125961, throughput 13.3731K wps
Begin Testing...
[Epoch 3] train avg loss 0.0127277, test acc 0.6635, test avg loss 0.632552, throughput 13.403K wps
Observed Improvement.
Begin Testing...
[Epoch 4 Batch 30/173] avg loss 0.0123327, throughput 13.661K wps
[Epoch 4 Batch 60/173] avg loss 0.0122858, throughput 13.2154K wps
[Epoch 4 Batch 90/173] avg loss 0.0123263, throughput 13.3396K wps
[Epoch 4 Batch 120/173] avg loss 0.0121932, throughput 13.3243K wps
[Epoch 4 Batch 150/173] avg loss 0.0124272, throughput 13.1568K wps
Begin Testing...
[Epoch 4] train avg loss 0.012307, test acc 0.6667, test avg loss 0.620546, throughput 13.3483K wps
Observed Improvement.
Begin Testing...
[Epoch 5 Batch 30/173] avg loss 0.0120515, throughput 13.6853K wps
[Epoch 5 Batch 60/173] avg loss 0.01197, throughput 13.2677K wps
[Epoch 5 Batch 90/173] avg loss 0.0117529, throughput 13.337K wps
[Epoch 5 Batch 120/173] avg loss 0.0120664, throughput 13.3683K wps
[Epoch 5 Batch 150/173] avg loss 0.0121069, throughput 13.3465K wps
Begin Testing...
[Epoch 5] train avg loss 0.0119785, test acc 0.6865, test avg loss 0.60496, throughput 13.3986K wps
Observed Improvement.
Begin Testing...
[Epoch 6 Batch 30/173] avg loss 0.0114762, throughput 13.6193K wps
[Epoch 6 Batch 60/173] avg loss 0.0116066, throughput 13.3921K wps
[Epoch 6 Batch 90/173] avg loss 0.0117324, throughput 13.3811K wps
[Epoch 6 Batch 120/173] avg loss 0.0115319, throughput 13.3264K wps
[Epoch 6 Batch 150/173] avg loss 0.0114738, throughput 13.3233K wps
Begin Testing...
[Epoch 6] train avg loss 0.0116124, test acc 0.7333, test avg loss 0.589564, throughput 13.4002K wps
Observed Improvement.
Begin Testing...
[Epoch 7 Batch 30/173] avg loss 0.0112279, throughput 13.532K wps
[Epoch 7 Batch 60/173] avg loss 0.011282, throughput 13.1487K wps
[Epoch 7 Batch 90/173] avg loss 0.011151, throughput 13.3367K wps
[Epoch 7 Batch 120/173] avg loss 0.0111162, throughput 13.353K wps
[Epoch 7 Batch 150/173] avg loss 0.0112236, throughput 13.3388K wps
Begin Testing...
[Epoch 7] train avg loss 0.0112493, test acc 0.6958, test avg loss 0.582975, throughput 13.3427K wps
[Epoch 8 Batch 30/173] avg loss 0.0112028, throughput 13.6175K wps
[Epoch 8 Batch 60/173] avg loss 0.0107719, throughput 13.231K wps
[Epoch 8 Batch 90/173] avg loss 0.0107939, throughput 13.2927K wps
[Epoch 8 Batch 120/173] avg loss 0.0110276, throughput 13.3195K wps
[Epoch 8 Batch 150/173] avg loss 0.01066, throughput 13.2724K wps
Begin Testing...
[Epoch 8] train avg loss 0.0108798, test acc 0.7271, test avg loss 0.565984, throughput 13.3546K wps
[Epoch 9 Batch 30/173] avg loss 0.0106192, throughput 13.6705K wps
[Epoch 9 Batch 60/173] avg loss 0.0103666, throughput 13.1978K wps
[Epoch 9 Batch 90/173] avg loss 0.010545, throughput 13.335K wps
[Epoch 9 Batch 120/173] avg loss 0.010728, throughput 13.2797K wps
[Epoch 9 Batch 150/173] avg loss 0.0103306, throughput 13.3285K wps
Begin Testing...
[Epoch 9] train avg loss 0.0105166, test acc 0.7385, test avg loss 0.549858, throughput 13.3614K wps
Observed Improvement.
Begin Testing...
[Epoch 10 Batch 30/173] avg loss 0.0100677, throughput 13.6412K wps
[Epoch 10 Batch 60/173] avg loss 0.0104342, throughput 13.2288K wps
[Epoch 10 Batch 90/173] avg loss 0.0103987, throughput 13.24K wps
[Epoch 10 Batch 120/173] avg loss 0.0101607, throughput 13.2525K wps
[Epoch 10 Batch 150/173] avg loss 0.0101383, throughput 13.191K wps
Begin Testing...
[Epoch 10] train avg loss 0.0102649, test acc 0.7500, test avg loss 0.536037, throughput 13.3066K wps
Observed Improvement.
Begin Testing...
[Epoch 11 Batch 30/173] avg loss 0.00978033, throughput 13.6187K wps
[Epoch 11 Batch 60/173] avg loss 0.00992172, throughput 13.1837K wps
[Epoch 11 Batch 90/173] avg loss 0.00988534, throughput 13.2487K wps
[Epoch 11 Batch 120/173] avg loss 0.00968766, throughput 13.1649K wps
[Epoch 11 Batch 150/173] avg loss 0.00980587, throughput 13.3415K wps
Begin Testing...
[Epoch 11] train avg loss 0.00982257, test acc 0.7354, test avg loss 0.532595, throughput 13.315K wps
[Epoch 12 Batch 30/173] avg loss 0.00966549, throughput 13.6111K wps
[Epoch 12 Batch 60/173] avg loss 0.00948078, throughput 13.1734K wps
[Epoch 12 Batch 90/173] avg loss 0.00952515, throughput 13.3549K wps
[Epoch 12 Batch 120/173] avg loss 0.00984185, throughput 13.1671K wps
[Epoch 12 Batch 150/173] avg loss 0.00958119, throughput 13.2275K wps
Begin Testing...
[Epoch 12] train avg loss 0.00961166, test acc 0.7604, test avg loss 0.516236, throughput 13.3011K wps
Observed Improvement.
Begin Testing...
[Epoch 13 Batch 30/173] avg loss 0.00944849, throughput 13.5464K wps
[Epoch 13 Batch 60/173] avg loss 0.00918096, throughput 13.2299K wps
[Epoch 13 Batch 90/173] avg loss 0.00917837, throughput 13.2517K wps
[Epoch 13 Batch 120/173] avg loss 0.00950389, throughput 13.2758K wps
[Epoch 13 Batch 150/173] avg loss 0.00919143, throughput 13.1882K wps
Begin Testing...
[Epoch 13] train avg loss 0.00934597, test acc 0.7552, test avg loss 0.509334, throughput 13.2982K wps
[Epoch 14 Batch 30/173] avg loss 0.00887352, throughput 13.6629K wps
[Epoch 14 Batch 60/173] avg loss 0.0092038, throughput 13.1987K wps
[Epoch 14 Batch 90/173] avg loss 0.00922967, throughput 13.1938K wps
[Epoch 14 Batch 120/173] avg loss 0.00870548, throughput 13.2856K wps
[Epoch 14 Batch 150/173] avg loss 0.00954777, throughput 13.1589K wps
Begin Testing...
[Epoch 14] train avg loss 0.0090817, test acc 0.7583, test avg loss 0.499564, throughput 13.2883K wps
[Epoch 15 Batch 30/173] avg loss 0.00892797, throughput 13.5505K wps
[Epoch 15 Batch 60/173] avg loss 0.00900814, throughput 13.1851K wps
[Epoch 15 Batch 90/173] avg loss 0.00898453, throughput 13.1693K wps
[Epoch 15 Batch 120/173] avg loss 0.00889227, throughput 13.1931K wps
[Epoch 15 Batch 150/173] avg loss 0.00875283, throughput 13.1668K wps
Begin Testing...
[Epoch 15] train avg loss 0.00892462, test acc 0.7604, test avg loss 0.496295, throughput 13.2526K wps
Observed Improvement.
Begin Testing...
[Epoch 16 Batch 30/173] avg loss 0.00872271, throughput 13.5785K wps
[Epoch 16 Batch 60/173] avg loss 0.00871608, throughput 13.1758K wps
[Epoch 16 Batch 90/173] avg loss 0.00843031, throughput 12.257K wps
[Epoch 16 Batch 120/173] avg loss 0.00871314, throughput 13.1289K wps
[Epoch 16 Batch 150/173] avg loss 0.00899805, throughput 13.2057K wps
Begin Testing...
[Epoch 16] train avg loss 0.0087068, test acc 0.7583, test avg loss 0.494512, throughput 13.0742K wps
[Epoch 17 Batch 30/173] avg loss 0.0081742, throughput 13.5341K wps
[Epoch 17 Batch 60/173] avg loss 0.00875089, throughput 13.1423K wps
[Epoch 17 Batch 90/173] avg loss 0.00823141, throughput 13.2216K wps
[Epoch 17 Batch 120/173] avg loss 0.0086916, throughput 13.1867K wps
[Epoch 17 Batch 150/173] avg loss 0.00857993, throughput 13.182K wps
Begin Testing...
[Epoch 17] train avg loss 0.00851159, test acc 0.7646, test avg loss 0.486124, throughput 13.2483K wps
Observed Improvement.
Begin Testing...
[Epoch 18 Batch 30/173] avg loss 0.00806357, throughput 13.5217K wps
[Epoch 18 Batch 60/173] avg loss 0.00804267, throughput 13.1804K wps
[Epoch 18 Batch 90/173] avg loss 0.00843838, throughput 13.1856K wps
[Epoch 18 Batch 120/173] avg loss 0.00846664, throughput 13.1717K wps
[Epoch 18 Batch 150/173] avg loss 0.00838712, throughput 13.1912K wps
Begin Testing...
[Epoch 18] train avg loss 0.00828373, test acc 0.7708, test avg loss 0.476769, throughput 13.2442K wps
Observed Improvement.
Begin Testing...
[Epoch 19 Batch 30/173] avg loss 0.00817383, throughput 13.4945K wps
[Epoch 19 Batch 60/173] avg loss 0.00816291, throughput 13.1565K wps
[Epoch 19 Batch 90/173] avg loss 0.00778568, throughput 13.1693K wps
[Epoch 19 Batch 120/173] avg loss 0.00804135, throughput 13.1352K wps
[Epoch 19 Batch 150/173] avg loss 0.00807651, throughput 13.1563K wps
Begin Testing...
[Epoch 19] train avg loss 0.00807475, test acc 0.7677, test avg loss 0.474933, throughput 13.2191K wps
[Epoch 20 Batch 30/173] avg loss 0.00804232, throughput 13.5301K wps
[Epoch 20 Batch 60/173] avg loss 0.00830731, throughput 13.1616K wps
[Epoch 20 Batch 90/173] avg loss 0.00754495, throughput 13.1365K wps
[Epoch 20 Batch 120/173] avg loss 0.00766106, throughput 13.173K wps
[Epoch 20 Batch 150/173] avg loss 0.00817862, throughput 13.1391K wps
Begin Testing...
[Epoch 20] train avg loss 0.00794759, test acc 0.7698, test avg loss 0.472468, throughput 13.217K wps
[Epoch 21 Batch 30/173] avg loss 0.00772597, throughput 13.437K wps
[Epoch 21 Batch 60/173] avg loss 0.00775137, throughput 13.0945K wps
[Epoch 21 Batch 90/173] avg loss 0.00761213, throughput 13.1617K wps
[Epoch 21 Batch 120/173] avg loss 0.0079916, throughput 13.1287K wps
[Epoch 21 Batch 150/173] avg loss 0.00774376, throughput 13.155K wps
Begin Testing...
[Epoch 21] train avg loss 0.00779034, test acc 0.7688, test avg loss 0.470554, throughput 13.1964K wps
[Epoch 22 Batch 30/173] avg loss 0.00749574, throughput 13.4677K wps
[Epoch 22 Batch 60/173] avg loss 0.00797758, throughput 13.1447K wps
[Epoch 22 Batch 90/173] avg loss 0.00756398, throughput 13.1556K wps
[Epoch 22 Batch 120/173] avg loss 0.00750746, throughput 13.1457K wps
[Epoch 22 Batch 150/173] avg loss 0.00785521, throughput 13.158K wps
Begin Testing...
[Epoch 22] train avg loss 0.00767721, test acc 0.7615, test avg loss 0.470854, throughput 13.2092K wps
[Epoch 23 Batch 30/173] avg loss 0.00767624, throughput 13.3769K wps
[Epoch 23 Batch 60/173] avg loss 0.00763617, throughput 13.133K wps
[Epoch 23 Batch 90/173] avg loss 0.00762791, throughput 13.1331K wps
[Epoch 23 Batch 120/173] avg loss 0.00715139, throughput 13.1308K wps
[Epoch 23 Batch 150/173] avg loss 0.00753451, throughput 13.1372K wps
Begin Testing...
[Epoch 23] train avg loss 0.00753824, test acc 0.7760, test avg loss 0.466496, throughput 13.1805K wps
Observed Improvement.
Begin Testing...
[Epoch 24 Batch 30/173] avg loss 0.00742626, throughput 13.517K wps
[Epoch 24 Batch 60/173] avg loss 0.00710272, throughput 13.1083K wps
[Epoch 24 Batch 90/173] avg loss 0.00712334, throughput 13.2076K wps
[Epoch 24 Batch 120/173] avg loss 0.00741861, throughput 13.1804K wps
[Epoch 24 Batch 150/173] avg loss 0.0073984, throughput 13.1854K wps
Begin Testing...
[Epoch 24] train avg loss 0.00736052, test acc 0.7708, test avg loss 0.464407, throughput 13.2404K wps
[Epoch 25 Batch 30/173] avg loss 0.00738116, throughput 13.5097K wps
[Epoch 25 Batch 60/173] avg loss 0.00723069, throughput 13.1077K wps
[Epoch 25 Batch 90/173] avg loss 0.00719341, throughput 13.171K wps
[Epoch 25 Batch 120/173] avg loss 0.00724383, throughput 13.15K wps
[Epoch 25 Batch 150/173] avg loss 0.00728076, throughput 13.1863K wps
Begin Testing...
[Epoch 25] train avg loss 0.00726218, test acc 0.7698, test avg loss 0.464781, throughput 13.2214K wps
[Epoch 26 Batch 30/173] avg loss 0.00675252, throughput 13.3418K wps
[Epoch 26 Batch 60/173] avg loss 0.00701486, throughput 12.8731K wps
[Epoch 26 Batch 90/173] avg loss 0.00677143, throughput 13.0806K wps
[Epoch 26 Batch 120/173] avg loss 0.00754317, throughput 12.9443K wps
[Epoch 26 Batch 150/173] avg loss 0.00715982, throughput 13.1387K wps
Begin Testing...
[Epoch 26] train avg loss 0.00708805, test acc 0.7698, test avg loss 0.462365, throughput 13.0892K wps
[Epoch 27 Batch 30/173] avg loss 0.00672063, throughput 13.4696K wps
[Epoch 27 Batch 60/173] avg loss 0.00696281, throughput 13.1703K wps
[Epoch 27 Batch 90/173] avg loss 0.00706911, throughput 13.1357K wps
[Epoch 27 Batch 120/173] avg loss 0.00681235, throughput 13.1507K wps
[Epoch 27 Batch 150/173] avg loss 0.0070275, throughput 13.172K wps
Begin Testing...
[Epoch 27] train avg loss 0.00688461, test acc 0.7729, test avg loss 0.461687, throughput 13.2087K wps
[Epoch 28 Batch 30/173] avg loss 0.00644825, throughput 13.5344K wps
[Epoch 28 Batch 60/173] avg loss 0.00707149, throughput 13.0609K wps
[Epoch 28 Batch 90/173] avg loss 0.00679793, throughput 13.172K wps
[Epoch 28 Batch 120/173] avg loss 0.00715125, throughput 13.1797K wps
[Epoch 28 Batch 150/173] avg loss 0.00721565, throughput 13.1145K wps
Begin Testing...
[Epoch 28] train avg loss 0.00688344, test acc 0.7698, test avg loss 0.460186, throughput 13.2105K wps
[Epoch 29 Batch 30/173] avg loss 0.00649559, throughput 13.5085K wps
[Epoch 29 Batch 60/173] avg loss 0.00675923, throughput 13.0926K wps
[Epoch 29 Batch 90/173] avg loss 0.00681049, throughput 13.1566K wps
[Epoch 29 Batch 120/173] avg loss 0.0065787, throughput 13.1062K wps
[Epoch 29 Batch 150/173] avg loss 0.00655453, throughput 13.1317K wps
Begin Testing...
[Epoch 29] train avg loss 0.00669423, test acc 0.7677, test avg loss 0.460647, throughput 13.1969K wps
[Epoch 30 Batch 30/173] avg loss 0.0063778, throughput 13.4507K wps
[Epoch 30 Batch 60/173] avg loss 0.00653867, throughput 13.0135K wps
[Epoch 30 Batch 90/173] avg loss 0.00622215, throughput 13.1412K wps
[Epoch 30 Batch 120/173] avg loss 0.0067841, throughput 13.1541K wps
[Epoch 30 Batch 150/173] avg loss 0.0065658, throughput 13.1674K wps
Begin Testing...
[Epoch 30] train avg loss 0.0065106, test acc 0.7760, test avg loss 0.456916, throughput 13.1851K wps
Observed Improvement.
Begin Testing...
[Epoch 31 Batch 30/173] avg loss 0.00667424, throughput 13.4885K wps
[Epoch 31 Batch 60/173] avg loss 0.00656502, throughput 13.0336K wps
[Epoch 31 Batch 90/173] avg loss 0.00622006, throughput 13.1783K wps
[Epoch 31 Batch 120/173] avg loss 0.0063609, throughput 13.1183K wps
[Epoch 31 Batch 150/173] avg loss 0.00605588, throughput 13.186K wps
Begin Testing...
[Epoch 31] train avg loss 0.00641054, test acc 0.7729, test avg loss 0.460772, throughput 13.1846K wps
[Epoch 32 Batch 30/173] avg loss 0.00604315, throughput 13.4603K wps
[Epoch 32 Batch 60/173] avg loss 0.00661153, throughput 13.0467K wps
[Epoch 32 Batch 90/173] avg loss 0.00586644, throughput 13.1354K wps
[Epoch 32 Batch 120/173] avg loss 0.00637962, throughput 13.0744K wps
[Epoch 32 Batch 150/173] avg loss 0.00666063, throughput 13.1183K wps
Begin Testing...
[Epoch 32] train avg loss 0.00631079, test acc 0.7760, test avg loss 0.452979, throughput 13.1652K wps
Observed Improvement.
Begin Testing...
[Epoch 33 Batch 30/173] avg loss 0.00603185, throughput 13.3405K wps
[Epoch 33 Batch 60/173] avg loss 0.00635407, throughput 13.036K wps
[Epoch 33 Batch 90/173] avg loss 0.00594697, throughput 13.0148K wps
[Epoch 33 Batch 120/173] avg loss 0.00637806, throughput 13.0674K wps
[Epoch 33 Batch 150/173] avg loss 0.00635537, throughput 13.067K wps
Begin Testing...
[Epoch 33] train avg loss 0.00616465, test acc 0.7708, test avg loss 0.454812, throughput 13.1035K wps
[Epoch 34 Batch 30/173] avg loss 0.00577034, throughput 13.4034K wps
[Epoch 34 Batch 60/173] avg loss 0.00620247, throughput 13.005K wps
[Epoch 34 Batch 90/173] avg loss 0.0059932, throughput 13.0303K wps
[Epoch 34 Batch 120/173] avg loss 0.0061161, throughput 13.0781K wps
[Epoch 34 Batch 150/173] avg loss 0.00622248, throughput 13.0869K wps
Begin Testing...
[Epoch 34] train avg loss 0.00607323, test acc 0.7802, test avg loss 0.453062, throughput 13.1046K wps
Observed Improvement.
Begin Testing...
[Epoch 35 Batch 30/173] avg loss 0.00592018, throughput 13.3184K wps
[Epoch 35 Batch 60/173] avg loss 0.00582066, throughput 13.0304K wps
[Epoch 35 Batch 90/173] avg loss 0.00586488, throughput 13.1286K wps
[Epoch 35 Batch 120/173] avg loss 0.0058946, throughput 13.1112K wps
[Epoch 35 Batch 150/173] avg loss 0.00634644, throughput 13.0274K wps
Begin Testing...
[Epoch 35] train avg loss 0.00598133, test acc 0.7740, test avg loss 0.453603, throughput 13.1117K wps
[Epoch 36 Batch 30/173] avg loss 0.00560481, throughput 13.4318K wps
[Epoch 36 Batch 60/173] avg loss 0.00551696, throughput 13.0262K wps
[Epoch 36 Batch 90/173] avg loss 0.00589286, throughput 13.0443K wps
[Epoch 36 Batch 120/173] avg loss 0.00596741, throughput 13.0825K wps
[Epoch 36 Batch 150/173] avg loss 0.00586302, throughput 13.1161K wps
Begin Testing...
[Epoch 36] train avg loss 0.00578654, test acc 0.7760, test avg loss 0.452461, throughput 13.1352K wps
[Epoch 37 Batch 30/173] avg loss 0.00566334, throughput 13.4694K wps
[Epoch 37 Batch 60/173] avg loss 0.00546962, throughput 13.0193K wps
[Epoch 37 Batch 90/173] avg loss 0.00592937, throughput 13.0165K wps
[Epoch 37 Batch 120/173] avg loss 0.00588093, throughput 13.0027K wps
[Epoch 37 Batch 150/173] avg loss 0.00594205, throughput 13.1086K wps
Begin Testing...
[Epoch 37] train avg loss 0.00570717, test acc 0.7792, test avg loss 0.45338, throughput 13.1079K wps
[Epoch 38 Batch 30/173] avg loss 0.00563397, throughput 13.3919K wps
[Epoch 38 Batch 60/173] avg loss 0.00538675, throughput 12.97K wps
[Epoch 38 Batch 90/173] avg loss 0.0056808, throughput 13.0156K wps
[Epoch 38 Batch 120/173] avg loss 0.00551007, throughput 13.0322K wps
[Epoch 38 Batch 150/173] avg loss 0.00565788, throughput 13.0687K wps
Begin Testing...
[Epoch 38] train avg loss 0.00554163, test acc 0.7760, test avg loss 0.455132, throughput 13.0899K wps
[Epoch 39 Batch 30/173] avg loss 0.00550745, throughput 13.4595K wps
[Epoch 39 Batch 60/173] avg loss 0.00558637, throughput 13.0183K wps
[Epoch 39 Batch 90/173] avg loss 0.00556495, throughput 13.0209K wps
[Epoch 39 Batch 120/173] avg loss 0.00538889, throughput 13.0544K wps
[Epoch 39 Batch 150/173] avg loss 0.00557051, throughput 13.0319K wps
Begin Testing...
[Epoch 39] train avg loss 0.00553107, test acc 0.7771, test avg loss 0.453784, throughput 13.1171K wps
[Epoch 40 Batch 30/173] avg loss 0.00581644, throughput 13.4288K wps
[Epoch 40 Batch 60/173] avg loss 0.0053967, throughput 13.0509K wps
[Epoch 40 Batch 90/173] avg loss 0.00523756, throughput 13.0569K wps
[Epoch 40 Batch 120/173] avg loss 0.00531302, throughput 13.0035K wps
[Epoch 40 Batch 150/173] avg loss 0.00551931, throughput 13.0279K wps
Begin Testing...
[Epoch 40] train avg loss 0.00545002, test acc 0.7771, test avg loss 0.455652, throughput 13.1014K wps
[Epoch 41 Batch 30/173] avg loss 0.00518146, throughput 13.4236K wps
[Epoch 41 Batch 60/173] avg loss 0.0052459, throughput 12.9852K wps
[Epoch 41 Batch 90/173] avg loss 0.00510528, throughput 13.0019K wps
[Epoch 41 Batch 120/173] avg loss 0.00554049, throughput 13.0003K wps
[Epoch 41 Batch 150/173] avg loss 0.00519943, throughput 13.0095K wps
Begin Testing...
[Epoch 41] train avg loss 0.00529167, test acc 0.7781, test avg loss 0.457808, throughput 13.0767K wps
[Epoch 42 Batch 30/173] avg loss 0.00500092, throughput 13.4316K wps
[Epoch 42 Batch 60/173] avg loss 0.00492734, throughput 13.0028K wps
[Epoch 42 Batch 90/173] avg loss 0.0048954, throughput 13.0271K wps
[Epoch 42 Batch 120/173] avg loss 0.0052777, throughput 13.0739K wps
[Epoch 42 Batch 150/173] avg loss 0.00495792, throughput 13.0389K wps
Begin Testing...
[Epoch 42] train avg loss 0.0051164, test acc 0.7729, test avg loss 0.455225, throughput 13.104K wps
[Epoch 43 Batch 30/173] avg loss 0.00504126, throughput 13.4286K wps
[Epoch 43 Batch 60/173] avg loss 0.00541761, throughput 13.0196K wps
[Epoch 43 Batch 90/173] avg loss 0.0052195, throughput 13.028K wps
[Epoch 43 Batch 120/173] avg loss 0.00509192, throughput 13.0025K wps
[Epoch 43 Batch 150/173] avg loss 0.00476639, throughput 13.0223K wps
Begin Testing...
[Epoch 43] train avg loss 0.00508985, test acc 0.7812, test avg loss 0.455502, throughput 13.0927K wps
Observed Improvement.
Begin Testing...
[Epoch 44 Batch 30/173] avg loss 0.00470727, throughput 13.3463K wps
[Epoch 44 Batch 60/173] avg loss 0.00488579, throughput 12.9621K wps
[Epoch 44 Batch 90/173] avg loss 0.00506715, throughput 12.9833K wps
[Epoch 44 Batch 120/173] avg loss 0.00490351, throughput 12.9977K wps
[Epoch 44 Batch 150/173] avg loss 0.00521203, throughput 12.9616K wps
Begin Testing...
[Epoch 44] train avg loss 0.00495613, test acc 0.7823, test avg loss 0.453426, throughput 13.039K wps
Observed Improvement.
Begin Testing...
[Epoch 45 Batch 30/173] avg loss 0.00483674, throughput 13.412K wps
[Epoch 45 Batch 60/173] avg loss 0.00481195, throughput 12.9956K wps
[Epoch 45 Batch 90/173] avg loss 0.00475305, throughput 12.962K wps
[Epoch 45 Batch 120/173] avg loss 0.00505132, throughput 12.973K wps
[Epoch 45 Batch 150/173] avg loss 0.00464129, throughput 12.9856K wps
Begin Testing...
[Epoch 45] train avg loss 0.00486184, test acc 0.7875, test avg loss 0.452812, throughput 13.0598K wps
Observed Improvement.
Begin Testing...
[Epoch 46 Batch 30/173] avg loss 0.00464891, throughput 13.3538K wps
[Epoch 46 Batch 60/173] avg loss 0.00466078, throughput 12.9208K wps
[Epoch 46 Batch 90/173] avg loss 0.00459316, throughput 12.9487K wps
[Epoch 46 Batch 120/173] avg loss 0.00486154, throughput 12.9874K wps
[Epoch 46 Batch 150/173] avg loss 0.00493791, throughput 12.9621K wps
Begin Testing...
[Epoch 46] train avg loss 0.00474305, test acc 0.7865, test avg loss 0.450764, throughput 13.0332K wps
[Epoch 47 Batch 30/173] avg loss 0.00475091, throughput 13.3296K wps
[Epoch 47 Batch 60/173] avg loss 0.00463307, throughput 13.0021K wps
[Epoch 47 Batch 90/173] avg loss 0.00480157, throughput 12.9906K wps
[Epoch 47 Batch 120/173] avg loss 0.00455245, throughput 13.0576K wps
[Epoch 47 Batch 150/173] avg loss 0.00474014, throughput 12.9347K wps
Begin Testing...
[Epoch 47] train avg loss 0.00468347, test acc 0.7792, test avg loss 0.453091, throughput 13.0558K wps
[Epoch 48 Batch 30/173] avg loss 0.0044363, throughput 13.3636K wps
[Epoch 48 Batch 60/173] avg loss 0.00439622, throughput 13.0142K wps
[Epoch 48 Batch 90/173] avg loss 0.00444731, throughput 13.003K wps
[Epoch 48 Batch 120/173] avg loss 0.00489273, throughput 13.001K wps
[Epoch 48 Batch 150/173] avg loss 0.00455601, throughput 12.9946K wps
Begin Testing...
[Epoch 48] train avg loss 0.00458396, test acc 0.7833, test avg loss 0.453578, throughput 13.0643K wps
[Epoch 49 Batch 30/173] avg loss 0.00450732, throughput 13.3643K wps
[Epoch 49 Batch 60/173] avg loss 0.00464974, throughput 12.8896K wps
[Epoch 49 Batch 90/173] avg loss 0.00442739, throughput 13.0175K wps
[Epoch 49 Batch 120/173] avg loss 0.00445156, throughput 12.9778K wps
[Epoch 49 Batch 150/173] avg loss 0.00435649, throughput 12.9654K wps
Begin Testing...
[Epoch 49] train avg loss 0.0045057, test acc 0.7844, test avg loss 0.451659, throughput 13.0348K wps
[Epoch 50 Batch 30/173] avg loss 0.0041396, throughput 13.3902K wps
[Epoch 50 Batch 60/173] avg loss 0.00460044, throughput 12.9568K wps
[Epoch 50 Batch 90/173] avg loss 0.00446978, throughput 12.9548K wps
[Epoch 50 Batch 120/173] avg loss 0.00448121, throughput 12.9903K wps
[Epoch 50 Batch 150/173] avg loss 0.00421903, throughput 12.9974K wps
Begin Testing...
[Epoch 50] train avg loss 0.00437656, test acc 0.7833, test avg loss 0.450645, throughput 13.0549K wps
[Epoch 51 Batch 30/173] avg loss 0.00448338, throughput 13.3762K wps
[Epoch 51 Batch 60/173] avg loss 0.00435114, throughput 12.843K wps
[Epoch 51 Batch 90/173] avg loss 0.00418208, throughput 12.9832K wps
[Epoch 51 Batch 120/173] avg loss 0.00429336, throughput 12.968K wps
[Epoch 51 Batch 150/173] avg loss 0.00430888, throughput 12.9638K wps
Begin Testing...
[Epoch 51] train avg loss 0.00429195, test acc 0.7875, test avg loss 0.453885, throughput 13.0204K wps
Observed Improvement.
Begin Testing...
[Epoch 52 Batch 30/173] avg loss 0.00417599, throughput 13.398K wps
[Epoch 52 Batch 60/173] avg loss 0.00420422, throughput 13.0163K wps
[Epoch 52 Batch 90/173] avg loss 0.00406156, throughput 13.0011K wps
[Epoch 52 Batch 120/173] avg loss 0.00442677, throughput 12.9633K wps
[Epoch 52 Batch 150/173] avg loss 0.00413953, throughput 12.9468K wps
Begin Testing...
[Epoch 52] train avg loss 0.00420197, test acc 0.7979, test avg loss 0.453806, throughput 13.0553K wps
Observed Improvement.
Begin Testing...
[Epoch 53 Batch 30/173] avg loss 0.00408502, throughput 13.3501K wps
[Epoch 53 Batch 60/173] avg loss 0.00419751, throughput 12.9052K wps
[Epoch 53 Batch 90/173] avg loss 0.00415746, throughput 12.9573K wps
[Epoch 53 Batch 120/173] avg loss 0.00418046, throughput 12.9524K wps
[Epoch 53 Batch 150/173] avg loss 0.00402702, throughput 12.9404K wps
Begin Testing...
[Epoch 53] train avg loss 0.00413677, test acc 0.7865, test avg loss 0.455001, throughput 13.0166K wps
[Epoch 54 Batch 30/173] avg loss 0.00373302, throughput 13.2693K wps
[Epoch 54 Batch 60/173] avg loss 0.0039991, throughput 12.9195K wps
[Epoch 54 Batch 90/173] avg loss 0.00437722, throughput 12.9438K wps
[Epoch 54 Batch 120/173] avg loss 0.00412703, throughput 12.9616K wps
[Epoch 54 Batch 150/173] avg loss 0.0039687, throughput 12.9327K wps
Begin Testing...
[Epoch 54] train avg loss 0.00407304, test acc 0.7823, test avg loss 0.453174, throughput 13.003K wps
[Epoch 55 Batch 30/173] avg loss 0.00395458, throughput 13.3473K wps
[Epoch 55 Batch 60/173] avg loss 0.00394598, throughput 12.8473K wps
[Epoch 55 Batch 90/173] avg loss 0.00410862, throughput 12.9634K wps
[Epoch 55 Batch 120/173] avg loss 0.00382014, throughput 12.9665K wps
[Epoch 55 Batch 150/173] avg loss 0.00366613, throughput 12.9608K wps
Begin Testing...
[Epoch 55] train avg loss 0.00395303, test acc 0.7875, test avg loss 0.454643, throughput 13.0135K wps
[Epoch 56 Batch 30/173] avg loss 0.00392289, throughput 13.2363K wps
[Epoch 56 Batch 60/173] avg loss 0.00395039, throughput 12.8714K wps
[Epoch 56 Batch 90/173] avg loss 0.00386342, throughput 12.9828K wps
[Epoch 56 Batch 120/173] avg loss 0.0035798, throughput 12.9751K wps
[Epoch 56 Batch 150/173] avg loss 0.00412631, throughput 12.9966K wps
Begin Testing...
[Epoch 56] train avg loss 0.00388252, test acc 0.7833, test avg loss 0.456549, throughput 13.0054K wps
[Epoch 57 Batch 30/173] avg loss 0.00369135, throughput 13.2682K wps
[Epoch 57 Batch 60/173] avg loss 0.00373794, throughput 12.9173K wps
[Epoch 57 Batch 90/173] avg loss 0.00378036, throughput 12.9649K wps
[Epoch 57 Batch 120/173] avg loss 0.00349169, throughput 12.9926K wps
[Epoch 57 Batch 150/173] avg loss 0.00363439, throughput 12.9649K wps
Begin Testing...
[Epoch 57] train avg loss 0.00371022, test acc 0.7885, test avg loss 0.456809, throughput 13.0158K wps
[Epoch 58 Batch 30/173] avg loss 0.00361373, throughput 13.3968K wps
[Epoch 58 Batch 60/173] avg loss 0.00340507, throughput 12.9369K wps
[Epoch 58 Batch 90/173] avg loss 0.00367416, throughput 12.9874K wps
[Epoch 58 Batch 120/173] avg loss 0.00402241, throughput 12.9914K wps
[Epoch 58 Batch 150/173] avg loss 0.00389278, throughput 12.969K wps
Begin Testing...
[Epoch 58] train avg loss 0.00369657, test acc 0.7875, test avg loss 0.451559, throughput 13.0467K wps
[Epoch 59 Batch 30/173] avg loss 0.00372463, throughput 13.249K wps
[Epoch 59 Batch 60/173] avg loss 0.0034176, throughput 12.8825K wps
[Epoch 59 Batch 90/173] avg loss 0.00378599, throughput 13.0106K wps
[Epoch 59 Batch 120/173] avg loss 0.00369932, throughput 12.9629K wps
[Epoch 59 Batch 150/173] avg loss 0.00335394, throughput 12.9904K wps
Begin Testing...
[Epoch 59] train avg loss 0.00360944, test acc 0.7927, test avg loss 0.457219, throughput 13.0166K wps
Test loss 0.436836, test acc 0.7992
Total time cost 175.88s
9596 1066
[Epoch 0 Batch 30/173] avg loss 0.0152292, throughput 11.7618K wps
[Epoch 0 Batch 60/173] avg loss 0.015037, throughput 12.8555K wps
[Epoch 0 Batch 90/173] avg loss 0.0148975, throughput 12.9713K wps
[Epoch 0 Batch 120/173] avg loss 0.014298, throughput 12.9844K wps
[Epoch 0 Batch 150/173] avg loss 0.0142362, throughput 12.9996K wps
Begin Testing...
[Epoch 0] train avg loss 0.0146826, test acc 0.5927, test avg loss 0.664683, throughput 12.7301K wps
Observed Improvement.
Begin Testing...
[Epoch 1 Batch 30/173] avg loss 0.0135781, throughput 13.3124K wps
[Epoch 1 Batch 60/173] avg loss 0.0139802, throughput 12.8433K wps
[Epoch 1 Batch 90/173] avg loss 0.0137881, throughput 12.9515K wps
[Epoch 1 Batch 120/173] avg loss 0.013535, throughput 12.9941K wps
[Epoch 1 Batch 150/173] avg loss 0.0135377, throughput 12.9679K wps
Begin Testing...
[Epoch 1] train avg loss 0.0136436, test acc 0.6344, test avg loss 0.649625, throughput 13.0093K wps
Observed Improvement.
Begin Testing...
[Epoch 2 Batch 30/173] avg loss 0.0131905, throughput 13.1452K wps
[Epoch 2 Batch 60/173] avg loss 0.013335, throughput 12.8923K wps
[Epoch 2 Batch 90/173] avg loss 0.0132137, throughput 12.9702K wps
[Epoch 2 Batch 120/173] avg loss 0.0129279, throughput 12.9491K wps
[Epoch 2 Batch 150/173] avg loss 0.0128778, throughput 12.9403K wps
Begin Testing...
[Epoch 2] train avg loss 0.0131084, test acc 0.6448, test avg loss 0.634636, throughput 12.9744K wps
Observed Improvement.
Begin Testing...
[Epoch 3 Batch 30/173] avg loss 0.0126329, throughput 13.1377K wps
[Epoch 3 Batch 60/173] avg loss 0.0127756, throughput 12.8326K wps
[Epoch 3 Batch 90/173] avg loss 0.0126299, throughput 12.9469K wps
[Epoch 3 Batch 120/173] avg loss 0.0124594, throughput 12.9915K wps
[Epoch 3 Batch 150/173] avg loss 0.0124217, throughput 12.9088K wps
Begin Testing...
[Epoch 3] train avg loss 0.0125917, test acc 0.6417, test avg loss 0.625238, throughput 12.9663K wps
[Epoch 4 Batch 30/173] avg loss 0.0122383, throughput 13.2896K wps
[Epoch 4 Batch 60/173] avg loss 0.0124238, throughput 12.8275K wps
[Epoch 4 Batch 90/173] avg loss 0.0122336, throughput 12.9589K wps
[Epoch 4 Batch 120/173] avg loss 0.0123432, throughput 12.9465K wps
[Epoch 4 Batch 150/173] avg loss 0.0122077, throughput 12.9692K wps
Begin Testing...
[Epoch 4] train avg loss 0.0122781, test acc 0.6802, test avg loss 0.608354, throughput 12.9936K wps
Observed Improvement.
Begin Testing...
[Epoch 5 Batch 30/173] avg loss 0.0122782, throughput 13.2217K wps
[Epoch 5 Batch 60/173] avg loss 0.0121302, throughput 12.886K wps
[Epoch 5 Batch 90/173] avg loss 0.0119926, throughput 12.9185K wps
[Epoch 5 Batch 120/173] avg loss 0.0118433, throughput 12.9837K wps
[Epoch 5 Batch 150/173] avg loss 0.011832, throughput 12.965K wps
Begin Testing...
[Epoch 5] train avg loss 0.0119814, test acc 0.6844, test avg loss 0.595843, throughput 12.9914K wps
Observed Improvement.
Begin Testing...
[Epoch 6 Batch 30/173] avg loss 0.0114532, throughput 13.1968K wps
[Epoch 6 Batch 60/173] avg loss 0.0117341, throughput 12.7937K wps
[Epoch 6 Batch 90/173] avg loss 0.0115126, throughput 12.9577K wps
[Epoch 6 Batch 120/173] avg loss 0.0118253, throughput 12.9275K wps
[Epoch 6 Batch 150/173] avg loss 0.0113511, throughput 12.9223K wps
Begin Testing...
[Epoch 6] train avg loss 0.0115779, test acc 0.7156, test avg loss 0.583671, throughput 12.9598K wps
Observed Improvement.
Begin Testing...
[Epoch 7 Batch 30/173] avg loss 0.0112032, throughput 13.3044K wps
[Epoch 7 Batch 60/173] avg loss 0.0113478, throughput 12.8456K wps
[Epoch 7 Batch 90/173] avg loss 0.011386, throughput 12.9614K wps
[Epoch 7 Batch 120/173] avg loss 0.0113221, throughput 12.9125K wps
[Epoch 7 Batch 150/173] avg loss 0.0113154, throughput 12.9351K wps
Begin Testing...
[Epoch 7] train avg loss 0.0113307, test acc 0.7198, test avg loss 0.571294, throughput 12.9956K wps
Observed Improvement.
Begin Testing...
[Epoch 8 Batch 30/173] avg loss 0.0108702, throughput 13.291K wps
[Epoch 8 Batch 60/173] avg loss 0.0109973, throughput 12.9014K wps
[Epoch 8 Batch 90/173] avg loss 0.0108421, throughput 12.9799K wps
[Epoch 8 Batch 120/173] avg loss 0.0110488, throughput 12.9469K wps
[Epoch 8 Batch 150/173] avg loss 0.0110163, throughput 12.9187K wps
Begin Testing...
[Epoch 8] train avg loss 0.0109481, test acc 0.7396, test avg loss 0.558713, throughput 13.0089K wps
Observed Improvement.
Begin Testing...
[Epoch 9 Batch 30/173] avg loss 0.0105844, throughput 13.3011K wps
[Epoch 9 Batch 60/173] avg loss 0.0107421, throughput 12.8386K wps
[Epoch 9 Batch 90/173] avg loss 0.0104786, throughput 12.9443K wps
[Epoch 9 Batch 120/173] avg loss 0.0103171, throughput 12.9575K wps
[Epoch 9 Batch 150/173] avg loss 0.0106638, throughput 12.94K wps
Begin Testing...
[Epoch 9] train avg loss 0.0105836, test acc 0.7427, test avg loss 0.544954, throughput 12.9973K wps
Observed Improvement.
Begin Testing...
[Epoch 10 Batch 30/173] avg loss 0.0106103, throughput 13.2943K wps
[Epoch 10 Batch 60/173] avg loss 0.0104338, throughput 12.8574K wps
[Epoch 10 Batch 90/173] avg loss 0.0104878, throughput 12.976K wps
[Epoch 10 Batch 120/173] avg loss 0.00990981, throughput 12.9541K wps
[Epoch 10 Batch 150/173] avg loss 0.00997924, throughput 12.9666K wps
Begin Testing...
[Epoch 10] train avg loss 0.0102743, test acc 0.7375, test avg loss 0.535761, throughput 13.0038K wps
[Epoch 11 Batch 30/173] avg loss 0.0101004, throughput 13.3172K wps
[Epoch 11 Batch 60/173] avg loss 0.00979428, throughput 12.8351K wps
[Epoch 11 Batch 90/173] avg loss 0.00979123, throughput 12.9771K wps
[Epoch 11 Batch 120/173] avg loss 0.0100119, throughput 12.9689K wps
[Epoch 11 Batch 150/173] avg loss 0.00994654, throughput 12.9568K wps
Begin Testing...
[Epoch 11] train avg loss 0.00997349, test acc 0.7385, test avg loss 0.530303, throughput 13.0054K wps
[Epoch 12 Batch 30/173] avg loss 0.00964997, throughput 13.2562K wps
[Epoch 12 Batch 60/173] avg loss 0.00975834, throughput 12.881K wps
[Epoch 12 Batch 90/173] avg loss 0.00984173, throughput 12.957K wps
[Epoch 12 Batch 120/173] avg loss 0.00963896, throughput 12.9279K wps
[Epoch 12 Batch 150/173] avg loss 0.00974373, throughput 12.9426K wps
Begin Testing...
[Epoch 12] train avg loss 0.00965088, test acc 0.7562, test avg loss 0.512517, throughput 12.9907K wps
Observed Improvement.
Begin Testing...
[Epoch 13 Batch 30/173] avg loss 0.00922695, throughput 13.3309K wps
[Epoch 13 Batch 60/173] avg loss 0.00960116, throughput 12.905K wps
[Epoch 13 Batch 90/173] avg loss 0.0091584, throughput 12.9456K wps
[Epoch 13 Batch 120/173] avg loss 0.00946733, throughput 12.9479K wps
[Epoch 13 Batch 150/173] avg loss 0.00949724, throughput 12.9643K wps
Begin Testing...
[Epoch 13] train avg loss 0.0093787, test acc 0.7531, test avg loss 0.509087, throughput 13.0127K wps
[Epoch 14 Batch 30/173] avg loss 0.00936114, throughput 13.3467K wps
[Epoch 14 Batch 60/173] avg loss 0.00934978, throughput 12.8803K wps
[Epoch 14 Batch 90/173] avg loss 0.00930321, throughput 12.9233K wps
[Epoch 14 Batch 120/173] avg loss 0.0090138, throughput 12.9177K wps
[Epoch 14 Batch 150/173] avg loss 0.00932916, throughput 12.9327K wps
Begin Testing...
[Epoch 14] train avg loss 0.00923914, test acc 0.7646, test avg loss 0.500219, throughput 12.9908K wps
Observed Improvement.
Begin Testing...
[Epoch 15 Batch 30/173] avg loss 0.00889574, throughput 13.263K wps
[Epoch 15 Batch 60/173] avg loss 0.00887032, throughput 12.9022K wps
[Epoch 15 Batch 90/173] avg loss 0.00907871, throughput 12.9541K wps
[Epoch 15 Batch 120/173] avg loss 0.00896236, throughput 12.9315K wps
[Epoch 15 Batch 150/173] avg loss 0.0086923, throughput 12.9394K wps
Begin Testing...
[Epoch 15] train avg loss 0.00890502, test acc 0.7750, test avg loss 0.492443, throughput 12.9904K wps
Observed Improvement.
Begin Testing...
[Epoch 16 Batch 30/173] avg loss 0.00880403, throughput 13.3055K wps
[Epoch 16 Batch 60/173] avg loss 0.00859188, throughput 12.7946K wps
[Epoch 16 Batch 90/173] avg loss 0.00836164, throughput 12.9464K wps
[Epoch 16 Batch 120/173] avg loss 0.00851168, throughput 12.9556K wps
[Epoch 16 Batch 150/173] avg loss 0.00886732, throughput 12.979K wps
Begin Testing...
[Epoch 16] train avg loss 0.00869962, test acc 0.7646, test avg loss 0.491083, throughput 12.9967K wps
[Epoch 17 Batch 30/173] avg loss 0.00853803, throughput 13.3524K wps
[Epoch 17 Batch 60/173] avg loss 0.00828442, throughput 12.9323K wps
[Epoch 17 Batch 90/173] avg loss 0.00847551, throughput 12.9352K wps
[Epoch 17 Batch 120/173] avg loss 0.00831875, throughput 12.9723K wps
[Epoch 17 Batch 150/173] avg loss 0.00890617, throughput 13.008K wps
Begin Testing...
[Epoch 17] train avg loss 0.00852547, test acc 0.7625, test avg loss 0.484428, throughput 13.0341K wps
[Epoch 18 Batch 30/173] avg loss 0.00820396, throughput 13.3418K wps
[Epoch 18 Batch 60/173] avg loss 0.00862128, throughput 12.832K wps
[Epoch 18 Batch 90/173] avg loss 0.00841677, throughput 12.9431K wps
[Epoch 18 Batch 120/173] avg loss 0.00831394, throughput 12.9436K wps
[Epoch 18 Batch 150/173] avg loss 0.00829101, throughput 12.9712K wps
Begin Testing...
[Epoch 18] train avg loss 0.00836705, test acc 0.7708, test avg loss 0.478399, throughput 13.001K wps
[Epoch 19 Batch 30/173] avg loss 0.00797731, throughput 13.1569K wps
[Epoch 19 Batch 60/173] avg loss 0.00806388, throughput 12.891K wps
[Epoch 19 Batch 90/173] avg loss 0.00799836, throughput 12.9418K wps
[Epoch 19 Batch 120/173] avg loss 0.00828359, throughput 12.9261K wps
[Epoch 19 Batch 150/173] avg loss 0.00816816, throughput 12.8971K wps
Begin Testing...
[Epoch 19] train avg loss 0.0082115, test acc 0.7635, test avg loss 0.480722, throughput 12.9583K wps
[Epoch 20 Batch 30/173] avg loss 0.00830761, throughput 13.2742K wps
[Epoch 20 Batch 60/173] avg loss 0.00802578, throughput 12.8125K wps
[Epoch 20 Batch 90/173] avg loss 0.00795506, throughput 12.8703K wps
[Epoch 20 Batch 120/173] avg loss 0.00773294, throughput 12.8605K wps
[Epoch 20 Batch 150/173] avg loss 0.00805489, throughput 12.9349K wps
Begin Testing...
[Epoch 20] train avg loss 0.00800905, test acc 0.7750, test avg loss 0.470316, throughput 12.9475K wps
Observed Improvement.
Begin Testing...
[Epoch 21 Batch 30/173] avg loss 0.00755878, throughput 13.325K wps
[Epoch 21 Batch 60/173] avg loss 0.00788849, throughput 12.8011K wps
[Epoch 21 Batch 90/173] avg loss 0.00770711, throughput 12.936K wps
[Epoch 21 Batch 120/173] avg loss 0.00795512, throughput 12.9265K wps
[Epoch 21 Batch 150/173] avg loss 0.00817099, throughput 12.9843K wps
Begin Testing...
[Epoch 21] train avg loss 0.00785167, test acc 0.7802, test avg loss 0.465014, throughput 12.9956K wps
Observed Improvement.
Begin Testing...
[Epoch 22 Batch 30/173] avg loss 0.00757227, throughput 13.2799K wps
[Epoch 22 Batch 60/173] avg loss 0.00758813, throughput 12.7462K wps
[Epoch 22 Batch 90/173] avg loss 0.00774703, throughput 12.9226K wps
[Epoch 22 Batch 120/173] avg loss 0.00794302, throughput 12.9233K wps
[Epoch 22 Batch 150/173] avg loss 0.00768833, throughput 12.9442K wps
Begin Testing...
[Epoch 22] train avg loss 0.00768997, test acc 0.7750, test avg loss 0.466904, throughput 12.9604K wps
[Epoch 23 Batch 30/173] avg loss 0.00717835, throughput 13.3271K wps
[Epoch 23 Batch 60/173] avg loss 0.00740941, throughput 12.8114K wps
[Epoch 23 Batch 90/173] avg loss 0.00789791, throughput 12.9438K wps
[Epoch 23 Batch 120/173] avg loss 0.00762343, throughput 12.9745K wps
[Epoch 23 Batch 150/173] avg loss 0.00716965, throughput 12.9545K wps
Begin Testing...
[Epoch 23] train avg loss 0.00747743, test acc 0.7698, test avg loss 0.468798, throughput 12.9968K wps
[Epoch 24 Batch 30/173] avg loss 0.00757202, throughput 13.2331K wps
[Epoch 24 Batch 60/173] avg loss 0.00752931, throughput 12.8083K wps
[Epoch 24 Batch 90/173] avg loss 0.00725671, throughput 12.9637K wps
[Epoch 24 Batch 120/173] avg loss 0.00702603, throughput 12.9537K wps
[Epoch 24 Batch 150/173] avg loss 0.0075811, throughput 12.9343K wps
Begin Testing...
[Epoch 24] train avg loss 0.00744351, test acc 0.7792, test avg loss 0.461787, throughput 12.9761K wps
[Epoch 25 Batch 30/173] avg loss 0.00746927, throughput 13.1551K wps
[Epoch 25 Batch 60/173] avg loss 0.00738579, throughput 12.826K wps
[Epoch 25 Batch 90/173] avg loss 0.00703574, throughput 12.9277K wps
[Epoch 25 Batch 120/173] avg loss 0.00733113, throughput 12.9349K wps
[Epoch 25 Batch 150/173] avg loss 0.00714938, throughput 12.9435K wps
Begin Testing...
[Epoch 25] train avg loss 0.00726606, test acc 0.7833, test avg loss 0.460491, throughput 12.9607K wps
Observed Improvement.
Begin Testing...
[Epoch 26 Batch 30/173] avg loss 0.00694838, throughput 13.3613K wps
[Epoch 26 Batch 60/173] avg loss 0.00742717, throughput 12.7431K wps
[Epoch 26 Batch 90/173] avg loss 0.00680323, throughput 12.9552K wps
[Epoch 26 Batch 120/173] avg loss 0.00733562, throughput 12.9592K wps
[Epoch 26 Batch 150/173] avg loss 0.00686396, throughput 12.9253K wps
Begin Testing...
[Epoch 26] train avg loss 0.00710963, test acc 0.7792, test avg loss 0.460635, throughput 12.985K wps
[Epoch 27 Batch 30/173] avg loss 0.00719649, throughput 13.1847K wps
[Epoch 27 Batch 60/173] avg loss 0.00676416, throughput 12.9228K wps
[Epoch 27 Batch 90/173] avg loss 0.00695758, throughput 12.955K wps
[Epoch 27 Batch 120/173] avg loss 0.00667442, throughput 12.9843K wps
[Epoch 27 Batch 150/173] avg loss 0.00725982, throughput 12.9744K wps
Begin Testing...
[Epoch 27] train avg loss 0.00696797, test acc 0.7771, test avg loss 0.461324, throughput 12.9997K wps
[Epoch 28 Batch 30/173] avg loss 0.00690231, throughput 13.0399K wps
[Epoch 28 Batch 60/173] avg loss 0.00666162, throughput 12.7013K wps
[Epoch 28 Batch 90/173] avg loss 0.00684556, throughput 12.7687K wps
[Epoch 28 Batch 120/173] avg loss 0.0067112, throughput 12.9543K wps
[Epoch 28 Batch 150/173] avg loss 0.00670379, throughput 12.9743K wps
Begin Testing...
[Epoch 28] train avg loss 0.00682157, test acc 0.7750, test avg loss 0.463266, throughput 12.8948K wps
[Epoch 29 Batch 30/173] avg loss 0.00662109, throughput 13.2786K wps
[Epoch 29 Batch 60/173] avg loss 0.00673312, throughput 12.7935K wps
[Epoch 29 Batch 90/173] avg loss 0.00673485, throughput 12.8297K wps
[Epoch 29 Batch 120/173] avg loss 0.00674987, throughput 12.9131K wps
[Epoch 29 Batch 150/173] avg loss 0.00675909, throughput 12.8773K wps
Begin Testing...
[Epoch 29] train avg loss 0.00670678, test acc 0.7771, test avg loss 0.462734, throughput 12.94K wps
[Epoch 30 Batch 30/173] avg loss 0.00665403, throughput 13.2954K wps
[Epoch 30 Batch 60/173] avg loss 0.00648865, throughput 12.8232K wps
[Epoch 30 Batch 90/173] avg loss 0.0064446, throughput 12.9536K wps
[Epoch 30 Batch 120/173] avg loss 0.00667004, throughput 13.0009K wps
[Epoch 30 Batch 150/173] avg loss 0.00625788, throughput 12.7974K wps
Begin Testing...
[Epoch 30] train avg loss 0.00655445, test acc 0.7844, test avg loss 0.457934, throughput 12.9673K wps
Observed Improvement.
Begin Testing...
[Epoch 31 Batch 30/173] avg loss 0.00649038, throughput 13.3074K wps
[Epoch 31 Batch 60/173] avg loss 0.0067185, throughput 12.7435K wps
[Epoch 31 Batch 90/173] avg loss 0.00623618, throughput 12.8492K wps
[Epoch 31 Batch 120/173] avg loss 0.00653005, throughput 12.9364K wps
[Epoch 31 Batch 150/173] avg loss 0.00636431, throughput 12.9474K wps
Begin Testing...
[Epoch 31] train avg loss 0.00641758, test acc 0.7823, test avg loss 0.457059, throughput 12.9586K wps
[Epoch 32 Batch 30/173] avg loss 0.00650854, throughput 13.2159K wps
[Epoch 32 Batch 60/173] avg loss 0.00673777, throughput 12.7645K wps
[Epoch 32 Batch 90/173] avg loss 0.00627216, throughput 12.9162K wps
[Epoch 32 Batch 120/173] avg loss 0.00595593, throughput 12.9613K wps
[Epoch 32 Batch 150/173] avg loss 0.0063348, throughput 12.9672K wps
Begin Testing...
[Epoch 32] train avg loss 0.00634035, test acc 0.7802, test avg loss 0.461022, throughput 12.9489K wps
[Epoch 33 Batch 30/173] avg loss 0.00630493, throughput 13.2561K wps
[Epoch 33 Batch 60/173] avg loss 0.00611025, throughput 12.8327K wps
[Epoch 33 Batch 90/173] avg loss 0.00618304, throughput 12.9722K wps
[Epoch 33 Batch 120/173] avg loss 0.00645856, throughput 12.9136K wps
[Epoch 33 Batch 150/173] avg loss 0.00599624, throughput 12.8044K wps
Begin Testing...
[Epoch 33] train avg loss 0.00619486, test acc 0.7917, test avg loss 0.456454, throughput 12.9479K wps
Observed Improvement.
Begin Testing...
[Epoch 34 Batch 30/173] avg loss 0.0058719, throughput 13.2856K wps
[Epoch 34 Batch 60/173] avg loss 0.00623594, throughput 12.8001K wps
[Epoch 34 Batch 90/173] avg loss 0.0064074, throughput 12.9207K wps
[Epoch 34 Batch 120/173] avg loss 0.00606852, throughput 12.9437K wps
[Epoch 34 Batch 150/173] avg loss 0.00604246, throughput 12.9139K wps
Begin Testing...
[Epoch 34] train avg loss 0.00614555, test acc 0.7865, test avg loss 0.452823, throughput 12.9696K wps
[Epoch 35 Batch 30/173] avg loss 0.00570663, throughput 13.2587K wps
[Epoch 35 Batch 60/173] avg loss 0.00590534, throughput 12.8277K wps
[Epoch 35 Batch 90/173] avg loss 0.00611385, throughput 12.9501K wps
[Epoch 35 Batch 120/173] avg loss 0.00595624, throughput 12.9363K wps
[Epoch 35 Batch 150/173] avg loss 0.00610243, throughput 12.9489K wps
Begin Testing...
[Epoch 35] train avg loss 0.00599153, test acc 0.7865, test avg loss 0.452735, throughput 12.9828K wps
[Epoch 36 Batch 30/173] avg loss 0.0059491, throughput 13.2807K wps
[Epoch 36 Batch 60/173] avg loss 0.00551969, throughput 12.8536K wps
[Epoch 36 Batch 90/173] avg loss 0.00577834, throughput 12.9682K wps
[Epoch 36 Batch 120/173] avg loss 0.00609349, throughput 12.906K wps
[Epoch 36 Batch 150/173] avg loss 0.00569965, throughput 12.9341K wps
Begin Testing...
[Epoch 36] train avg loss 0.00583535, test acc 0.7802, test avg loss 0.455048, throughput 12.9883K wps
[Epoch 37 Batch 30/173] avg loss 0.00547297, throughput 13.2703K wps
[Epoch 37 Batch 60/173] avg loss 0.00598629, throughput 12.8105K wps
[Epoch 37 Batch 90/173] avg loss 0.00566083, throughput 12.9703K wps
[Epoch 37 Batch 120/173] avg loss 0.00610216, throughput 12.9556K wps
[Epoch 37 Batch 150/173] avg loss 0.00582433, throughput 12.9667K wps
Begin Testing...
[Epoch 37] train avg loss 0.00577063, test acc 0.7729, test avg loss 0.460898, throughput 12.9961K wps
[Epoch 38 Batch 30/173] avg loss 0.00556016, throughput 13.3101K wps
[Epoch 38 Batch 60/173] avg loss 0.00550609, throughput 12.8224K wps
[Epoch 38 Batch 90/173] avg loss 0.00562732, throughput 12.9624K wps
[Epoch 38 Batch 120/173] avg loss 0.00570072, throughput 12.9232K wps
[Epoch 38 Batch 150/173] avg loss 0.00578046, throughput 12.7786K wps
Begin Testing...
[Epoch 38] train avg loss 0.00558806, test acc 0.7833, test avg loss 0.455255, throughput 12.9663K wps
[Epoch 39 Batch 30/173] avg loss 0.00546131, throughput 13.1746K wps
[Epoch 39 Batch 60/173] avg loss 0.00538219, throughput 12.8407K wps
[Epoch 39 Batch 90/173] avg loss 0.00562371, throughput 12.9224K wps
[Epoch 39 Batch 120/173] avg loss 0.00559729, throughput 12.9436K wps
[Epoch 39 Batch 150/173] avg loss 0.00558224, throughput 12.9403K wps
Begin Testing...
[Epoch 39] train avg loss 0.00552787, test acc 0.7823, test avg loss 0.45467, throughput 12.9631K wps
[Epoch 40 Batch 30/173] avg loss 0.0054545, throughput 13.2115K wps
[Epoch 40 Batch 60/173] avg loss 0.00562267, throughput 12.8227K wps
[Epoch 40 Batch 90/173] avg loss 0.00507709, throughput 12.9539K wps
[Epoch 40 Batch 120/173] avg loss 0.00572181, throughput 12.9921K wps
[Epoch 40 Batch 150/173] avg loss 0.00538492, throughput 12.8179K wps
Begin Testing...
[Epoch 40] train avg loss 0.0054625, test acc 0.7865, test avg loss 0.455294, throughput 12.9625K wps
[Epoch 41 Batch 30/173] avg loss 0.00542276, throughput 13.3227K wps
[Epoch 41 Batch 60/173] avg loss 0.00545169, throughput 12.8104K wps
[Epoch 41 Batch 90/173] avg loss 0.00546759, throughput 12.9335K wps
[Epoch 41 Batch 120/173] avg loss 0.0053314, throughput 12.9701K wps
[Epoch 41 Batch 150/173] avg loss 0.00527679, throughput 12.9772K wps
Begin Testing...
[Epoch 41] train avg loss 0.00535499, test acc 0.7885, test avg loss 0.452254, throughput 12.9988K wps
[Epoch 42 Batch 30/173] avg loss 0.0049029, throughput 13.1992K wps
[Epoch 42 Batch 60/173] avg loss 0.00504269, throughput 12.7827K wps
[Epoch 42 Batch 90/173] avg loss 0.00521222, throughput 12.9004K wps
[Epoch 42 Batch 120/173] avg loss 0.00497889, throughput 12.8308K wps
[Epoch 42 Batch 150/173] avg loss 0.00492567, throughput 12.8555K wps
Begin Testing...
[Epoch 42] train avg loss 0.00508279, test acc 0.7781, test avg loss 0.452465, throughput 12.9239K wps
[Epoch 43 Batch 30/173] avg loss 0.00519872, throughput 13.1968K wps
[Epoch 43 Batch 60/173] avg loss 0.00510043, throughput 12.7815K wps
[Epoch 43 Batch 90/173] avg loss 0.00516467, throughput 12.9461K wps
[Epoch 43 Batch 120/173] avg loss 0.00493881, throughput 12.9303K wps
[Epoch 43 Batch 150/173] avg loss 0.00521776, throughput 12.9514K wps
Begin Testing...
[Epoch 43] train avg loss 0.0051267, test acc 0.7750, test avg loss 0.463696, throughput 12.9547K wps
[Epoch 44 Batch 30/173] avg loss 0.00501701, throughput 13.1891K wps
[Epoch 44 Batch 60/173] avg loss 0.00475751, throughput 12.7529K wps
[Epoch 44 Batch 90/173] avg loss 0.00507656, throughput 12.9068K wps
[Epoch 44 Batch 120/173] avg loss 0.00485195, throughput 12.931K wps
[Epoch 44 Batch 150/173] avg loss 0.00493067, throughput 12.9175K wps
Begin Testing...
[Epoch 44] train avg loss 0.00497849, test acc 0.7802, test avg loss 0.458845, throughput 12.924K wps
[Epoch 45 Batch 30/173] avg loss 0.00489228, throughput 13.2K wps
[Epoch 45 Batch 60/173] avg loss 0.00468165, throughput 12.7789K wps
[Epoch 45 Batch 90/173] avg loss 0.00472042, throughput 12.9164K wps
[Epoch 45 Batch 120/173] avg loss 0.0047817, throughput 12.9158K wps
[Epoch 45 Batch 150/173] avg loss 0.00531316, throughput 12.9453K wps
Begin Testing...
[Epoch 45] train avg loss 0.00489724, test acc 0.7823, test avg loss 0.457988, throughput 12.9501K wps
[Epoch 46 Batch 30/173] avg loss 0.00468842, throughput 13.1995K wps
[Epoch 46 Batch 60/173] avg loss 0.00475317, throughput 12.8084K wps
[Epoch 46 Batch 90/173] avg loss 0.00468282, throughput 12.9593K wps
[Epoch 46 Batch 120/173] avg loss 0.00466676, throughput 12.7691K wps
[Epoch 46 Batch 150/173] avg loss 0.00478985, throughput 12.9674K wps
Begin Testing...
[Epoch 46] train avg loss 0.00471024, test acc 0.7865, test avg loss 0.461647, throughput 12.9442K wps
[Epoch 47 Batch 30/173] avg loss 0.00484329, throughput 13.1879K wps
[Epoch 47 Batch 60/173] avg loss 0.00463943, throughput 12.7896K wps
[Epoch 47 Batch 90/173] avg loss 0.00450173, throughput 12.801K wps
[Epoch 47 Batch 120/173] avg loss 0.00436027, throughput 12.9604K wps
[Epoch 47 Batch 150/173] avg loss 0.00458065, throughput 12.828K wps
Begin Testing...
[Epoch 47] train avg loss 0.00465262, test acc 0.7885, test avg loss 0.457629, throughput 12.9134K wps
[Epoch 48 Batch 30/173] avg loss 0.00433227, throughput 13.3232K wps
[Epoch 48 Batch 60/173] avg loss 0.00473509, throughput 12.7764K wps
[Epoch 48 Batch 90/173] avg loss 0.00464387, throughput 12.7551K wps
[Epoch 48 Batch 120/173] avg loss 0.0046356, throughput 12.8054K wps
[Epoch 48 Batch 150/173] avg loss 0.0046602, throughput 12.8107K wps
Begin Testing...
[Epoch 48] train avg loss 0.00454566, test acc 0.7885, test avg loss 0.457302, throughput 12.8816K wps
[Epoch 49 Batch 30/173] avg loss 0.00457741, throughput 13.2243K wps
[Epoch 49 Batch 60/173] avg loss 0.00434479, throughput 12.8401K wps
[Epoch 49 Batch 90/173] avg loss 0.00447994, throughput 12.8599K wps
[Epoch 49 Batch 120/173] avg loss 0.00436165, throughput 12.7696K wps
[Epoch 49 Batch 150/173] avg loss 0.00448306, throughput 12.9263K wps
Begin Testing...
[Epoch 49] train avg loss 0.00447029, test acc 0.7917, test avg loss 0.457924, throughput 12.9121K wps
Observed Improvement.
Begin Testing...
[Epoch 50 Batch 30/173] avg loss 0.00435583, throughput 13.2513K wps
[Epoch 50 Batch 60/173] avg loss 0.00437526, throughput 12.7965K wps
[Epoch 50 Batch 90/173] avg loss 0.00456938, throughput 12.9285K wps
[Epoch 50 Batch 120/173] avg loss 0.00436426, throughput 12.9126K wps
[Epoch 50 Batch 150/173] avg loss 0.00430349, throughput 12.9995K wps
Begin Testing...
[Epoch 50] train avg loss 0.00442033, test acc 0.7760, test avg loss 0.477812, throughput 12.9778K wps
[Epoch 51 Batch 30/173] avg loss 0.0040417, throughput 13.2291K wps
[Epoch 51 Batch 60/173] avg loss 0.00412759, throughput 12.794K wps
[Epoch 51 Batch 90/173] avg loss 0.00432129, throughput 12.8629K wps
[Epoch 51 Batch 120/173] avg loss 0.00443461, throughput 12.9592K wps
[Epoch 51 Batch 150/173] avg loss 0.00423884, throughput 12.8156K wps
Begin Testing...
[Epoch 51] train avg loss 0.00424023, test acc 0.7885, test avg loss 0.456841, throughput 12.9185K wps
[Epoch 52 Batch 30/173] avg loss 0.00403328, throughput 13.2519K wps
[Epoch 52 Batch 60/173] avg loss 0.00454696, throughput 12.7543K wps
[Epoch 52 Batch 90/173] avg loss 0.00408829, throughput 12.9715K wps
[Epoch 52 Batch 120/173] avg loss 0.00443207, throughput 12.9244K wps
[Epoch 52 Batch 150/173] avg loss 0.00405271, throughput 12.9273K wps
Begin Testing...
[Epoch 52] train avg loss 0.0042075, test acc 0.7906, test avg loss 0.459623, throughput 12.9653K wps
[Epoch 53 Batch 30/173] avg loss 0.00442795, throughput 13.2256K wps
[Epoch 53 Batch 60/173] avg loss 0.00404942, throughput 12.8118K wps
[Epoch 53 Batch 90/173] avg loss 0.00426836, throughput 12.9256K wps
[Epoch 53 Batch 120/173] avg loss 0.00384331, throughput 12.9657K wps
[Epoch 53 Batch 150/173] avg loss 0.00388063, throughput 12.9716K wps
Begin Testing...
[Epoch 53] train avg loss 0.00414721, test acc 0.7854, test avg loss 0.458914, throughput 12.9726K wps
[Epoch 54 Batch 30/173] avg loss 0.00414422, throughput 13.1011K wps
[Epoch 54 Batch 60/173] avg loss 0.00396758, throughput 12.877K wps
[Epoch 54 Batch 90/173] avg loss 0.00404299, throughput 12.9468K wps
[Epoch 54 Batch 120/173] avg loss 0.00416702, throughput 12.954K wps
[Epoch 54 Batch 150/173] avg loss 0.00417509, throughput 12.9324K wps
Begin Testing...
[Epoch 54] train avg loss 0.00408097, test acc 0.7937, test avg loss 0.456581, throughput 12.9641K wps
Observed Improvement.
Begin Testing...
[Epoch 55 Batch 30/173] avg loss 0.00383879, throughput 13.3104K wps
[Epoch 55 Batch 60/173] avg loss 0.00388904, throughput 12.8208K wps
[Epoch 55 Batch 90/173] avg loss 0.00387697, throughput 12.9338K wps
[Epoch 55 Batch 120/173] avg loss 0.00376065, throughput 12.944K wps
[Epoch 55 Batch 150/173] avg loss 0.00397757, throughput 12.9907K wps
Begin Testing...
[Epoch 55] train avg loss 0.00392401, test acc 0.7875, test avg loss 0.46662, throughput 12.9963K wps
[Epoch 56 Batch 30/173] avg loss 0.00376378, throughput 13.2984K wps
[Epoch 56 Batch 60/173] avg loss 0.00398746, throughput 12.7991K wps
[Epoch 56 Batch 90/173] avg loss 0.00375447, throughput 12.9602K wps
[Epoch 56 Batch 120/173] avg loss 0.00379219, throughput 12.888K wps
[Epoch 56 Batch 150/173] avg loss 0.00394814, throughput 12.9595K wps
Begin Testing...
[Epoch 56] train avg loss 0.00385722, test acc 0.7927, test avg loss 0.463154, throughput 12.984K wps
[Epoch 57 Batch 30/173] avg loss 0.00356171, throughput 13.2475K wps
[Epoch 57 Batch 60/173] avg loss 0.00395626, throughput 12.8436K wps
[Epoch 57 Batch 90/173] avg loss 0.00386819, throughput 13.0047K wps
[Epoch 57 Batch 120/173] avg loss 0.00375633, throughput 12.9987K wps
[Epoch 57 Batch 150/173] avg loss 0.0036804, throughput 12.9665K wps
Begin Testing...
[Epoch 57] train avg loss 0.00374732, test acc 0.7802, test avg loss 0.474795, throughput 12.9888K wps
[Epoch 58 Batch 30/173] avg loss 0.00353479, throughput 13.2803K wps
[Epoch 58 Batch 60/173] avg loss 0.00367683, throughput 12.7687K wps
[Epoch 58 Batch 90/173] avg loss 0.00365566, throughput 12.8988K wps
[Epoch 58 Batch 120/173] avg loss 0.00360591, throughput 12.9529K wps
[Epoch 58 Batch 150/173] avg loss 0.00398232, throughput 12.9349K wps
Begin Testing...
[Epoch 58] train avg loss 0.00369688, test acc 0.7802, test avg loss 0.466659, throughput 12.9715K wps
[Epoch 59 Batch 30/173] avg loss 0.00373147, throughput 13.2478K wps
[Epoch 59 Batch 60/173] avg loss 0.00365347, throughput 12.7675K wps
[Epoch 59 Batch 90/173] avg loss 0.00336023, throughput 12.9836K wps
[Epoch 59 Batch 120/173] avg loss 0.00344199, throughput 12.9223K wps
[Epoch 59 Batch 150/173] avg loss 0.0036154, throughput 12.981K wps
Begin Testing...
[Epoch 59] train avg loss 0.00356144, test acc 0.7885, test avg loss 0.471678, throughput 12.9558K wps
Test loss 0.437224, test acc 0.7974
Total time cost 174.94s
9596 1066
[Epoch 0 Batch 30/173] avg loss 0.0152363, throughput 11.7067K wps
[Epoch 0 Batch 60/173] avg loss 0.0148721, throughput 12.7973K wps
[Epoch 0 Batch 90/173] avg loss 0.0149196, throughput 12.86K wps
[Epoch 0 Batch 120/173] avg loss 0.0144451, throughput 12.9612K wps
[Epoch 0 Batch 150/173] avg loss 0.0147371, throughput 12.7688K wps
Begin Testing...
[Epoch 0] train avg loss 0.0147347, test acc 0.5833, test avg loss 0.669076, throughput 12.6383K wps
Observed Improvement.
Begin Testing...
[Epoch 1 Batch 30/173] avg loss 0.0138651, throughput 13.2966K wps
[Epoch 1 Batch 60/173] avg loss 0.0137036, throughput 12.78K wps
[Epoch 1 Batch 90/173] avg loss 0.0137273, throughput 12.8693K wps
[Epoch 1 Batch 120/173] avg loss 0.0133841, throughput 12.8931K wps
[Epoch 1 Batch 150/173] avg loss 0.0135879, throughput 12.8673K wps
Begin Testing...
[Epoch 1] train avg loss 0.0136623, test acc 0.6198, test avg loss 0.656362, throughput 12.9201K wps
Observed Improvement.
Begin Testing...
[Epoch 2 Batch 30/173] avg loss 0.0131913, throughput 13.2756K wps
[Epoch 2 Batch 60/173] avg loss 0.0132994, throughput 12.795K wps
[Epoch 2 Batch 90/173] avg loss 0.0132643, throughput 12.9251K wps
[Epoch 2 Batch 120/173] avg loss 0.0130178, throughput 12.8127K wps
[Epoch 2 Batch 150/173] avg loss 0.0131641, throughput 12.8067K wps
Begin Testing...
[Epoch 2] train avg loss 0.0131951, test acc 0.6510, test avg loss 0.642331, throughput 12.9234K wps
Observed Improvement.
Begin Testing...
[Epoch 3 Batch 30/173] avg loss 0.0131481, throughput 13.3K wps
[Epoch 3 Batch 60/173] avg loss 0.0128074, throughput 12.7993K wps
[Epoch 3 Batch 90/173] avg loss 0.0126622, throughput 12.8078K wps
[Epoch 3 Batch 120/173] avg loss 0.0128161, throughput 12.8439K wps
[Epoch 3 Batch 150/173] avg loss 0.0128786, throughput 12.7991K wps
Begin Testing...
[Epoch 3] train avg loss 0.0128703, test acc 0.6573, test avg loss 0.631575, throughput 12.8947K wps
Observed Improvement.
Begin Testing...
[Epoch 4 Batch 30/173] avg loss 0.0125037, throughput 13.1474K wps
[Epoch 4 Batch 60/173] avg loss 0.0124793, throughput 12.7988K wps
[Epoch 4 Batch 90/173] avg loss 0.0124097, throughput 12.7701K wps
[Epoch 4 Batch 120/173] avg loss 0.0122808, throughput 12.7619K wps
[Epoch 4 Batch 150/173] avg loss 0.0125238, throughput 12.7407K wps
Begin Testing...
[Epoch 4] train avg loss 0.0124481, test acc 0.6906, test avg loss 0.616717, throughput 12.856K wps
Observed Improvement.
Begin Testing...
[Epoch 5 Batch 30/173] avg loss 0.0123363, throughput 13.2709K wps
[Epoch 5 Batch 60/173] avg loss 0.0122012, throughput 12.7811K wps
[Epoch 5 Batch 90/173] avg loss 0.0119516, throughput 12.7762K wps
[Epoch 5 Batch 120/173] avg loss 0.0120247, throughput 12.8708K wps
[Epoch 5 Batch 150/173] avg loss 0.0121276, throughput 12.7714K wps
Begin Testing...
[Epoch 5] train avg loss 0.0121366, test acc 0.6802, test avg loss 0.60505, throughput 12.88K wps
[Epoch 6 Batch 30/173] avg loss 0.0118421, throughput 13.2529K wps
[Epoch 6 Batch 60/173] avg loss 0.0117915, throughput 12.6621K wps
[Epoch 6 Batch 90/173] avg loss 0.0116505, throughput 12.9333K wps
[Epoch 6 Batch 120/173] avg loss 0.0120754, throughput 12.8234K wps
[Epoch 6 Batch 150/173] avg loss 0.0116965, throughput 12.9147K wps
Begin Testing...
[Epoch 6] train avg loss 0.0117924, test acc 0.7198, test avg loss 0.586626, throughput 12.9028K wps
Observed Improvement.
Begin Testing...
[Epoch 7 Batch 30/173] avg loss 0.0115089, throughput 13.2589K wps
[Epoch 7 Batch 60/173] avg loss 0.0113099, throughput 12.7803K wps
[Epoch 7 Batch 90/173] avg loss 0.0116943, throughput 12.786K wps
[Epoch 7 Batch 120/173] avg loss 0.011342, throughput 12.7812K wps
[Epoch 7 Batch 150/173] avg loss 0.0112863, throughput 12.7941K wps
Begin Testing...
[Epoch 7] train avg loss 0.0113925, test acc 0.7344, test avg loss 0.571464, throughput 12.8626K wps
Observed Improvement.
Begin Testing...
[Epoch 8 Batch 30/173] avg loss 0.0113911, throughput 13.291K wps
[Epoch 8 Batch 60/173] avg loss 0.0109869, throughput 12.7982K wps
[Epoch 8 Batch 90/173] avg loss 0.0112121, throughput 12.7829K wps
[Epoch 8 Batch 120/173] avg loss 0.0110124, throughput 12.8362K wps
[Epoch 8 Batch 150/173] avg loss 0.0109786, throughput 12.9852K wps
Begin Testing...
[Epoch 8] train avg loss 0.0111349, test acc 0.7385, test avg loss 0.560672, throughput 12.9291K wps
Observed Improvement.
Begin Testing...
[Epoch 9 Batch 30/173] avg loss 0.0108551, throughput 13.2456K wps
[Epoch 9 Batch 60/173] avg loss 0.0107543, throughput 12.7973K wps
[Epoch 9 Batch 90/173] avg loss 0.0106109, throughput 12.8544K wps
[Epoch 9 Batch 120/173] avg loss 0.0106375, throughput 12.9758K wps
[Epoch 9 Batch 150/173] avg loss 0.0106887, throughput 12.8178K wps
Begin Testing...
[Epoch 9] train avg loss 0.0107149, test acc 0.7292, test avg loss 0.550351, throughput 12.9396K wps
[Epoch 10 Batch 30/173] avg loss 0.0105364, throughput 13.2698K wps
[Epoch 10 Batch 60/173] avg loss 0.0104034, throughput 12.855K wps
[Epoch 10 Batch 90/173] avg loss 0.0104934, throughput 12.9496K wps
[Epoch 10 Batch 120/173] avg loss 0.0105157, throughput 12.818K wps
[Epoch 10 Batch 150/173] avg loss 0.01002, throughput 12.7904K wps
Begin Testing...
[Epoch 10] train avg loss 0.0104133, test acc 0.7615, test avg loss 0.528145, throughput 12.9286K wps
Observed Improvement.
Begin Testing...
[Epoch 11 Batch 30/173] avg loss 0.0100983, throughput 13.1965K wps
[Epoch 11 Batch 60/173] avg loss 0.0100821, throughput 12.7935K wps
[Epoch 11 Batch 90/173] avg loss 0.0101261, throughput 12.7605K wps
[Epoch 11 Batch 120/173] avg loss 0.0101295, throughput 12.8178K wps
[Epoch 11 Batch 150/173] avg loss 0.0100571, throughput 12.9365K wps
Begin Testing...
[Epoch 11] train avg loss 0.0100712, test acc 0.7740, test avg loss 0.51464, throughput 12.8877K wps
Observed Improvement.
Begin Testing...
[Epoch 12 Batch 30/173] avg loss 0.00971092, throughput 13.3221K wps
[Epoch 12 Batch 60/173] avg loss 0.00952121, throughput 12.8056K wps
[Epoch 12 Batch 90/173] avg loss 0.0098299, throughput 12.9722K wps
[Epoch 12 Batch 120/173] avg loss 0.010008, throughput 12.9302K wps
[Epoch 12 Batch 150/173] avg loss 0.0100165, throughput 12.8192K wps
Begin Testing...
[Epoch 12] train avg loss 0.0098281, test acc 0.7604, test avg loss 0.504278, throughput 12.947K wps
[Epoch 13 Batch 30/173] avg loss 0.00913436, throughput 13.2397K wps
[Epoch 13 Batch 60/173] avg loss 0.00982822, throughput 12.8151K wps
[Epoch 13 Batch 90/173] avg loss 0.00935429, throughput 12.8329K wps
[Epoch 13 Batch 120/173] avg loss 0.00945847, throughput 12.958K wps
[Epoch 13 Batch 150/173] avg loss 0.00926425, throughput 12.9709K wps
Begin Testing...
[Epoch 13] train avg loss 0.00944529, test acc 0.7729, test avg loss 0.4887, throughput 12.9635K wps
[Epoch 14 Batch 30/173] avg loss 0.00919779, throughput 13.2541K wps
[Epoch 14 Batch 60/173] avg loss 0.00934159, throughput 12.7802K wps
[Epoch 14 Batch 90/173] avg loss 0.009223, throughput 12.9636K wps
[Epoch 14 Batch 120/173] avg loss 0.00939514, throughput 12.8223K wps
[Epoch 14 Batch 150/173] avg loss 0.00912786, throughput 12.8357K wps
Begin Testing...
[Epoch 14] train avg loss 0.00927313, test acc 0.7729, test avg loss 0.480575, throughput 12.9308K wps
[Epoch 15 Batch 30/173] avg loss 0.00900165, throughput 13.2641K wps
[Epoch 15 Batch 60/173] avg loss 0.00893564, throughput 12.8473K wps
[Epoch 15 Batch 90/173] avg loss 0.0091121, throughput 12.9742K wps
[Epoch 15 Batch 120/173] avg loss 0.00904362, throughput 12.8151K wps
[Epoch 15 Batch 150/173] avg loss 0.0090722, throughput 12.8022K wps
Begin Testing...
[Epoch 15] train avg loss 0.00906418, test acc 0.7823, test avg loss 0.474472, throughput 12.946K wps
Observed Improvement.
Begin Testing...
[Epoch 16 Batch 30/173] avg loss 0.00873811, throughput 13.3145K wps
[Epoch 16 Batch 60/173] avg loss 0.0090792, throughput 12.7916K wps
[Epoch 16 Batch 90/173] avg loss 0.00889637, throughput 12.9429K wps
[Epoch 16 Batch 120/173] avg loss 0.00859402, throughput 12.9482K wps
[Epoch 16 Batch 150/173] avg loss 0.00875199, throughput 12.9388K wps
Begin Testing...
[Epoch 16] train avg loss 0.00881051, test acc 0.7792, test avg loss 0.466061, throughput 12.9858K wps
[Epoch 17 Batch 30/173] avg loss 0.00866418, throughput 13.1929K wps
[Epoch 17 Batch 60/173] avg loss 0.00824751, throughput 12.8226K wps
[Epoch 17 Batch 90/173] avg loss 0.00890821, throughput 12.9122K wps
[Epoch 17 Batch 120/173] avg loss 0.00867075, throughput 12.9557K wps
[Epoch 17 Batch 150/173] avg loss 0.0083999, throughput 12.8122K wps
Begin Testing...
[Epoch 17] train avg loss 0.0086424, test acc 0.7823, test avg loss 0.465273, throughput 12.933K wps
Observed Improvement.
Begin Testing...
[Epoch 18 Batch 30/173] avg loss 0.00851518, throughput 13.337K wps
[Epoch 18 Batch 60/173] avg loss 0.00809266, throughput 12.8099K wps
[Epoch 18 Batch 90/173] avg loss 0.00855018, throughput 12.8557K wps
[Epoch 18 Batch 120/173] avg loss 0.0085201, throughput 12.9575K wps
[Epoch 18 Batch 150/173] avg loss 0.00836847, throughput 12.9012K wps
Begin Testing...
[Epoch 18] train avg loss 0.00839971, test acc 0.7823, test avg loss 0.453696, throughput 12.9512K wps
Observed Improvement.
Begin Testing...
[Epoch 19 Batch 30/173] avg loss 0.00803144, throughput 13.2699K wps
[Epoch 19 Batch 60/173] avg loss 0.00847641, throughput 12.7848K wps
[Epoch 19 Batch 90/173] avg loss 0.00808634, throughput 12.9385K wps
[Epoch 19 Batch 120/173] avg loss 0.00784241, throughput 12.9134K wps
[Epoch 19 Batch 150/173] avg loss 0.00829465, throughput 12.9248K wps
Begin Testing...
[Epoch 19] train avg loss 0.00820616, test acc 0.7844, test avg loss 0.449122, throughput 12.969K wps
Observed Improvement.
Begin Testing...
[Epoch 20 Batch 30/173] avg loss 0.00826826, throughput 13.122K wps
[Epoch 20 Batch 60/173] avg loss 0.00791061, throughput 12.76K wps
[Epoch 20 Batch 90/173] avg loss 0.00788199, throughput 12.8988K wps
[Epoch 20 Batch 120/173] avg loss 0.00803085, throughput 12.7294K wps
[Epoch 20 Batch 150/173] avg loss 0.00819191, throughput 12.8635K wps
Begin Testing...
[Epoch 20] train avg loss 0.00808403, test acc 0.7792, test avg loss 0.448454, throughput 12.8797K wps
[Epoch 21 Batch 30/173] avg loss 0.00790113, throughput 13.2546K wps
[Epoch 21 Batch 60/173] avg loss 0.00811302, throughput 12.7785K wps
[Epoch 21 Batch 90/173] avg loss 0.00784821, throughput 12.8155K wps
[Epoch 21 Batch 120/173] avg loss 0.00792618, throughput 12.9477K wps
[Epoch 21 Batch 150/173] avg loss 0.00773243, throughput 12.782K wps
Begin Testing...
[Epoch 21] train avg loss 0.00793734, test acc 0.7844, test avg loss 0.44505, throughput 12.9015K wps
Observed Improvement.
Begin Testing...
[Epoch 22 Batch 30/173] avg loss 0.00766276, throughput 13.2854K wps
[Epoch 22 Batch 60/173] avg loss 0.00753772, throughput 12.8117K wps
[Epoch 22 Batch 90/173] avg loss 0.0079445, throughput 12.8116K wps
[Epoch 22 Batch 120/173] avg loss 0.00780572, throughput 12.7709K wps
[Epoch 22 Batch 150/173] avg loss 0.00788943, throughput 12.7951K wps
Begin Testing...
[Epoch 22] train avg loss 0.00779393, test acc 0.7917, test avg loss 0.440571, throughput 12.8816K wps
Observed Improvement.
Begin Testing...
[Epoch 23 Batch 30/173] avg loss 0.00756888, throughput 13.2657K wps
[Epoch 23 Batch 60/173] avg loss 0.00753843, throughput 12.7884K wps
[Epoch 23 Batch 90/173] avg loss 0.0076124, throughput 12.8499K wps
[Epoch 23 Batch 120/173] avg loss 0.00750785, throughput 12.8473K wps
[Epoch 23 Batch 150/173] avg loss 0.0074081, throughput 12.8521K wps
Begin Testing...
[Epoch 23] train avg loss 0.00761335, test acc 0.7948, test avg loss 0.435864, throughput 12.9044K wps
Observed Improvement.
Begin Testing...
[Epoch 24 Batch 30/173] avg loss 0.00739155, throughput 13.2986K wps
[Epoch 24 Batch 60/173] avg loss 0.00743935, throughput 12.8895K wps
[Epoch 24 Batch 90/173] avg loss 0.00755543, throughput 12.8754K wps
[Epoch 24 Batch 120/173] avg loss 0.00741736, throughput 12.7799K wps
[Epoch 24 Batch 150/173] avg loss 0.00756232, throughput 12.7733K wps
Begin Testing...
[Epoch 24] train avg loss 0.00746594, test acc 0.7896, test avg loss 0.435057, throughput 12.9171K wps
[Epoch 25 Batch 30/173] avg loss 0.00741538, throughput 13.2227K wps
[Epoch 25 Batch 60/173] avg loss 0.00697252, throughput 12.797K wps
[Epoch 25 Batch 90/173] avg loss 0.0073571, throughput 12.9081K wps
[Epoch 25 Batch 120/173] avg loss 0.00751855, throughput 12.8189K wps
[Epoch 25 Batch 150/173] avg loss 0.00740041, throughput 12.9736K wps
Begin Testing...
[Epoch 25] train avg loss 0.00731978, test acc 0.7885, test avg loss 0.431761, throughput 12.9255K wps
[Epoch 26 Batch 30/173] avg loss 0.0071519, throughput 13.2319K wps
[Epoch 26 Batch 60/173] avg loss 0.00717342, throughput 12.8276K wps
[Epoch 26 Batch 90/173] avg loss 0.00700779, throughput 12.9374K wps
[Epoch 26 Batch 120/173] avg loss 0.00750258, throughput 12.8456K wps
[Epoch 26 Batch 150/173] avg loss 0.0068568, throughput 12.9225K wps
Begin Testing...
[Epoch 26] train avg loss 0.00717795, test acc 0.7906, test avg loss 0.432609, throughput 12.9551K wps
[Epoch 27 Batch 30/173] avg loss 0.00742258, throughput 13.2065K wps
[Epoch 27 Batch 60/173] avg loss 0.007188, throughput 12.7871K wps
[Epoch 27 Batch 90/173] avg loss 0.00682556, throughput 12.8079K wps
[Epoch 27 Batch 120/173] avg loss 0.00702771, throughput 12.9205K wps
[Epoch 27 Batch 150/173] avg loss 0.00713378, throughput 12.8726K wps
Begin Testing...
[Epoch 27] train avg loss 0.00713119, test acc 0.7844, test avg loss 0.432339, throughput 12.8991K wps
[Epoch 28 Batch 30/173] avg loss 0.00704115, throughput 13.2629K wps
[Epoch 28 Batch 60/173] avg loss 0.0068311, throughput 12.7857K wps
[Epoch 28 Batch 90/173] avg loss 0.0067168, throughput 12.8457K wps
[Epoch 28 Batch 120/173] avg loss 0.00687707, throughput 12.8731K wps
[Epoch 28 Batch 150/173] avg loss 0.00709755, throughput 12.897K wps
Begin Testing...
[Epoch 28] train avg loss 0.00693436, test acc 0.7854, test avg loss 0.425984, throughput 12.9349K wps
[Epoch 29 Batch 30/173] avg loss 0.00689896, throughput 13.2017K wps
[Epoch 29 Batch 60/173] avg loss 0.00646123, throughput 12.7575K wps
[Epoch 29 Batch 90/173] avg loss 0.00683432, throughput 12.7253K wps
[Epoch 29 Batch 120/173] avg loss 0.00690353, throughput 12.7978K wps
[Epoch 29 Batch 150/173] avg loss 0.00686894, throughput 12.8264K wps
Begin Testing...
[Epoch 29] train avg loss 0.00680197, test acc 0.7865, test avg loss 0.426965, throughput 12.845K wps
[Epoch 30 Batch 30/173] avg loss 0.00700972, throughput 13.2627K wps
[Epoch 30 Batch 60/173] avg loss 0.00662128, throughput 12.7813K wps
[Epoch 30 Batch 90/173] avg loss 0.00635511, throughput 12.8685K wps
[Epoch 30 Batch 120/173] avg loss 0.00664798, throughput 12.8389K wps
[Epoch 30 Batch 150/173] avg loss 0.00648268, throughput 12.9327K wps
Begin Testing...
[Epoch 30] train avg loss 0.00665304, test acc 0.7854, test avg loss 0.430391, throughput 12.9452K wps
[Epoch 31 Batch 30/173] avg loss 0.00705923, throughput 13.2178K wps
[Epoch 31 Batch 60/173] avg loss 0.00633604, throughput 12.7629K wps
[Epoch 31 Batch 90/173] avg loss 0.00614708, throughput 12.776K wps
[Epoch 31 Batch 120/173] avg loss 0.00695019, throughput 12.8698K wps
[Epoch 31 Batch 150/173] avg loss 0.00671672, throughput 12.8506K wps
Begin Testing...
[Epoch 31] train avg loss 0.00660347, test acc 0.7896, test avg loss 0.425492, throughput 12.9005K wps
[Epoch 32 Batch 30/173] avg loss 0.00608872, throughput 13.3244K wps
[Epoch 32 Batch 60/173] avg loss 0.00633392, throughput 12.8601K wps
[Epoch 32 Batch 90/173] avg loss 0.00653973, throughput 12.9121K wps
[Epoch 32 Batch 120/173] avg loss 0.00675393, throughput 12.854K wps
[Epoch 32 Batch 150/173] avg loss 0.00681246, throughput 12.772K wps
Begin Testing...
[Epoch 32] train avg loss 0.00647918, test acc 0.7812, test avg loss 0.422077, throughput 12.9418K wps
[Epoch 33 Batch 30/173] avg loss 0.00626536, throughput 13.3265K wps
[Epoch 33 Batch 60/173] avg loss 0.00639422, throughput 12.86K wps
[Epoch 33 Batch 90/173] avg loss 0.00610759, throughput 12.8807K wps
[Epoch 33 Batch 120/173] avg loss 0.00640321, throughput 12.8882K wps
[Epoch 33 Batch 150/173] avg loss 0.00634219, throughput 12.9171K wps
Begin Testing...
[Epoch 33] train avg loss 0.006352, test acc 0.7937, test avg loss 0.419566, throughput 12.9701K wps
[Epoch 34 Batch 30/173] avg loss 0.00612503, throughput 13.2992K wps
[Epoch 34 Batch 60/173] avg loss 0.00659696, throughput 12.8162K wps
[Epoch 34 Batch 90/173] avg loss 0.00627384, throughput 12.9694K wps
[Epoch 34 Batch 120/173] avg loss 0.0063328, throughput 12.8866K wps
[Epoch 34 Batch 150/173] avg loss 0.00615513, throughput 12.8878K wps
Begin Testing...
[Epoch 34] train avg loss 0.00626921, test acc 0.7854, test avg loss 0.41892, throughput 12.9467K wps
[Epoch 35 Batch 30/173] avg loss 0.00612311, throughput 13.2833K wps
[Epoch 35 Batch 60/173] avg loss 0.00588545, throughput 12.7659K wps
[Epoch 35 Batch 90/173] avg loss 0.00609755, throughput 12.7991K wps
[Epoch 35 Batch 120/173] avg loss 0.00601561, throughput 12.9497K wps
[Epoch 35 Batch 150/173] avg loss 0.00625789, throughput 12.7786K wps
Begin Testing...
[Epoch 35] train avg loss 0.00605923, test acc 0.7885, test avg loss 0.424293, throughput 12.9096K wps
[Epoch 36 Batch 30/173] avg loss 0.00581697, throughput 13.1943K wps
[Epoch 36 Batch 60/173] avg loss 0.00633516, throughput 12.8192K wps
[Epoch 36 Batch 90/173] avg loss 0.0058833, throughput 12.9374K wps
[Epoch 36 Batch 120/173] avg loss 0.00587444, throughput 12.9664K wps
[Epoch 36 Batch 150/173] avg loss 0.00602645, throughput 12.8314K wps
Begin Testing...
[Epoch 36] train avg loss 0.00595829, test acc 0.7906, test avg loss 0.429912, throughput 12.9538K wps
[Epoch 37 Batch 30/173] avg loss 0.00533411, throughput 13.262K wps
[Epoch 37 Batch 60/173] avg loss 0.0059099, throughput 12.7982K wps
[Epoch 37 Batch 90/173] avg loss 0.00589036, throughput 12.9349K wps
[Epoch 37 Batch 120/173] avg loss 0.00581432, throughput 12.8208K wps
[Epoch 37 Batch 150/173] avg loss 0.00598414, throughput 12.7903K wps
Begin Testing...
[Epoch 37] train avg loss 0.00580998, test acc 0.7937, test avg loss 0.424072, throughput 12.9037K wps
[Epoch 38 Batch 30/173] avg loss 0.00557901, throughput 13.3174K wps
[Epoch 38 Batch 60/173] avg loss 0.00584059, throughput 12.8197K wps
[Epoch 38 Batch 90/173] avg loss 0.00554931, throughput 12.8789K wps
[Epoch 38 Batch 120/173] avg loss 0.0055937, throughput 12.8141K wps
[Epoch 38 Batch 150/173] avg loss 0.00580461, throughput 12.7865K wps
Begin Testing...
[Epoch 38] train avg loss 0.00572495, test acc 0.7917, test avg loss 0.421366, throughput 12.9182K wps
[Epoch 39 Batch 30/173] avg loss 0.00574616, throughput 13.2917K wps
[Epoch 39 Batch 60/173] avg loss 0.00569012, throughput 12.7887K wps
[Epoch 39 Batch 90/173] avg loss 0.00546517, throughput 12.787K wps
[Epoch 39 Batch 120/173] avg loss 0.0053932, throughput 12.9084K wps
[Epoch 39 Batch 150/173] avg loss 0.00581096, throughput 12.8487K wps
Begin Testing...
[Epoch 39] train avg loss 0.00559219, test acc 0.7927, test avg loss 0.41936, throughput 12.9187K wps
[Epoch 40 Batch 30/173] avg loss 0.00562782, throughput 13.2284K wps
[Epoch 40 Batch 60/173] avg loss 0.0055108, throughput 12.8214K wps
[Epoch 40 Batch 90/173] avg loss 0.00540169, throughput 12.8053K wps
[Epoch 40 Batch 120/173] avg loss 0.00538184, throughput 12.9066K wps
[Epoch 40 Batch 150/173] avg loss 0.00567139, throughput 12.8113K wps
Begin Testing...
[Epoch 40] train avg loss 0.0055152, test acc 0.7906, test avg loss 0.422004, throughput 12.9037K wps
[Epoch 41 Batch 30/173] avg loss 0.00553076, throughput 13.2249K wps
[Epoch 41 Batch 60/173] avg loss 0.00530168, throughput 12.7473K wps
[Epoch 41 Batch 90/173] avg loss 0.00526591, throughput 12.8182K wps
[Epoch 41 Batch 120/173] avg loss 0.00543891, throughput 12.8159K wps
[Epoch 41 Batch 150/173] avg loss 0.00508619, throughput 12.8535K wps
Begin Testing...
[Epoch 41] train avg loss 0.00539676, test acc 0.7865, test avg loss 0.417986, throughput 12.902K wps
[Epoch 42 Batch 30/173] avg loss 0.00535341, throughput 13.2325K wps
[Epoch 42 Batch 60/173] avg loss 0.00523547, throughput 12.8278K wps
[Epoch 42 Batch 90/173] avg loss 0.00507757, throughput 12.8166K wps
[Epoch 42 Batch 120/173] avg loss 0.005078, throughput 12.9296K wps
[Epoch 42 Batch 150/173] avg loss 0.00564022, throughput 12.7423K wps
Begin Testing...
[Epoch 42] train avg loss 0.0052853, test acc 0.7896, test avg loss 0.421648, throughput 12.9149K wps
[Epoch 43 Batch 30/173] avg loss 0.0052969, throughput 13.2328K wps
[Epoch 43 Batch 60/173] avg loss 0.00508022, throughput 12.7755K wps
[Epoch 43 Batch 90/173] avg loss 0.00490712, throughput 12.8543K wps
[Epoch 43 Batch 120/173] avg loss 0.00516156, throughput 12.971K wps
[Epoch 43 Batch 150/173] avg loss 0.00523948, throughput 12.8693K wps
Begin Testing...
[Epoch 43] train avg loss 0.00519403, test acc 0.7885, test avg loss 0.423997, throughput 12.9259K wps
[Epoch 44 Batch 30/173] avg loss 0.00526478, throughput 13.2727K wps
[Epoch 44 Batch 60/173] avg loss 0.00487546, throughput 12.7451K wps
[Epoch 44 Batch 90/173] avg loss 0.00511545, throughput 12.9153K wps
[Epoch 44 Batch 120/173] avg loss 0.00480904, throughput 12.7922K wps
[Epoch 44 Batch 150/173] avg loss 0.0051406, throughput 12.9103K wps
Begin Testing...
[Epoch 44] train avg loss 0.00510149, test acc 0.7937, test avg loss 0.425254, throughput 12.9331K wps
[Epoch 45 Batch 30/173] avg loss 0.0046878, throughput 13.2123K wps
[Epoch 45 Batch 60/173] avg loss 0.00501943, throughput 12.7987K wps
[Epoch 45 Batch 90/173] avg loss 0.00493449, throughput 12.9556K wps
[Epoch 45 Batch 120/173] avg loss 0.00482693, throughput 12.89K wps
[Epoch 45 Batch 150/173] avg loss 0.00495282, throughput 12.9156K wps
Begin Testing...
[Epoch 45] train avg loss 0.00497996, test acc 0.7917, test avg loss 0.421913, throughput 12.9576K wps
[Epoch 46 Batch 30/173] avg loss 0.00494001, throughput 13.2062K wps
[Epoch 46 Batch 60/173] avg loss 0.00479662, throughput 12.8442K wps
[Epoch 46 Batch 90/173] avg loss 0.0046663, throughput 12.926K wps
[Epoch 46 Batch 120/173] avg loss 0.00484836, throughput 12.9555K wps
[Epoch 46 Batch 150/173] avg loss 0.00522013, throughput 12.8394K wps
Begin Testing...
[Epoch 46] train avg loss 0.00488532, test acc 0.7906, test avg loss 0.422629, throughput 12.9556K wps
[Epoch 47 Batch 30/173] avg loss 0.00435781, throughput 13.2954K wps
[Epoch 47 Batch 60/173] avg loss 0.00505179, throughput 12.8334K wps
[Epoch 47 Batch 90/173] avg loss 0.00483855, throughput 12.8215K wps
[Epoch 47 Batch 120/173] avg loss 0.00494631, throughput 12.8079K wps
[Epoch 47 Batch 150/173] avg loss 0.00492982, throughput 12.8955K wps
Begin Testing...
[Epoch 47] train avg loss 0.00479879, test acc 0.8010, test avg loss 0.416351, throughput 12.9263K wps
Observed Improvement.
Begin Testing...
[Epoch 48 Batch 30/173] avg loss 0.00491343, throughput 13.2903K wps
[Epoch 48 Batch 60/173] avg loss 0.00456172, throughput 12.8246K wps
[Epoch 48 Batch 90/173] avg loss 0.00460143, throughput 12.9666K wps
[Epoch 48 Batch 120/173] avg loss 0.00469605, throughput 12.8217K wps
[Epoch 48 Batch 150/173] avg loss 0.00474427, throughput 12.9772K wps
Begin Testing...
[Epoch 48] train avg loss 0.0047333, test acc 0.7948, test avg loss 0.424954, throughput 12.9695K wps
[Epoch 49 Batch 30/173] avg loss 0.00458756, throughput 13.2055K wps
[Epoch 49 Batch 60/173] avg loss 0.00444615, throughput 12.8158K wps
[Epoch 49 Batch 90/173] avg loss 0.00454764, throughput 12.9256K wps
[Epoch 49 Batch 120/173] avg loss 0.00432894, throughput 12.9449K wps
[Epoch 49 Batch 150/173] avg loss 0.00447933, throughput 12.9203K wps
Begin Testing...
[Epoch 49] train avg loss 0.00452895, test acc 0.7990, test avg loss 0.420166, throughput 12.9624K wps
[Epoch 50 Batch 30/173] avg loss 0.0042275, throughput 13.2024K wps
[Epoch 50 Batch 60/173] avg loss 0.004532, throughput 12.8659K wps
[Epoch 50 Batch 90/173] avg loss 0.00457122, throughput 12.957K wps
[Epoch 50 Batch 120/173] avg loss 0.0045985, throughput 12.8471K wps
[Epoch 50 Batch 150/173] avg loss 0.0046207, throughput 12.8429K wps
Begin Testing...
[Epoch 50] train avg loss 0.00451995, test acc 0.7937, test avg loss 0.42857, throughput 12.9466K wps
[Epoch 51 Batch 30/173] avg loss 0.00385433, throughput 13.245K wps
[Epoch 51 Batch 60/173] avg loss 0.00459687, throughput 12.7748K wps
[Epoch 51 Batch 90/173] avg loss 0.00437944, throughput 12.8811K wps
[Epoch 51 Batch 120/173] avg loss 0.00460732, throughput 12.8162K wps
[Epoch 51 Batch 150/173] avg loss 0.00454251, throughput 12.862K wps
Begin Testing...
[Epoch 51] train avg loss 0.00441601, test acc 0.7896, test avg loss 0.424704, throughput 12.9155K wps
[Epoch 52 Batch 30/173] avg loss 0.00429883, throughput 13.2159K wps
[Epoch 52 Batch 60/173] avg loss 0.00409457, throughput 12.8819K wps
[Epoch 52 Batch 90/173] avg loss 0.00446392, throughput 12.928K wps
[Epoch 52 Batch 120/173] avg loss 0.00403176, throughput 12.9593K wps
[Epoch 52 Batch 150/173] avg loss 0.00428642, throughput 12.922K wps
Begin Testing...
[Epoch 52] train avg loss 0.00430443, test acc 0.7896, test avg loss 0.430633, throughput 12.977K wps
[Epoch 53 Batch 30/173] avg loss 0.00414652, throughput 13.1667K wps
[Epoch 53 Batch 60/173] avg loss 0.0047178, throughput 12.7772K wps
[Epoch 53 Batch 90/173] avg loss 0.00421683, throughput 12.9149K wps
[Epoch 53 Batch 120/173] avg loss 0.00456702, throughput 12.9409K wps
[Epoch 53 Batch 150/173] avg loss 0.0042151, throughput 12.9268K wps
Begin Testing...
[Epoch 53] train avg loss 0.00436294, test acc 0.7990, test avg loss 0.423607, throughput 12.9508K wps
[Epoch 54 Batch 30/173] avg loss 0.00388598, throughput 13.2413K wps
[Epoch 54 Batch 60/173] avg loss 0.00432001, throughput 12.7891K wps
[Epoch 54 Batch 90/173] avg loss 0.00397464, throughput 12.9794K wps
[Epoch 54 Batch 120/173] avg loss 0.00417503, throughput 12.9625K wps
[Epoch 54 Batch 150/173] avg loss 0.00417777, throughput 12.966K wps
Begin Testing...
[Epoch 54] train avg loss 0.00415625, test acc 0.7969, test avg loss 0.431825, throughput 12.989K wps
[Epoch 55 Batch 30/173] avg loss 0.00402769, throughput 13.1281K wps
[Epoch 55 Batch 60/173] avg loss 0.00408693, throughput 12.7986K wps
[Epoch 55 Batch 90/173] avg loss 0.00385873, throughput 12.9477K wps
[Epoch 55 Batch 120/173] avg loss 0.0039918, throughput 12.9187K wps
[Epoch 55 Batch 150/173] avg loss 0.00413231, throughput 12.941K wps
Begin Testing...
[Epoch 55] train avg loss 0.00405402, test acc 0.8000, test avg loss 0.425602, throughput 12.9327K wps
[Epoch 56 Batch 30/173] avg loss 0.00409554, throughput 13.2707K wps
[Epoch 56 Batch 60/173] avg loss 0.00383233, throughput 12.7801K wps
[Epoch 56 Batch 90/173] avg loss 0.00380897, throughput 12.9636K wps
[Epoch 56 Batch 120/173] avg loss 0.00389721, throughput 12.8357K wps
[Epoch 56 Batch 150/173] avg loss 0.00404535, throughput 12.9642K wps
Begin Testing...
[Epoch 56] train avg loss 0.00397028, test acc 0.7958, test avg loss 0.423431, throughput 12.965K wps
[Epoch 57 Batch 30/173] avg loss 0.00377483, throughput 13.1917K wps
[Epoch 57 Batch 60/173] avg loss 0.00373187, throughput 12.7843K wps
[Epoch 57 Batch 90/173] avg loss 0.00382123, throughput 12.7978K wps
[Epoch 57 Batch 120/173] avg loss 0.00396672, throughput 12.8221K wps
[Epoch 57 Batch 150/173] avg loss 0.00373263, throughput 12.9618K wps
Begin Testing...
[Epoch 57] train avg loss 0.00383055, test acc 0.7927, test avg loss 0.428831, throughput 12.9199K wps
[Epoch 58 Batch 30/173] avg loss 0.00398798, throughput 13.2398K wps
[Epoch 58 Batch 60/173] avg loss 0.00367449, throughput 12.7803K wps
[Epoch 58 Batch 90/173] avg loss 0.00350762, throughput 12.8707K wps
[Epoch 58 Batch 120/173] avg loss 0.00382455, throughput 12.8869K wps
[Epoch 58 Batch 150/173] avg loss 0.00363564, throughput 12.7739K wps
Begin Testing...
[Epoch 58] train avg loss 0.00373173, test acc 0.7969, test avg loss 0.426456, throughput 12.911K wps
[Epoch 59 Batch 30/173] avg loss 0.00380517, throughput 13.1492K wps
[Epoch 59 Batch 60/173] avg loss 0.00378381, throughput 12.7577K wps
[Epoch 59 Batch 90/173] avg loss 0.00357442, throughput 12.8699K wps
[Epoch 59 Batch 120/173] avg loss 0.00362383, throughput 12.8182K wps
[Epoch 59 Batch 150/173] avg loss 0.00386244, throughput 12.7823K wps
Begin Testing...
[Epoch 59] train avg loss 0.00370645, test acc 0.7979, test avg loss 0.425792, throughput 12.864K wps
Test loss 0.459335, test acc 0.7917
Total time cost 175.35s
9596 1066
[Epoch 0 Batch 30/173] avg loss 0.0157419, throughput 11.6219K wps
[Epoch 0 Batch 60/173] avg loss 0.015423, throughput 12.7828K wps
[Epoch 0 Batch 90/173] avg loss 0.0149993, throughput 12.7851K wps
[Epoch 0 Batch 120/173] avg loss 0.0148328, throughput 12.8433K wps
[Epoch 0 Batch 150/173] avg loss 0.0143967, throughput 12.8006K wps
Begin Testing...
[Epoch 0] train avg loss 0.015012, test acc 0.5552, test avg loss 0.685852, throughput 12.578K wps
Observed Improvement.
Begin Testing...
[Epoch 1 Batch 30/173] avg loss 0.0141715, throughput 13.1596K wps
[Epoch 1 Batch 60/173] avg loss 0.0141194, throughput 12.7671K wps
[Epoch 1 Batch 90/173] avg loss 0.0136524, throughput 12.8092K wps
[Epoch 1 Batch 120/173] avg loss 0.013962, throughput 12.8015K wps
[Epoch 1 Batch 150/173] avg loss 0.0136764, throughput 12.7577K wps
Begin Testing...
[Epoch 1] train avg loss 0.0138641, test acc 0.5646, test avg loss 0.666483, throughput 12.8522K wps
Observed Improvement.
Begin Testing...
[Epoch 2 Batch 30/173] avg loss 0.0133367, throughput 13.2714K wps
[Epoch 2 Batch 60/173] avg loss 0.0134424, throughput 12.808K wps
[Epoch 2 Batch 90/173] avg loss 0.0132265, throughput 12.7972K wps
[Epoch 2 Batch 120/173] avg loss 0.0130401, throughput 12.7949K wps
[Epoch 2 Batch 150/173] avg loss 0.0131003, throughput 12.7986K wps
Begin Testing...
[Epoch 2] train avg loss 0.013236, test acc 0.5979, test avg loss 0.651043, throughput 12.8817K wps
Observed Improvement.
Begin Testing...
[Epoch 3 Batch 30/173] avg loss 0.0128378, throughput 13.2248K wps
[Epoch 3 Batch 60/173] avg loss 0.0130178, throughput 12.8452K wps
[Epoch 3 Batch 90/173] avg loss 0.0129827, throughput 12.8259K wps
[Epoch 3 Batch 120/173] avg loss 0.0126713, throughput 12.7642K wps
[Epoch 3 Batch 150/173] avg loss 0.0127506, throughput 12.7792K wps
Begin Testing...
[Epoch 3] train avg loss 0.0128077, test acc 0.6219, test avg loss 0.634731, throughput 12.8781K wps
Observed Improvement.
Begin Testing...
[Epoch 4 Batch 30/173] avg loss 0.0123354, throughput 13.1644K wps
[Epoch 4 Batch 60/173] avg loss 0.0123361, throughput 12.8315K wps
[Epoch 4 Batch 90/173] avg loss 0.0124461, throughput 12.8348K wps
[Epoch 4 Batch 120/173] avg loss 0.0123376, throughput 12.7487K wps
[Epoch 4 Batch 150/173] avg loss 0.0124635, throughput 12.7772K wps
Begin Testing...
[Epoch 4] train avg loss 0.012376, test acc 0.6375, test avg loss 0.619504, throughput 12.8622K wps
Observed Improvement.
Begin Testing...
[Epoch 5 Batch 30/173] avg loss 0.0122453, throughput 13.2296K wps
[Epoch 5 Batch 60/173] avg loss 0.0118571, throughput 12.6557K wps
[Epoch 5 Batch 90/173] avg loss 0.0119882, throughput 12.8701K wps
[Epoch 5 Batch 120/173] avg loss 0.0119003, throughput 12.7805K wps
[Epoch 5 Batch 150/173] avg loss 0.0118624, throughput 12.7871K wps
Begin Testing...
[Epoch 5] train avg loss 0.0119893, test acc 0.6844, test avg loss 0.598291, throughput 12.8565K wps
Observed Improvement.
Begin Testing...
[Epoch 6 Batch 30/173] avg loss 0.0115996, throughput 13.2566K wps
[Epoch 6 Batch 60/173] avg loss 0.011916, throughput 12.8432K wps
[Epoch 6 Batch 90/173] avg loss 0.0116318, throughput 12.8979K wps
[Epoch 6 Batch 120/173] avg loss 0.0113005, throughput 12.8191K wps
[Epoch 6 Batch 150/173] avg loss 0.0118678, throughput 12.7888K wps
Begin Testing...
[Epoch 6] train avg loss 0.011615, test acc 0.6875, test avg loss 0.585978, throughput 12.9372K wps
Observed Improvement.
Begin Testing...
[Epoch 7 Batch 30/173] avg loss 0.0114706, throughput 13.2261K wps
[Epoch 7 Batch 60/173] avg loss 0.0115463, throughput 12.812K wps
[Epoch 7 Batch 90/173] avg loss 0.0111877, throughput 12.8936K wps
[Epoch 7 Batch 120/173] avg loss 0.0113131, throughput 12.8491K wps
[Epoch 7 Batch 150/173] avg loss 0.0114477, throughput 12.8315K wps
Begin Testing...
[Epoch 7] train avg loss 0.0113817, test acc 0.7198, test avg loss 0.568895, throughput 12.932K wps
Observed Improvement.
Begin Testing...
[Epoch 8 Batch 30/173] avg loss 0.0109364, throughput 13.244K wps
[Epoch 8 Batch 60/173] avg loss 0.011156, throughput 12.8001K wps
[Epoch 8 Batch 90/173] avg loss 0.0109696, throughput 12.8213K wps
[Epoch 8 Batch 120/173] avg loss 0.0109624, throughput 12.8165K wps
[Epoch 8 Batch 150/173] avg loss 0.0108862, throughput 12.7883K wps
Begin Testing...
[Epoch 8] train avg loss 0.0109941, test acc 0.7406, test avg loss 0.552702, throughput 12.8797K wps
Observed Improvement.
Begin Testing...
[Epoch 9 Batch 30/173] avg loss 0.0105783, throughput 13.3134K wps
[Epoch 9 Batch 60/173] avg loss 0.0106385, throughput 12.8887K wps
[Epoch 9 Batch 90/173] avg loss 0.0103431, throughput 12.8493K wps
[Epoch 9 Batch 120/173] avg loss 0.0107677, throughput 12.7988K wps
[Epoch 9 Batch 150/173] avg loss 0.0106698, throughput 12.954K wps
Begin Testing...
[Epoch 9] train avg loss 0.0106211, test acc 0.7448, test avg loss 0.538908, throughput 12.9487K wps
Observed Improvement.
Begin Testing...
[Epoch 10 Batch 30/173] avg loss 0.0100922, throughput 13.2106K wps
[Epoch 10 Batch 60/173] avg loss 0.0104273, throughput 12.7958K wps
[Epoch 10 Batch 90/173] avg loss 0.010299, throughput 12.9792K wps
[Epoch 10 Batch 120/173] avg loss 0.0101252, throughput 12.7718K wps
[Epoch 10 Batch 150/173] avg loss 0.0101273, throughput 12.9153K wps
Begin Testing...
[Epoch 10] train avg loss 0.0102461, test acc 0.7760, test avg loss 0.526335, throughput 12.9296K wps
Observed Improvement.
Begin Testing...
[Epoch 11 Batch 30/173] avg loss 0.0101365, throughput 13.2666K wps
[Epoch 11 Batch 60/173] avg loss 0.010009, throughput 12.8082K wps
[Epoch 11 Batch 90/173] avg loss 0.00991723, throughput 12.8538K wps
[Epoch 11 Batch 120/173] avg loss 0.00998297, throughput 12.9295K wps
[Epoch 11 Batch 150/173] avg loss 0.00991425, throughput 12.7988K wps
Begin Testing...
[Epoch 11] train avg loss 0.0100031, test acc 0.7656, test avg loss 0.513972, throughput 12.9118K wps
[Epoch 12 Batch 30/173] avg loss 0.00971923, throughput 13.1827K wps
[Epoch 12 Batch 60/173] avg loss 0.00989818, throughput 12.787K wps
[Epoch 12 Batch 90/173] avg loss 0.00965192, throughput 12.9492K wps
[Epoch 12 Batch 120/173] avg loss 0.00959229, throughput 12.9916K wps
[Epoch 12 Batch 150/173] avg loss 0.00940998, throughput 12.8598K wps
Begin Testing...
[Epoch 12] train avg loss 0.00963714, test acc 0.7719, test avg loss 0.499148, throughput 12.9339K wps
[Epoch 13 Batch 30/173] avg loss 0.00942034, throughput 13.1954K wps
[Epoch 13 Batch 60/173] avg loss 0.00929509, throughput 12.7744K wps
[Epoch 13 Batch 90/173] avg loss 0.00967072, throughput 12.8241K wps
[Epoch 13 Batch 120/173] avg loss 0.00966481, throughput 12.8659K wps
[Epoch 13 Batch 150/173] avg loss 0.00930514, throughput 12.7527K wps
Begin Testing...
[Epoch 13] train avg loss 0.00941907, test acc 0.7792, test avg loss 0.496771, throughput 12.867K wps
Observed Improvement.
Begin Testing...
[Epoch 14 Batch 30/173] avg loss 0.00937524, throughput 13.2028K wps
[Epoch 14 Batch 60/173] avg loss 0.00892572, throughput 12.7664K wps
[Epoch 14 Batch 90/173] avg loss 0.00931387, throughput 12.7455K wps
[Epoch 14 Batch 120/173] avg loss 0.00912669, throughput 12.9099K wps
[Epoch 14 Batch 150/173] avg loss 0.00915253, throughput 12.8089K wps
Begin Testing...
[Epoch 14] train avg loss 0.00921035, test acc 0.7719, test avg loss 0.482724, throughput 12.8941K wps
[Epoch 15 Batch 30/173] avg loss 0.00902089, throughput 13.2485K wps
[Epoch 15 Batch 60/173] avg loss 0.00896838, throughput 12.788K wps
[Epoch 15 Batch 90/173] avg loss 0.00899948, throughput 12.9543K wps
[Epoch 15 Batch 120/173] avg loss 0.00896729, throughput 12.8714K wps
[Epoch 15 Batch 150/173] avg loss 0.00917391, throughput 12.9097K wps
Begin Testing...
[Epoch 15] train avg loss 0.008986, test acc 0.7729, test avg loss 0.477935, throughput 12.9326K wps
[Epoch 16 Batch 30/173] avg loss 0.00874842, throughput 13.2871K wps
[Epoch 16 Batch 60/173] avg loss 0.00851943, throughput 12.7947K wps
[Epoch 16 Batch 90/173] avg loss 0.00893272, throughput 12.8002K wps
[Epoch 16 Batch 120/173] avg loss 0.00907801, throughput 12.8062K wps
[Epoch 16 Batch 150/173] avg loss 0.00842872, throughput 12.8101K wps
Begin Testing...
[Epoch 16] train avg loss 0.00878606, test acc 0.7792, test avg loss 0.471384, throughput 12.9134K wps
Observed Improvement.
Begin Testing...
[Epoch 17 Batch 30/173] avg loss 0.00817488, throughput 13.2107K wps
[Epoch 17 Batch 60/173] avg loss 0.00862872, throughput 12.8327K wps
[Epoch 17 Batch 90/173] avg loss 0.00846964, throughput 12.9421K wps
[Epoch 17 Batch 120/173] avg loss 0.00858703, throughput 12.797K wps
[Epoch 17 Batch 150/173] avg loss 0.0086745, throughput 12.809K wps
Begin Testing...
[Epoch 17] train avg loss 0.0085637, test acc 0.7823, test avg loss 0.464742, throughput 12.8993K wps
Observed Improvement.
Begin Testing...
[Epoch 18 Batch 30/173] avg loss 0.00835282, throughput 13.2031K wps
[Epoch 18 Batch 60/173] avg loss 0.00846973, throughput 12.842K wps
[Epoch 18 Batch 90/173] avg loss 0.00848814, throughput 12.8724K wps
[Epoch 18 Batch 120/173] avg loss 0.00818899, throughput 12.8224K wps
[Epoch 18 Batch 150/173] avg loss 0.00845082, throughput 12.8248K wps
Begin Testing...
[Epoch 18] train avg loss 0.00837989, test acc 0.7937, test avg loss 0.461506, throughput 12.9001K wps
Observed Improvement.
Begin Testing...
[Epoch 19 Batch 30/173] avg loss 0.00817169, throughput 13.2901K wps
[Epoch 19 Batch 60/173] avg loss 0.00834736, throughput 12.8412K wps
[Epoch 19 Batch 90/173] avg loss 0.00801462, throughput 12.9598K wps
[Epoch 19 Batch 120/173] avg loss 0.00824782, throughput 12.8015K wps
[Epoch 19 Batch 150/173] avg loss 0.00840555, throughput 12.883K wps
Begin Testing...
[Epoch 19] train avg loss 0.00819717, test acc 0.7812, test avg loss 0.454376, throughput 12.9376K wps
[Epoch 20 Batch 30/173] avg loss 0.00793182, throughput 13.2613K wps
[Epoch 20 Batch 60/173] avg loss 0.00790392, throughput 12.8997K wps
[Epoch 20 Batch 90/173] avg loss 0.00769655, throughput 12.8264K wps
[Epoch 20 Batch 120/173] avg loss 0.00820704, throughput 12.8938K wps
[Epoch 20 Batch 150/173] avg loss 0.00828881, throughput 12.8042K wps
Begin Testing...
[Epoch 20] train avg loss 0.00797724, test acc 0.7833, test avg loss 0.454797, throughput 12.92K wps
[Epoch 21 Batch 30/173] avg loss 0.00766252, throughput 13.2008K wps
[Epoch 21 Batch 60/173] avg loss 0.00794432, throughput 12.7587K wps
[Epoch 21 Batch 90/173] avg loss 0.00786586, throughput 12.8851K wps
[Epoch 21 Batch 120/173] avg loss 0.00787014, throughput 12.8529K wps
[Epoch 21 Batch 150/173] avg loss 0.00797825, throughput 12.9545K wps
Begin Testing...
[Epoch 21] train avg loss 0.00791431, test acc 0.7937, test avg loss 0.450426, throughput 12.9157K wps
Observed Improvement.
Begin Testing...
[Epoch 22 Batch 30/173] avg loss 0.00765097, throughput 13.2232K wps
[Epoch 22 Batch 60/173] avg loss 0.00777245, throughput 12.8286K wps
[Epoch 22 Batch 90/173] avg loss 0.00786165, throughput 12.7652K wps
[Epoch 22 Batch 120/173] avg loss 0.00767917, throughput 12.8909K wps
[Epoch 22 Batch 150/173] avg loss 0.00783889, throughput 12.8407K wps
Begin Testing...
[Epoch 22] train avg loss 0.00772701, test acc 0.7937, test avg loss 0.445797, throughput 12.9016K wps
Observed Improvement.
Begin Testing...
[Epoch 23 Batch 30/173] avg loss 0.00732942, throughput 13.2595K wps
[Epoch 23 Batch 60/173] avg loss 0.00734097, throughput 12.8435K wps
[Epoch 23 Batch 90/173] avg loss 0.00744604, throughput 12.9603K wps
[Epoch 23 Batch 120/173] avg loss 0.00753529, throughput 12.9504K wps
[Epoch 23 Batch 150/173] avg loss 0.00751726, throughput 12.9759K wps
Begin Testing...
[Epoch 23] train avg loss 0.00747149, test acc 0.7812, test avg loss 0.448726, throughput 12.9788K wps
[Epoch 24 Batch 30/173] avg loss 0.00752979, throughput 13.253K wps
[Epoch 24 Batch 60/173] avg loss 0.00737198, throughput 12.7924K wps
[Epoch 24 Batch 90/173] avg loss 0.00696378, throughput 12.8057K wps
[Epoch 24 Batch 120/173] avg loss 0.00730442, throughput 12.8211K wps
[Epoch 24 Batch 150/173] avg loss 0.00766971, throughput 12.8344K wps
Begin Testing...
[Epoch 24] train avg loss 0.00737829, test acc 0.7958, test avg loss 0.442379, throughput 12.9116K wps
Observed Improvement.
Begin Testing...
[Epoch 25 Batch 30/173] avg loss 0.00732694, throughput 13.2694K wps
[Epoch 25 Batch 60/173] avg loss 0.00691747, throughput 12.8335K wps
[Epoch 25 Batch 90/173] avg loss 0.00758362, throughput 12.9549K wps
[Epoch 25 Batch 120/173] avg loss 0.00695614, throughput 12.8712K wps
[Epoch 25 Batch 150/173] avg loss 0.00720916, throughput 12.9069K wps
Begin Testing...
[Epoch 25] train avg loss 0.00725387, test acc 0.7927, test avg loss 0.439659, throughput 12.971K wps
[Epoch 26 Batch 30/173] avg loss 0.00726395, throughput 13.2341K wps
[Epoch 26 Batch 60/173] avg loss 0.00720293, throughput 12.8156K wps
[Epoch 26 Batch 90/173] avg loss 0.00727967, throughput 12.9416K wps
[Epoch 26 Batch 120/173] avg loss 0.00725806, throughput 12.9554K wps
[Epoch 26 Batch 150/173] avg loss 0.00742731, throughput 12.9559K wps
Begin Testing...
[Epoch 26] train avg loss 0.00720092, test acc 0.7833, test avg loss 0.438388, throughput 12.961K wps
[Epoch 27 Batch 30/173] avg loss 0.00702161, throughput 13.2635K wps
[Epoch 27 Batch 60/173] avg loss 0.00704422, throughput 12.8339K wps
[Epoch 27 Batch 90/173] avg loss 0.00667444, throughput 12.7552K wps
[Epoch 27 Batch 120/173] avg loss 0.00666501, throughput 12.963K wps
[Epoch 27 Batch 150/173] avg loss 0.00695533, throughput 12.9779K wps
Begin Testing...
[Epoch 27] train avg loss 0.00692961, test acc 0.7885, test avg loss 0.435459, throughput 12.9614K wps
[Epoch 28 Batch 30/173] avg loss 0.00673614, throughput 13.2671K wps
[Epoch 28 Batch 60/173] avg loss 0.00683919, throughput 12.7923K wps
[Epoch 28 Batch 90/173] avg loss 0.00701582, throughput 12.9036K wps
[Epoch 28 Batch 120/173] avg loss 0.00672866, throughput 12.9741K wps
[Epoch 28 Batch 150/173] avg loss 0.00686219, throughput 12.9653K wps
Begin Testing...
[Epoch 28] train avg loss 0.0068603, test acc 0.8000, test avg loss 0.433169, throughput 12.9815K wps
Observed Improvement.
Begin Testing...
[Epoch 29 Batch 30/173] avg loss 0.00657978, throughput 13.2977K wps
[Epoch 29 Batch 60/173] avg loss 0.0067534, throughput 12.7447K wps
[Epoch 29 Batch 90/173] avg loss 0.00661849, throughput 12.7751K wps
[Epoch 29 Batch 120/173] avg loss 0.00685296, throughput 12.932K wps
[Epoch 29 Batch 150/173] avg loss 0.00681584, throughput 12.9312K wps
Begin Testing...
[Epoch 29] train avg loss 0.00670338, test acc 0.8000, test avg loss 0.430814, throughput 12.9428K wps
Observed Improvement.
Begin Testing...
[Epoch 30 Batch 30/173] avg loss 0.00647898, throughput 13.2542K wps
[Epoch 30 Batch 60/173] avg loss 0.00644008, throughput 12.8295K wps
[Epoch 30 Batch 90/173] avg loss 0.00657038, throughput 12.9895K wps
[Epoch 30 Batch 120/173] avg loss 0.00643756, throughput 12.8709K wps
[Epoch 30 Batch 150/173] avg loss 0.00677817, throughput 12.8694K wps
Begin Testing...
[Epoch 30] train avg loss 0.00654136, test acc 0.7937, test avg loss 0.431011, throughput 12.9509K wps
[Epoch 31 Batch 30/173] avg loss 0.00643583, throughput 13.0846K wps
[Epoch 31 Batch 60/173] avg loss 0.00643495, throughput 12.7392K wps
[Epoch 31 Batch 90/173] avg loss 0.00670903, throughput 12.8397K wps
[Epoch 31 Batch 120/173] avg loss 0.00664973, throughput 12.9441K wps
[Epoch 31 Batch 150/173] avg loss 0.00610118, throughput 12.9615K wps
Begin Testing...
[Epoch 31] train avg loss 0.00653599, test acc 0.7948, test avg loss 0.432505, throughput 12.927K wps
[Epoch 32 Batch 30/173] avg loss 0.00616676, throughput 13.3406K wps
[Epoch 32 Batch 60/173] avg loss 0.00636595, throughput 12.7842K wps
[Epoch 32 Batch 90/173] avg loss 0.00676699, throughput 12.929K wps
[Epoch 32 Batch 120/173] avg loss 0.00616483, throughput 12.9227K wps
[Epoch 32 Batch 150/173] avg loss 0.00633614, throughput 12.9167K wps
Begin Testing...
[Epoch 32] train avg loss 0.00639127, test acc 0.7937, test avg loss 0.431847, throughput 12.977K wps
[Epoch 33 Batch 30/173] avg loss 0.00607893, throughput 13.2112K wps
[Epoch 33 Batch 60/173] avg loss 0.00643469, throughput 12.7805K wps
[Epoch 33 Batch 90/173] avg loss 0.00647781, throughput 12.896K wps
[Epoch 33 Batch 120/173] avg loss 0.0060986, throughput 12.8955K wps
[Epoch 33 Batch 150/173] avg loss 0.00640064, throughput 12.821K wps
Begin Testing...
[Epoch 33] train avg loss 0.00630737, test acc 0.7906, test avg loss 0.43087, throughput 12.907K wps
[Epoch 34 Batch 30/173] avg loss 0.00599407, throughput 13.2109K wps
[Epoch 34 Batch 60/173] avg loss 0.00621972, throughput 12.8084K wps
[Epoch 34 Batch 90/173] avg loss 0.00600268, throughput 12.7779K wps
[Epoch 34 Batch 120/173] avg loss 0.00585322, throughput 12.7805K wps
[Epoch 34 Batch 150/173] avg loss 0.00629095, throughput 12.9412K wps
Begin Testing...
[Epoch 34] train avg loss 0.00605431, test acc 0.7896, test avg loss 0.429921, throughput 12.9106K wps
[Epoch 35 Batch 30/173] avg loss 0.00612826, throughput 13.2452K wps
[Epoch 35 Batch 60/173] avg loss 0.00603488, throughput 12.7559K wps
[Epoch 35 Batch 90/173] avg loss 0.00587309, throughput 12.9769K wps
[Epoch 35 Batch 120/173] avg loss 0.00620339, throughput 12.9846K wps
[Epoch 35 Batch 150/173] avg loss 0.00584681, throughput 12.8696K wps
Begin Testing...
[Epoch 35] train avg loss 0.00599455, test acc 0.8010, test avg loss 0.428255, throughput 12.9386K wps
Observed Improvement.
Begin Testing...
[Epoch 36 Batch 30/173] avg loss 0.00560618, throughput 13.2284K wps
[Epoch 36 Batch 60/173] avg loss 0.00587602, throughput 12.7892K wps
[Epoch 36 Batch 90/173] avg loss 0.00595086, throughput 12.8981K wps
[Epoch 36 Batch 120/173] avg loss 0.00595466, throughput 12.8021K wps
[Epoch 36 Batch 150/173] avg loss 0.00595475, throughput 12.9189K wps
Begin Testing...
[Epoch 36] train avg loss 0.00592177, test acc 0.7906, test avg loss 0.428497, throughput 12.919K wps
[Epoch 37 Batch 30/173] avg loss 0.00583212, throughput 13.296K wps
[Epoch 37 Batch 60/173] avg loss 0.00560307, throughput 12.8247K wps
[Epoch 37 Batch 90/173] avg loss 0.00585908, throughput 12.8133K wps
[Epoch 37 Batch 120/173] avg loss 0.00567025, throughput 12.9167K wps
[Epoch 37 Batch 150/173] avg loss 0.00590815, throughput 12.8847K wps
Begin Testing...
[Epoch 37] train avg loss 0.00580168, test acc 0.8073, test avg loss 0.426845, throughput 12.9232K wps
Observed Improvement.
Begin Testing...
[Epoch 38 Batch 30/173] avg loss 0.0055656, throughput 13.2436K wps
[Epoch 38 Batch 60/173] avg loss 0.00571723, throughput 12.8006K wps
[Epoch 38 Batch 90/173] avg loss 0.00570182, throughput 12.9197K wps
[Epoch 38 Batch 120/173] avg loss 0.00552139, throughput 12.799K wps
[Epoch 38 Batch 150/173] avg loss 0.00589571, throughput 12.7868K wps
Begin Testing...
[Epoch 38] train avg loss 0.00566934, test acc 0.7969, test avg loss 0.42361, throughput 12.9055K wps
[Epoch 39 Batch 30/173] avg loss 0.00543709, throughput 13.1864K wps
[Epoch 39 Batch 60/173] avg loss 0.00563853, throughput 12.7601K wps
[Epoch 39 Batch 90/173] avg loss 0.00571228, throughput 12.7873K wps
[Epoch 39 Batch 120/173] avg loss 0.00563464, throughput 12.782K wps
[Epoch 39 Batch 150/173] avg loss 0.00551811, throughput 12.9018K wps
Begin Testing...
[Epoch 39] train avg loss 0.00555806, test acc 0.7969, test avg loss 0.425688, throughput 12.8806K wps
[Epoch 40 Batch 30/173] avg loss 0.00566505, throughput 13.2591K wps
[Epoch 40 Batch 60/173] avg loss 0.0054401, throughput 12.8138K wps
[Epoch 40 Batch 90/173] avg loss 0.00543482, throughput 12.8846K wps
[Epoch 40 Batch 120/173] avg loss 0.00526231, throughput 12.8221K wps
[Epoch 40 Batch 150/173] avg loss 0.00560903, throughput 12.8713K wps
Begin Testing...
[Epoch 40] train avg loss 0.0054679, test acc 0.7948, test avg loss 0.423595, throughput 12.9091K wps
[Epoch 41 Batch 30/173] avg loss 0.00526038, throughput 13.3138K wps
[Epoch 41 Batch 60/173] avg loss 0.00520324, throughput 12.7178K wps
[Epoch 41 Batch 90/173] avg loss 0.00540373, throughput 12.9313K wps
[Epoch 41 Batch 120/173] avg loss 0.00530231, throughput 12.9521K wps
[Epoch 41 Batch 150/173] avg loss 0.00542209, throughput 12.9224K wps
Begin Testing...
[Epoch 41] train avg loss 0.00535325, test acc 0.8000, test avg loss 0.42448, throughput 12.9667K wps
[Epoch 42 Batch 30/173] avg loss 0.00524518, throughput 13.0913K wps
[Epoch 42 Batch 60/173] avg loss 0.00503403, throughput 12.7901K wps
[Epoch 42 Batch 90/173] avg loss 0.00516054, throughput 12.8871K wps
[Epoch 42 Batch 120/173] avg loss 0.00522181, throughput 12.8393K wps
[Epoch 42 Batch 150/173] avg loss 0.00541023, throughput 12.9084K wps
Begin Testing...
[Epoch 42] train avg loss 0.00521839, test acc 0.8063, test avg loss 0.42293, throughput 12.9104K wps
[Epoch 43 Batch 30/173] avg loss 0.00508196, throughput 13.1567K wps
[Epoch 43 Batch 60/173] avg loss 0.0050084, throughput 12.7715K wps
[Epoch 43 Batch 90/173] avg loss 0.00536724, throughput 12.9644K wps
[Epoch 43 Batch 120/173] avg loss 0.00512728, throughput 12.9099K wps
[Epoch 43 Batch 150/173] avg loss 0.0050046, throughput 12.7544K wps
Begin Testing...
[Epoch 43] train avg loss 0.00510997, test acc 0.8021, test avg loss 0.423138, throughput 12.9105K wps
[Epoch 44 Batch 30/173] avg loss 0.00513589, throughput 13.1745K wps
[Epoch 44 Batch 60/173] avg loss 0.00477052, throughput 12.771K wps
[Epoch 44 Batch 90/173] avg loss 0.00519944, throughput 12.7761K wps
[Epoch 44 Batch 120/173] avg loss 0.00492969, throughput 12.8341K wps
[Epoch 44 Batch 150/173] avg loss 0.00511445, throughput 12.9033K wps
Begin Testing...
[Epoch 44] train avg loss 0.00501613, test acc 0.8010, test avg loss 0.425724, throughput 12.8764K wps
[Epoch 45 Batch 30/173] avg loss 0.00477915, throughput 13.2331K wps
[Epoch 45 Batch 60/173] avg loss 0.00451849, throughput 12.7646K wps
[Epoch 45 Batch 90/173] avg loss 0.00486515, throughput 12.9126K wps
[Epoch 45 Batch 120/173] avg loss 0.00514032, throughput 12.9488K wps
[Epoch 45 Batch 150/173] avg loss 0.00489976, throughput 12.9475K wps
Begin Testing...
[Epoch 45] train avg loss 0.00483742, test acc 0.8094, test avg loss 0.42957, throughput 12.9633K wps
Observed Improvement.
Begin Testing...
[Epoch 46 Batch 30/173] avg loss 0.00482895, throughput 13.2432K wps
[Epoch 46 Batch 60/173] avg loss 0.00481221, throughput 12.723K wps
[Epoch 46 Batch 90/173] avg loss 0.0045122, throughput 12.9635K wps
[Epoch 46 Batch 120/173] avg loss 0.00470883, throughput 12.9237K wps
[Epoch 46 Batch 150/173] avg loss 0.00478771, throughput 12.9376K wps
Begin Testing...
[Epoch 46] train avg loss 0.00478475, test acc 0.8031, test avg loss 0.425839, throughput 12.9506K wps
[Epoch 47 Batch 30/173] avg loss 0.00463576, throughput 13.2901K wps
[Epoch 47 Batch 60/173] avg loss 0.00419427, throughput 12.8232K wps
[Epoch 47 Batch 90/173] avg loss 0.0045615, throughput 12.9731K wps
[Epoch 47 Batch 120/173] avg loss 0.00508581, throughput 12.9493K wps
[Epoch 47 Batch 150/173] avg loss 0.00467079, throughput 12.9472K wps
Begin Testing...
[Epoch 47] train avg loss 0.004697, test acc 0.8052, test avg loss 0.423944, throughput 13.0006K wps
[Epoch 48 Batch 30/173] avg loss 0.00459379, throughput 13.1631K wps
[Epoch 48 Batch 60/173] avg loss 0.00465743, throughput 12.7997K wps
[Epoch 48 Batch 90/173] avg loss 0.004548, throughput 12.8985K wps
[Epoch 48 Batch 120/173] avg loss 0.00468072, throughput 12.9862K wps
[Epoch 48 Batch 150/173] avg loss 0.00464593, throughput 12.9847K wps
Begin Testing...
[Epoch 48] train avg loss 0.00465177, test acc 0.7896, test avg loss 0.426972, throughput 12.9671K wps
[Epoch 49 Batch 30/173] avg loss 0.0044965, throughput 13.3299K wps
[Epoch 49 Batch 60/173] avg loss 0.00463945, throughput 12.8123K wps
[Epoch 49 Batch 90/173] avg loss 0.00469526, throughput 12.9281K wps
[Epoch 49 Batch 120/173] avg loss 0.00451143, throughput 12.9742K wps
[Epoch 49 Batch 150/173] avg loss 0.00437381, throughput 12.9694K wps
Begin Testing...
[Epoch 49] train avg loss 0.00451515, test acc 0.7875, test avg loss 0.433508, throughput 12.9959K wps
[Epoch 50 Batch 30/173] avg loss 0.0046805, throughput 13.249K wps
[Epoch 50 Batch 60/173] avg loss 0.00445539, throughput 12.8521K wps
[Epoch 50 Batch 90/173] avg loss 0.00403498, throughput 12.9064K wps
[Epoch 50 Batch 120/173] avg loss 0.00472827, throughput 12.8175K wps
[Epoch 50 Batch 150/173] avg loss 0.00420302, throughput 12.8584K wps
Begin Testing...
[Epoch 50] train avg loss 0.00442469, test acc 0.8073, test avg loss 0.42522, throughput 12.9207K wps
[Epoch 51 Batch 30/173] avg loss 0.00427726, throughput 13.2708K wps
[Epoch 51 Batch 60/173] avg loss 0.00433925, throughput 12.8692K wps
[Epoch 51 Batch 90/173] avg loss 0.00449583, throughput 12.9777K wps
[Epoch 51 Batch 120/173] avg loss 0.00446927, throughput 12.9687K wps
[Epoch 51 Batch 150/173] avg loss 0.00426865, throughput 12.8892K wps
Begin Testing...
[Epoch 51] train avg loss 0.00437209, test acc 0.8083, test avg loss 0.423159, throughput 12.9954K wps
[Epoch 52 Batch 30/173] avg loss 0.00420484, throughput 13.2607K wps
[Epoch 52 Batch 60/173] avg loss 0.0044664, throughput 12.7786K wps
[Epoch 52 Batch 90/173] avg loss 0.00435243, throughput 12.8132K wps
[Epoch 52 Batch 120/173] avg loss 0.00416668, throughput 12.9425K wps
[Epoch 52 Batch 150/173] avg loss 0.00417794, throughput 12.9726K wps
Begin Testing...
[Epoch 52] train avg loss 0.00425102, test acc 0.8083, test avg loss 0.428273, throughput 12.9398K wps
[Epoch 53 Batch 30/173] avg loss 0.00414837, throughput 13.2135K wps
[Epoch 53 Batch 60/173] avg loss 0.00426217, throughput 12.7941K wps
[Epoch 53 Batch 90/173] avg loss 0.00400035, throughput 12.9855K wps
[Epoch 53 Batch 120/173] avg loss 0.00403658, throughput 13.0004K wps
[Epoch 53 Batch 150/173] avg loss 0.00462299, throughput 12.7379K wps
Begin Testing...
[Epoch 53] train avg loss 0.00418601, test acc 0.8052, test avg loss 0.427399, throughput 12.9524K wps
[Epoch 54 Batch 30/173] avg loss 0.00404801, throughput 13.2618K wps
[Epoch 54 Batch 60/173] avg loss 0.00404502, throughput 12.7681K wps
[Epoch 54 Batch 90/173] avg loss 0.00418687, throughput 12.856K wps
[Epoch 54 Batch 120/173] avg loss 0.00390462, throughput 12.8787K wps
[Epoch 54 Batch 150/173] avg loss 0.00406297, throughput 12.863K wps
Begin Testing...
[Epoch 54] train avg loss 0.00404572, test acc 0.8052, test avg loss 0.430812, throughput 12.9325K wps
[Epoch 55 Batch 30/173] avg loss 0.00407843, throughput 13.1917K wps
[Epoch 55 Batch 60/173] avg loss 0.00379487, throughput 12.8899K wps
[Epoch 55 Batch 90/173] avg loss 0.00420294, throughput 12.9152K wps
[Epoch 55 Batch 120/173] avg loss 0.00380609, throughput 12.9005K wps
[Epoch 55 Batch 150/173] avg loss 0.00388048, throughput 12.9492K wps
Begin Testing...
[Epoch 55] train avg loss 0.00399168, test acc 0.8052, test avg loss 0.426798, throughput 12.974K wps
[Epoch 56 Batch 30/173] avg loss 0.00358698, throughput 13.2117K wps
[Epoch 56 Batch 60/173] avg loss 0.00384616, throughput 12.7813K wps
[Epoch 56 Batch 90/173] avg loss 0.00405149, throughput 12.7799K wps
[Epoch 56 Batch 120/173] avg loss 0.00380257, throughput 12.8663K wps
[Epoch 56 Batch 150/173] avg loss 0.00386149, throughput 12.8986K wps
Begin Testing...
[Epoch 56] train avg loss 0.00382517, test acc 0.8063, test avg loss 0.428528, throughput 12.9112K wps
[Epoch 57 Batch 30/173] avg loss 0.00366026, throughput 13.1289K wps
[Epoch 57 Batch 60/173] avg loss 0.00376307, throughput 12.7707K wps
[Epoch 57 Batch 90/173] avg loss 0.00355402, throughput 12.7778K wps
[Epoch 57 Batch 120/173] avg loss 0.00408563, throughput 12.9235K wps
[Epoch 57 Batch 150/173] avg loss 0.00384255, throughput 12.7794K wps
Begin Testing...
[Epoch 57] train avg loss 0.0037828, test acc 0.7990, test avg loss 0.431537, throughput 12.8861K wps
[Epoch 58 Batch 30/173] avg loss 0.00352421, throughput 13.1335K wps
[Epoch 58 Batch 60/173] avg loss 0.00377035, throughput 12.7907K wps
[Epoch 58 Batch 90/173] avg loss 0.00381548, throughput 12.9048K wps
[Epoch 58 Batch 120/173] avg loss 0.00364067, throughput 12.8326K wps
[Epoch 58 Batch 150/173] avg loss 0.00391465, throughput 12.8468K wps
Begin Testing...
[Epoch 58] train avg loss 0.00374492, test acc 0.8000, test avg loss 0.433506, throughput 12.8882K wps
[Epoch 59 Batch 30/173] avg loss 0.00339691, throughput 13.2935K wps
[Epoch 59 Batch 60/173] avg loss 0.00346132, throughput 12.7647K wps
[Epoch 59 Batch 90/173] avg loss 0.00387849, throughput 12.9165K wps
[Epoch 59 Batch 120/173] avg loss 0.00389996, throughput 12.8522K wps
[Epoch 59 Batch 150/173] avg loss 0.00359376, throughput 12.7713K wps
Begin Testing...
[Epoch 59] train avg loss 0.00367893, test acc 0.8031, test avg loss 0.433526, throughput 12.9021K wps
Test loss 0.430243, test acc 0.8086
Total time cost 176.34s
9596 1066
[Epoch 0 Batch 30/173] avg loss 0.0151524, throughput 11.6175K wps
[Epoch 0 Batch 60/173] avg loss 0.0150427, throughput 12.792K wps
[Epoch 0 Batch 90/173] avg loss 0.0147196, throughput 12.7494K wps
[Epoch 0 Batch 120/173] avg loss 0.0143724, throughput 12.7777K wps
[Epoch 0 Batch 150/173] avg loss 0.0145958, throughput 12.8502K wps
Begin Testing...
[Epoch 0] train avg loss 0.0147268, test acc 0.5802, test avg loss 0.672383, throughput 12.57K wps
Observed Improvement.
Begin Testing...
[Epoch 1 Batch 30/173] avg loss 0.0137053, throughput 13.2086K wps
[Epoch 1 Batch 60/173] avg loss 0.0136243, throughput 12.7427K wps
[Epoch 1 Batch 90/173] avg loss 0.0136274, throughput 12.8448K wps
[Epoch 1 Batch 120/173] avg loss 0.0136326, throughput 12.8126K wps
[Epoch 1 Batch 150/173] avg loss 0.0136055, throughput 12.7767K wps
Begin Testing...
[Epoch 1] train avg loss 0.013647, test acc 0.6125, test avg loss 0.657152, throughput 12.8695K wps
Observed Improvement.
Begin Testing...
[Epoch 2 Batch 30/173] avg loss 0.0132757, throughput 13.2177K wps
[Epoch 2 Batch 60/173] avg loss 0.0131856, throughput 12.7925K wps
[Epoch 2 Batch 90/173] avg loss 0.0132863, throughput 12.8418K wps
[Epoch 2 Batch 120/173] avg loss 0.0130683, throughput 12.7755K wps
[Epoch 2 Batch 150/173] avg loss 0.0130289, throughput 12.8104K wps
Begin Testing...
[Epoch 2] train avg loss 0.0131675, test acc 0.6250, test avg loss 0.649466, throughput 12.8783K wps
Observed Improvement.
Begin Testing...
[Epoch 3 Batch 30/173] avg loss 0.0127144, throughput 13.2227K wps
[Epoch 3 Batch 60/173] avg loss 0.0129495, throughput 12.7068K wps
[Epoch 3 Batch 90/173] avg loss 0.0127554, throughput 12.9534K wps
[Epoch 3 Batch 120/173] avg loss 0.0127797, throughput 12.8121K wps
[Epoch 3 Batch 150/173] avg loss 0.0126972, throughput 12.8302K wps
Begin Testing...
[Epoch 3] train avg loss 0.0127617, test acc 0.6542, test avg loss 0.635802, throughput 12.8932K wps
Observed Improvement.
Begin Testing...
[Epoch 4 Batch 30/173] avg loss 0.012181, throughput 13.2331K wps
[Epoch 4 Batch 60/173] avg loss 0.0123511, throughput 12.7686K wps
[Epoch 4 Batch 90/173] avg loss 0.0125008, throughput 12.8885K wps
[Epoch 4 Batch 120/173] avg loss 0.0122616, throughput 12.91K wps
[Epoch 4 Batch 150/173] avg loss 0.0124184, throughput 12.8223K wps
Begin Testing...
[Epoch 4] train avg loss 0.0123452, test acc 0.6750, test avg loss 0.622023, throughput 12.9278K wps
Observed Improvement.
Begin Testing...
[Epoch 5 Batch 30/173] avg loss 0.0119048, throughput 13.2309K wps
[Epoch 5 Batch 60/173] avg loss 0.0119659, throughput 12.7978K wps
[Epoch 5 Batch 90/173] avg loss 0.0120538, throughput 12.752K wps
[Epoch 5 Batch 120/173] avg loss 0.0121289, throughput 12.804K wps
[Epoch 5 Batch 150/173] avg loss 0.0119629, throughput 12.7977K wps
Begin Testing...
[Epoch 5] train avg loss 0.0120006, test acc 0.6927, test avg loss 0.610899, throughput 12.8704K wps
Observed Improvement.
Begin Testing...
[Epoch 6 Batch 30/173] avg loss 0.011775, throughput 13.2124K wps
[Epoch 6 Batch 60/173] avg loss 0.0116298, throughput 12.7732K wps
[Epoch 6 Batch 90/173] avg loss 0.011752, throughput 12.8094K wps
[Epoch 6 Batch 120/173] avg loss 0.0117439, throughput 12.8543K wps
[Epoch 6 Batch 150/173] avg loss 0.0115773, throughput 12.8582K wps
Begin Testing...
[Epoch 6] train avg loss 0.0116888, test acc 0.6823, test avg loss 0.600591, throughput 12.8922K wps
[Epoch 7 Batch 30/173] avg loss 0.01155, throughput 13.1906K wps
[Epoch 7 Batch 60/173] avg loss 0.011502, throughput 12.766K wps
[Epoch 7 Batch 90/173] avg loss 0.0113015, throughput 12.9384K wps
[Epoch 7 Batch 120/173] avg loss 0.0110419, throughput 12.9223K wps
[Epoch 7 Batch 150/173] avg loss 0.011514, throughput 12.8003K wps
Begin Testing...
[Epoch 7] train avg loss 0.0113836, test acc 0.7156, test avg loss 0.585941, throughput 12.9286K wps
Observed Improvement.
Begin Testing...
[Epoch 8 Batch 30/173] avg loss 0.0109868, throughput 13.2909K wps
[Epoch 8 Batch 60/173] avg loss 0.0111387, throughput 12.8136K wps
[Epoch 8 Batch 90/173] avg loss 0.010957, throughput 12.9714K wps
[Epoch 8 Batch 120/173] avg loss 0.010787, throughput 12.9653K wps
[Epoch 8 Batch 150/173] avg loss 0.0109498, throughput 12.7726K wps
Begin Testing...
[Epoch 8] train avg loss 0.0109599, test acc 0.7250, test avg loss 0.571707, throughput 12.9611K wps
Observed Improvement.
Begin Testing...
[Epoch 9 Batch 30/173] avg loss 0.0106114, throughput 13.3125K wps
[Epoch 9 Batch 60/173] avg loss 0.0104068, throughput 12.7864K wps
[Epoch 9 Batch 90/173] avg loss 0.0105999, throughput 12.9226K wps
[Epoch 9 Batch 120/173] avg loss 0.0106639, throughput 12.9601K wps
[Epoch 9 Batch 150/173] avg loss 0.0107762, throughput 12.8485K wps
Begin Testing...
[Epoch 9] train avg loss 0.0106359, test acc 0.7458, test avg loss 0.550974, throughput 12.943K wps
Observed Improvement.
Begin Testing...
[Epoch 10 Batch 30/173] avg loss 0.0104571, throughput 13.2382K wps
[Epoch 10 Batch 60/173] avg loss 0.0102906, throughput 12.8277K wps
[Epoch 10 Batch 90/173] avg loss 0.0105694, throughput 12.9755K wps
[Epoch 10 Batch 120/173] avg loss 0.0101364, throughput 12.821K wps
[Epoch 10 Batch 150/173] avg loss 0.0100027, throughput 12.7679K wps
Begin Testing...
[Epoch 10] train avg loss 0.0102823, test acc 0.7469, test avg loss 0.541372, throughput 12.9067K wps
Observed Improvement.
Begin Testing...
[Epoch 11 Batch 30/173] avg loss 0.0099515, throughput 13.2218K wps
[Epoch 11 Batch 60/173] avg loss 0.0100327, throughput 12.8671K wps
[Epoch 11 Batch 90/173] avg loss 0.00980991, throughput 12.9421K wps
[Epoch 11 Batch 120/173] avg loss 0.010027, throughput 12.7701K wps
[Epoch 11 Batch 150/173] avg loss 0.0098386, throughput 12.8064K wps
Begin Testing...
[Epoch 11] train avg loss 0.00996404, test acc 0.7531, test avg loss 0.527742, throughput 12.9087K wps
Observed Improvement.
Begin Testing...
[Epoch 12 Batch 30/173] avg loss 0.00966528, throughput 13.2524K wps
[Epoch 12 Batch 60/173] avg loss 0.00953796, throughput 12.6733K wps
[Epoch 12 Batch 90/173] avg loss 0.00985969, throughput 12.9766K wps
[Epoch 12 Batch 120/173] avg loss 0.00950065, throughput 12.9555K wps
[Epoch 12 Batch 150/173] avg loss 0.00972353, throughput 12.7808K wps
Begin Testing...
[Epoch 12] train avg loss 0.00966567, test acc 0.7615, test avg loss 0.51724, throughput 12.906K wps
Observed Improvement.
Begin Testing...
[Epoch 13 Batch 30/173] avg loss 0.00919454, throughput 13.1998K wps
[Epoch 13 Batch 60/173] avg loss 0.00944392, throughput 12.7899K wps
[Epoch 13 Batch 90/173] avg loss 0.00953906, throughput 12.9306K wps
[Epoch 13 Batch 120/173] avg loss 0.00941792, throughput 12.761K wps
[Epoch 13 Batch 150/173] avg loss 0.00937444, throughput 12.8058K wps
Begin Testing...
[Epoch 13] train avg loss 0.00945623, test acc 0.7646, test avg loss 0.509236, throughput 12.8973K wps
Observed Improvement.
Begin Testing...
[Epoch 14 Batch 30/173] avg loss 0.00899179, throughput 13.2067K wps
[Epoch 14 Batch 60/173] avg loss 0.00904377, throughput 12.8684K wps
[Epoch 14 Batch 90/173] avg loss 0.00918283, throughput 12.8768K wps
[Epoch 14 Batch 120/173] avg loss 0.00903011, throughput 12.7202K wps
[Epoch 14 Batch 150/173] avg loss 0.0090286, throughput 12.8996K wps
Begin Testing...
[Epoch 14] train avg loss 0.00909493, test acc 0.7667, test avg loss 0.500426, throughput 12.9179K wps
Observed Improvement.
Begin Testing...
[Epoch 15 Batch 30/173] avg loss 0.00865062, throughput 13.2812K wps
[Epoch 15 Batch 60/173] avg loss 0.0091304, throughput 12.7809K wps
[Epoch 15 Batch 90/173] avg loss 0.00844185, throughput 12.7705K wps
[Epoch 15 Batch 120/173] avg loss 0.00884389, throughput 12.8014K wps
[Epoch 15 Batch 150/173] avg loss 0.00861288, throughput 12.9645K wps
Begin Testing...
[Epoch 15] train avg loss 0.00879633, test acc 0.7656, test avg loss 0.493261, throughput 12.9226K wps
[Epoch 16 Batch 30/173] avg loss 0.00869348, throughput 13.2987K wps
[Epoch 16 Batch 60/173] avg loss 0.00867016, throughput 12.7858K wps
[Epoch 16 Batch 90/173] avg loss 0.00874878, throughput 12.8082K wps
[Epoch 16 Batch 120/173] avg loss 0.00882192, throughput 12.8015K wps
[Epoch 16 Batch 150/173] avg loss 0.00834557, throughput 12.8077K wps
Begin Testing...
[Epoch 16] train avg loss 0.00866746, test acc 0.7719, test avg loss 0.486526, throughput 12.9088K wps
Observed Improvement.
Begin Testing...
[Epoch 17 Batch 30/173] avg loss 0.00871555, throughput 13.2244K wps
[Epoch 17 Batch 60/173] avg loss 0.00862859, throughput 12.7931K wps
[Epoch 17 Batch 90/173] avg loss 0.00840512, throughput 12.7482K wps
[Epoch 17 Batch 120/173] avg loss 0.00846066, throughput 12.8935K wps
[Epoch 17 Batch 150/173] avg loss 0.00835496, throughput 12.7783K wps
Begin Testing...
[Epoch 17] train avg loss 0.0085397, test acc 0.7708, test avg loss 0.485966, throughput 12.8777K wps
[Epoch 18 Batch 30/173] avg loss 0.00818135, throughput 13.1772K wps
[Epoch 18 Batch 60/173] avg loss 0.00821621, throughput 12.7915K wps
[Epoch 18 Batch 90/173] avg loss 0.00868828, throughput 12.8058K wps
[Epoch 18 Batch 120/173] avg loss 0.00830405, throughput 12.7816K wps
[Epoch 18 Batch 150/173] avg loss 0.00811452, throughput 12.8412K wps
Begin Testing...
[Epoch 18] train avg loss 0.00833557, test acc 0.7688, test avg loss 0.478227, throughput 12.8582K wps
[Epoch 19 Batch 30/173] avg loss 0.00802665, throughput 13.1904K wps
[Epoch 19 Batch 60/173] avg loss 0.00809132, throughput 12.7961K wps
[Epoch 19 Batch 90/173] avg loss 0.00807883, throughput 12.7882K wps
[Epoch 19 Batch 120/173] avg loss 0.00800446, throughput 12.8228K wps
[Epoch 19 Batch 150/173] avg loss 0.00794487, throughput 12.8759K wps
Begin Testing...
[Epoch 19] train avg loss 0.00807185, test acc 0.7646, test avg loss 0.479814, throughput 12.8811K wps
[Epoch 20 Batch 30/173] avg loss 0.00819262, throughput 13.2007K wps
[Epoch 20 Batch 60/173] avg loss 0.00788458, throughput 12.7828K wps
[Epoch 20 Batch 90/173] avg loss 0.00778894, throughput 12.8786K wps
[Epoch 20 Batch 120/173] avg loss 0.00804572, throughput 12.8307K wps
[Epoch 20 Batch 150/173] avg loss 0.00791009, throughput 12.7622K wps
Begin Testing...
[Epoch 20] train avg loss 0.00794528, test acc 0.7719, test avg loss 0.472251, throughput 12.8958K wps
Observed Improvement.
Begin Testing...
[Epoch 21 Batch 30/173] avg loss 0.00799556, throughput 13.2512K wps
[Epoch 21 Batch 60/173] avg loss 0.00803876, throughput 12.6999K wps
[Epoch 21 Batch 90/173] avg loss 0.00774099, throughput 12.9354K wps
[Epoch 21 Batch 120/173] avg loss 0.00759676, throughput 12.8907K wps
[Epoch 21 Batch 150/173] avg loss 0.00770579, throughput 12.7829K wps
Begin Testing...
[Epoch 21] train avg loss 0.00776494, test acc 0.7781, test avg loss 0.470593, throughput 12.8979K wps
Observed Improvement.
Begin Testing...
[Epoch 22 Batch 30/173] avg loss 0.00749638, throughput 13.2121K wps
[Epoch 22 Batch 60/173] avg loss 0.00781318, throughput 12.7802K wps
[Epoch 22 Batch 90/173] avg loss 0.00764918, throughput 12.7703K wps
[Epoch 22 Batch 120/173] avg loss 0.00754806, throughput 12.7996K wps
[Epoch 22 Batch 150/173] avg loss 0.00772536, throughput 12.9079K wps
Begin Testing...
[Epoch 22] train avg loss 0.00760349, test acc 0.7833, test avg loss 0.472667, throughput 12.887K wps
Observed Improvement.
Begin Testing...
[Epoch 23 Batch 30/173] avg loss 0.00700628, throughput 13.1185K wps
[Epoch 23 Batch 60/173] avg loss 0.00785841, throughput 12.7588K wps
[Epoch 23 Batch 90/173] avg loss 0.00737391, throughput 12.9387K wps
[Epoch 23 Batch 120/173] avg loss 0.00741421, throughput 12.8161K wps
[Epoch 23 Batch 150/173] avg loss 0.00737714, throughput 12.9148K wps
Begin Testing...
[Epoch 23] train avg loss 0.00742686, test acc 0.7729, test avg loss 0.46763, throughput 12.8887K wps
[Epoch 24 Batch 30/173] avg loss 0.00764571, throughput 13.2376K wps
[Epoch 24 Batch 60/173] avg loss 0.00709496, throughput 12.7669K wps
[Epoch 24 Batch 90/173] avg loss 0.00749439, throughput 12.9157K wps
[Epoch 24 Batch 120/173] avg loss 0.00720169, throughput 12.9278K wps
[Epoch 24 Batch 150/173] avg loss 0.00719533, throughput 12.9131K wps
Begin Testing...
[Epoch 24] train avg loss 0.00733866, test acc 0.7708, test avg loss 0.465478, throughput 12.9364K wps
[Epoch 25 Batch 30/173] avg loss 0.00696051, throughput 13.1152K wps
[Epoch 25 Batch 60/173] avg loss 0.00709862, throughput 12.7344K wps
[Epoch 25 Batch 90/173] avg loss 0.00715208, throughput 12.9282K wps
[Epoch 25 Batch 120/173] avg loss 0.00740778, throughput 12.8994K wps
[Epoch 25 Batch 150/173] avg loss 0.00713214, throughput 12.7715K wps
Begin Testing...
[Epoch 25] train avg loss 0.00716664, test acc 0.7812, test avg loss 0.460902, throughput 12.8943K wps
[Epoch 26 Batch 30/173] avg loss 0.00689925, throughput 13.3385K wps
[Epoch 26 Batch 60/173] avg loss 0.00704611, throughput 12.7563K wps
[Epoch 26 Batch 90/173] avg loss 0.0070697, throughput 12.8042K wps
[Epoch 26 Batch 120/173] avg loss 0.00720208, throughput 12.864K wps
[Epoch 26 Batch 150/173] avg loss 0.00680805, throughput 12.945K wps
Begin Testing...
[Epoch 26] train avg loss 0.00704472, test acc 0.7781, test avg loss 0.462703, throughput 12.9282K wps
[Epoch 27 Batch 30/173] avg loss 0.00683671, throughput 13.244K wps
[Epoch 27 Batch 60/173] avg loss 0.00727766, throughput 12.7964K wps
[Epoch 27 Batch 90/173] avg loss 0.00699965, throughput 12.9301K wps
[Epoch 27 Batch 120/173] avg loss 0.00707946, throughput 12.8758K wps
[Epoch 27 Batch 150/173] avg loss 0.00695549, throughput 12.7996K wps
Begin Testing...
[Epoch 27] train avg loss 0.00701148, test acc 0.7833, test avg loss 0.461034, throughput 12.9238K wps
Observed Improvement.
Begin Testing...
[Epoch 28 Batch 30/173] avg loss 0.0067351, throughput 13.1676K wps
[Epoch 28 Batch 60/173] avg loss 0.00673381, throughput 12.8392K wps
[Epoch 28 Batch 90/173] avg loss 0.00696579, throughput 12.9185K wps
[Epoch 28 Batch 120/173] avg loss 0.00680854, throughput 12.7957K wps
[Epoch 28 Batch 150/173] avg loss 0.00647138, throughput 12.8486K wps
Begin Testing...
[Epoch 28] train avg loss 0.00673665, test acc 0.7802, test avg loss 0.458574, throughput 12.9012K wps
[Epoch 29 Batch 30/173] avg loss 0.00664245, throughput 13.2409K wps
[Epoch 29 Batch 60/173] avg loss 0.00656205, throughput 12.7978K wps
[Epoch 29 Batch 90/173] avg loss 0.00672899, throughput 12.9289K wps
[Epoch 29 Batch 120/173] avg loss 0.00683076, throughput 12.8851K wps
[Epoch 29 Batch 150/173] avg loss 0.00650514, throughput 12.8247K wps
Begin Testing...
[Epoch 29] train avg loss 0.00665653, test acc 0.7781, test avg loss 0.460484, throughput 12.9371K wps
[Epoch 30 Batch 30/173] avg loss 0.00622721, throughput 13.2769K wps
[Epoch 30 Batch 60/173] avg loss 0.00681436, throughput 12.7887K wps
[Epoch 30 Batch 90/173] avg loss 0.00659083, throughput 12.9342K wps
[Epoch 30 Batch 120/173] avg loss 0.00651901, throughput 12.9695K wps
[Epoch 30 Batch 150/173] avg loss 0.00643092, throughput 12.8014K wps
Begin Testing...
[Epoch 30] train avg loss 0.00651403, test acc 0.7771, test avg loss 0.460924, throughput 12.9403K wps
[Epoch 31 Batch 30/173] avg loss 0.00670716, throughput 13.241K wps
[Epoch 31 Batch 60/173] avg loss 0.00630979, throughput 12.7951K wps
[Epoch 31 Batch 90/173] avg loss 0.00632582, throughput 12.9806K wps
[Epoch 31 Batch 120/173] avg loss 0.00635289, throughput 12.865K wps
[Epoch 31 Batch 150/173] avg loss 0.00636553, throughput 12.7929K wps
Begin Testing...
[Epoch 31] train avg loss 0.00639246, test acc 0.7823, test avg loss 0.454522, throughput 12.9158K wps
[Epoch 32 Batch 30/173] avg loss 0.00595197, throughput 13.1268K wps
[Epoch 32 Batch 60/173] avg loss 0.00626357, throughput 12.7425K wps
[Epoch 32 Batch 90/173] avg loss 0.00622767, throughput 12.8391K wps
[Epoch 32 Batch 120/173] avg loss 0.0066223, throughput 12.6079K wps
[Epoch 32 Batch 150/173] avg loss 0.00634274, throughput 12.9293K wps
Begin Testing...
[Epoch 32] train avg loss 0.00627788, test acc 0.7771, test avg loss 0.460345, throughput 12.8481K wps
[Epoch 33 Batch 30/173] avg loss 0.00596772, throughput 13.2481K wps
[Epoch 33 Batch 60/173] avg loss 0.00616452, throughput 12.7753K wps
[Epoch 33 Batch 90/173] avg loss 0.00624277, throughput 12.931K wps
[Epoch 33 Batch 120/173] avg loss 0.00632832, throughput 12.7877K wps
[Epoch 33 Batch 150/173] avg loss 0.0062896, throughput 12.7946K wps
Begin Testing...
[Epoch 33] train avg loss 0.00620223, test acc 0.7750, test avg loss 0.45581, throughput 12.8941K wps
[Epoch 34 Batch 30/173] avg loss 0.00591691, throughput 13.2161K wps
[Epoch 34 Batch 60/173] avg loss 0.00582343, throughput 12.8027K wps
[Epoch 34 Batch 90/173] avg loss 0.00609009, throughput 12.8043K wps
[Epoch 34 Batch 120/173] avg loss 0.00621081, throughput 12.7943K wps
[Epoch 34 Batch 150/173] avg loss 0.00594148, throughput 12.7731K wps
Begin Testing...
[Epoch 34] train avg loss 0.00602792, test acc 0.7844, test avg loss 0.455501, throughput 12.8825K wps
Observed Improvement.
Begin Testing...
[Epoch 35 Batch 30/173] avg loss 0.00588506, throughput 13.2645K wps
[Epoch 35 Batch 60/173] avg loss 0.00610468, throughput 12.7903K wps
[Epoch 35 Batch 90/173] avg loss 0.00592211, throughput 12.9178K wps
[Epoch 35 Batch 120/173] avg loss 0.00597714, throughput 12.8503K wps
[Epoch 35 Batch 150/173] avg loss 0.0056153, throughput 12.8836K wps
Begin Testing...
[Epoch 35] train avg loss 0.00595106, test acc 0.7771, test avg loss 0.458984, throughput 12.9284K wps
[Epoch 36 Batch 30/173] avg loss 0.00551817, throughput 13.2055K wps
[Epoch 36 Batch 60/173] avg loss 0.00558653, throughput 12.8221K wps
[Epoch 36 Batch 90/173] avg loss 0.00565287, throughput 12.853K wps
[Epoch 36 Batch 120/173] avg loss 0.00630687, throughput 12.8185K wps
[Epoch 36 Batch 150/173] avg loss 0.00587604, throughput 12.7702K wps
Begin Testing...
[Epoch 36] train avg loss 0.00578451, test acc 0.7698, test avg loss 0.457527, throughput 12.8899K wps
[Epoch 37 Batch 30/173] avg loss 0.00549006, throughput 13.2612K wps
[Epoch 37 Batch 60/173] avg loss 0.00549616, throughput 12.7949K wps
[Epoch 37 Batch 90/173] avg loss 0.00564213, throughput 12.7959K wps
[Epoch 37 Batch 120/173] avg loss 0.00608558, throughput 12.7565K wps
[Epoch 37 Batch 150/173] avg loss 0.00591311, throughput 12.748K wps
Begin Testing...
[Epoch 37] train avg loss 0.00574142, test acc 0.7708, test avg loss 0.453361, throughput 12.8809K wps
[Epoch 38 Batch 30/173] avg loss 0.00536974, throughput 13.1966K wps
[Epoch 38 Batch 60/173] avg loss 0.00532566, throughput 12.819K wps
[Epoch 38 Batch 90/173] avg loss 0.00608859, throughput 12.7825K wps
[Epoch 38 Batch 120/173] avg loss 0.0055087, throughput 12.89K wps
[Epoch 38 Batch 150/173] avg loss 0.00530299, throughput 12.7278K wps
Begin Testing...
[Epoch 38] train avg loss 0.00554422, test acc 0.7792, test avg loss 0.458608, throughput 12.8692K wps
[Epoch 39 Batch 30/173] avg loss 0.00543057, throughput 13.167K wps
[Epoch 39 Batch 60/173] avg loss 0.00555977, throughput 12.7516K wps
[Epoch 39 Batch 90/173] avg loss 0.00538133, throughput 12.9238K wps
[Epoch 39 Batch 120/173] avg loss 0.00534248, throughput 12.8039K wps
[Epoch 39 Batch 150/173] avg loss 0.00554916, throughput 12.9519K wps
Begin Testing...
[Epoch 39] train avg loss 0.00550404, test acc 0.7812, test avg loss 0.454195, throughput 12.9265K wps
[Epoch 40 Batch 30/173] avg loss 0.00509005, throughput 13.3488K wps
[Epoch 40 Batch 60/173] avg loss 0.00518551, throughput 12.7882K wps
[Epoch 40 Batch 90/173] avg loss 0.00526839, throughput 12.7815K wps
[Epoch 40 Batch 120/173] avg loss 0.00564546, throughput 12.7968K wps
[Epoch 40 Batch 150/173] avg loss 0.00548266, throughput 12.7894K wps
Begin Testing...
[Epoch 40] train avg loss 0.0053259, test acc 0.7760, test avg loss 0.455419, throughput 12.8887K wps
[Epoch 41 Batch 30/173] avg loss 0.00534508, throughput 13.2139K wps
[Epoch 41 Batch 60/173] avg loss 0.00506123, throughput 12.798K wps
[Epoch 41 Batch 90/173] avg loss 0.00516898, throughput 12.9214K wps
[Epoch 41 Batch 120/173] avg loss 0.00528502, throughput 12.8585K wps
[Epoch 41 Batch 150/173] avg loss 0.0052356, throughput 12.7837K wps
Begin Testing...
[Epoch 41] train avg loss 0.00528614, test acc 0.7771, test avg loss 0.455079, throughput 12.9009K wps
[Epoch 42 Batch 30/173] avg loss 0.00511145, throughput 13.2846K wps
[Epoch 42 Batch 60/173] avg loss 0.00516392, throughput 12.7616K wps
[Epoch 42 Batch 90/173] avg loss 0.0053988, throughput 12.9349K wps
[Epoch 42 Batch 120/173] avg loss 0.00506197, throughput 12.9109K wps
[Epoch 42 Batch 150/173] avg loss 0.00489847, throughput 12.8356K wps
Begin Testing...
[Epoch 42] train avg loss 0.005119, test acc 0.7812, test avg loss 0.453967, throughput 12.9491K wps
[Epoch 43 Batch 30/173] avg loss 0.00487853, throughput 13.261K wps
[Epoch 43 Batch 60/173] avg loss 0.00511951, throughput 12.7897K wps
[Epoch 43 Batch 90/173] avg loss 0.00511121, throughput 12.9559K wps
[Epoch 43 Batch 120/173] avg loss 0.00492639, throughput 12.9811K wps
[Epoch 43 Batch 150/173] avg loss 0.00520825, throughput 12.9645K wps
Begin Testing...
[Epoch 43] train avg loss 0.00507661, test acc 0.7760, test avg loss 0.456868, throughput 12.9694K wps
[Epoch 44 Batch 30/173] avg loss 0.00498666, throughput 13.36K wps
[Epoch 44 Batch 60/173] avg loss 0.00476046, throughput 12.8748K wps
[Epoch 44 Batch 90/173] avg loss 0.00489349, throughput 12.9309K wps
[Epoch 44 Batch 120/173] avg loss 0.0049188, throughput 12.9521K wps
[Epoch 44 Batch 150/173] avg loss 0.00530008, throughput 12.9536K wps
Begin Testing...
[Epoch 44] train avg loss 0.00494476, test acc 0.7781, test avg loss 0.458981, throughput 12.9899K wps
[Epoch 45 Batch 30/173] avg loss 0.00471308, throughput 13.1971K wps
[Epoch 45 Batch 60/173] avg loss 0.0046661, throughput 12.782K wps
[Epoch 45 Batch 90/173] avg loss 0.00526665, throughput 12.9314K wps
[Epoch 45 Batch 120/173] avg loss 0.0050314, throughput 12.9195K wps
[Epoch 45 Batch 150/173] avg loss 0.0046355, throughput 12.9296K wps
Begin Testing...
[Epoch 45] train avg loss 0.00488144, test acc 0.7698, test avg loss 0.46003, throughput 12.9525K wps
[Epoch 46 Batch 30/173] avg loss 0.00463045, throughput 13.1912K wps
[Epoch 46 Batch 60/173] avg loss 0.00457944, throughput 12.7694K wps
[Epoch 46 Batch 90/173] avg loss 0.00475371, throughput 12.9308K wps
[Epoch 46 Batch 120/173] avg loss 0.00494359, throughput 12.9715K wps
[Epoch 46 Batch 150/173] avg loss 0.00484599, throughput 12.9248K wps
Begin Testing...
[Epoch 46] train avg loss 0.00472342, test acc 0.7750, test avg loss 0.455345, throughput 12.956K wps
[Epoch 47 Batch 30/173] avg loss 0.00448674, throughput 13.1708K wps
[Epoch 47 Batch 60/173] avg loss 0.0045311, throughput 12.8273K wps
[Epoch 47 Batch 90/173] avg loss 0.00462122, throughput 12.7838K wps
[Epoch 47 Batch 120/173] avg loss 0.00434015, throughput 12.9184K wps
[Epoch 47 Batch 150/173] avg loss 0.00493087, throughput 12.9445K wps
Begin Testing...
[Epoch 47] train avg loss 0.00462597, test acc 0.7760, test avg loss 0.460254, throughput 12.9293K wps
[Epoch 48 Batch 30/173] avg loss 0.00457363, throughput 13.1857K wps
[Epoch 48 Batch 60/173] avg loss 0.00448209, throughput 12.8089K wps
[Epoch 48 Batch 90/173] avg loss 0.00452993, throughput 12.9574K wps
[Epoch 48 Batch 120/173] avg loss 0.00444495, throughput 12.9292K wps
[Epoch 48 Batch 150/173] avg loss 0.00457171, throughput 12.8264K wps
Begin Testing...
[Epoch 48] train avg loss 0.00450409, test acc 0.7771, test avg loss 0.455645, throughput 12.9248K wps
[Epoch 49 Batch 30/173] avg loss 0.00430992, throughput 13.1752K wps
[Epoch 49 Batch 60/173] avg loss 0.0045806, throughput 12.7701K wps
[Epoch 49 Batch 90/173] avg loss 0.00460319, throughput 12.7391K wps
[Epoch 49 Batch 120/173] avg loss 0.00448334, throughput 12.9458K wps
[Epoch 49 Batch 150/173] avg loss 0.00432586, throughput 12.9336K wps
Begin Testing...
[Epoch 49] train avg loss 0.0044843, test acc 0.7729, test avg loss 0.458359, throughput 12.9165K wps
[Epoch 50 Batch 30/173] avg loss 0.00399165, throughput 13.1712K wps
[Epoch 50 Batch 60/173] avg loss 0.00446413, throughput 12.8825K wps
[Epoch 50 Batch 90/173] avg loss 0.00446258, throughput 12.9401K wps
[Epoch 50 Batch 120/173] avg loss 0.00447266, throughput 12.9532K wps
[Epoch 50 Batch 150/173] avg loss 0.00445407, throughput 12.7864K wps
Begin Testing...
[Epoch 50] train avg loss 0.00440257, test acc 0.7823, test avg loss 0.456984, throughput 12.9404K wps
[Epoch 51 Batch 30/173] avg loss 0.0043485, throughput 13.0924K wps
[Epoch 51 Batch 60/173] avg loss 0.00444854, throughput 12.8801K wps
[Epoch 51 Batch 90/173] avg loss 0.00423502, throughput 12.9376K wps
[Epoch 51 Batch 120/173] avg loss 0.00425346, throughput 12.8629K wps
[Epoch 51 Batch 150/173] avg loss 0.00392564, throughput 12.7717K wps
Begin Testing...
[Epoch 51] train avg loss 0.00423883, test acc 0.7760, test avg loss 0.456457, throughput 12.8939K wps
[Epoch 52 Batch 30/173] avg loss 0.00375363, throughput 13.2099K wps
[Epoch 52 Batch 60/173] avg loss 0.00421807, throughput 12.855K wps
[Epoch 52 Batch 90/173] avg loss 0.00429947, throughput 12.8586K wps
[Epoch 52 Batch 120/173] avg loss 0.00424825, throughput 12.8417K wps
[Epoch 52 Batch 150/173] avg loss 0.00428055, throughput 12.9148K wps
Begin Testing...
[Epoch 52] train avg loss 0.00417416, test acc 0.7698, test avg loss 0.459893, throughput 12.9266K wps
[Epoch 53 Batch 30/173] avg loss 0.00401279, throughput 13.2773K wps
[Epoch 53 Batch 60/173] avg loss 0.00418852, throughput 12.8493K wps
[Epoch 53 Batch 90/173] avg loss 0.00422255, throughput 12.8751K wps
[Epoch 53 Batch 120/173] avg loss 0.00404822, throughput 12.8945K wps
[Epoch 53 Batch 150/173] avg loss 0.00415702, throughput 12.7912K wps
Begin Testing...
[Epoch 53] train avg loss 0.00411011, test acc 0.7771, test avg loss 0.46449, throughput 12.9178K wps
[Epoch 54 Batch 30/173] avg loss 0.0042004, throughput 13.19K wps
[Epoch 54 Batch 60/173] avg loss 0.00382151, throughput 12.7394K wps
[Epoch 54 Batch 90/173] avg loss 0.00392023, throughput 12.9646K wps
[Epoch 54 Batch 120/173] avg loss 0.00419067, throughput 12.8239K wps
[Epoch 54 Batch 150/173] avg loss 0.00381288, throughput 12.881K wps
Begin Testing...
[Epoch 54] train avg loss 0.00402107, test acc 0.7812, test avg loss 0.460522, throughput 12.9081K wps
[Epoch 55 Batch 30/173] avg loss 0.00376142, throughput 13.1932K wps
[Epoch 55 Batch 60/173] avg loss 0.0040127, throughput 12.7384K wps
[Epoch 55 Batch 90/173] avg loss 0.00450781, throughput 12.8139K wps
[Epoch 55 Batch 120/173] avg loss 0.00382088, throughput 12.808K wps
[Epoch 55 Batch 150/173] avg loss 0.00406865, throughput 12.9043K wps
Begin Testing...
[Epoch 55] train avg loss 0.00404982, test acc 0.7875, test avg loss 0.463137, throughput 12.8942K wps
Observed Improvement.
Begin Testing...
[Epoch 56 Batch 30/173] avg loss 0.00372365, throughput 13.2996K wps
[Epoch 56 Batch 60/173] avg loss 0.00398371, throughput 12.7834K wps
[Epoch 56 Batch 90/173] avg loss 0.00385591, throughput 12.8703K wps
[Epoch 56 Batch 120/173] avg loss 0.0038927, throughput 12.7778K wps
[Epoch 56 Batch 150/173] avg loss 0.00374345, throughput 12.7832K wps
Begin Testing...
[Epoch 56] train avg loss 0.00386282, test acc 0.7812, test avg loss 0.467242, throughput 12.9078K wps
[Epoch 57 Batch 30/173] avg loss 0.00384315, throughput 13.1889K wps
[Epoch 57 Batch 60/173] avg loss 0.0038752, throughput 12.8124K wps
[Epoch 57 Batch 90/173] avg loss 0.00366811, throughput 12.9398K wps
[Epoch 57 Batch 120/173] avg loss 0.00352958, throughput 12.7884K wps
[Epoch 57 Batch 150/173] avg loss 0.00367608, throughput 12.9554K wps
Begin Testing...
[Epoch 57] train avg loss 0.0037496, test acc 0.7812, test avg loss 0.46485, throughput 12.939K wps
[Epoch 58 Batch 30/173] avg loss 0.00369646, throughput 13.1637K wps
[Epoch 58 Batch 60/173] avg loss 0.00365296, throughput 12.8508K wps
[Epoch 58 Batch 90/173] avg loss 0.00365918, throughput 12.8152K wps
[Epoch 58 Batch 120/173] avg loss 0.00348735, throughput 12.7875K wps
[Epoch 58 Batch 150/173] avg loss 0.00388489, throughput 12.85K wps
Begin Testing...
[Epoch 58] train avg loss 0.00366761, test acc 0.7885, test avg loss 0.458851, throughput 12.9048K wps
Observed Improvement.
Begin Testing...
[Epoch 59 Batch 30/173] avg loss 0.00340658, throughput 13.2361K wps
[Epoch 59 Batch 60/173] avg loss 0.00343367, throughput 12.822K wps
[Epoch 59 Batch 90/173] avg loss 0.00364744, throughput 12.9759K wps
[Epoch 59 Batch 120/173] avg loss 0.0037046, throughput 12.9596K wps
[Epoch 59 Batch 150/173] avg loss 0.00349184, throughput 12.8561K wps
Begin Testing...
[Epoch 59] train avg loss 0.00358287, test acc 0.7906, test avg loss 0.463738, throughput 12.9489K wps
Observed Improvement.
Begin Testing...
Test loss 0.448524, test acc 0.7946
Total time cost 176.51s
9596 1066
[Epoch 0 Batch 30/173] avg loss 0.0152428, throughput 11.6744K wps
[Epoch 0 Batch 60/173] avg loss 0.0147207, throughput 12.7668K wps
[Epoch 0 Batch 90/173] avg loss 0.0145617, throughput 12.7693K wps
[Epoch 0 Batch 120/173] avg loss 0.0145026, throughput 12.8133K wps
[Epoch 0 Batch 150/173] avg loss 0.014528, throughput 12.9356K wps
Begin Testing...
[Epoch 0] train avg loss 0.0146153, test acc 0.6010, test avg loss 0.66883, throughput 12.6016K wps
Observed Improvement.
Begin Testing...
[Epoch 1 Batch 30/173] avg loss 0.0141462, throughput 13.2669K wps
[Epoch 1 Batch 60/173] avg loss 0.0141538, throughput 12.7306K wps
[Epoch 1 Batch 90/173] avg loss 0.0138595, throughput 12.8247K wps
[Epoch 1 Batch 120/173] avg loss 0.0135839, throughput 12.7659K wps
[Epoch 1 Batch 150/173] avg loss 0.0133585, throughput 12.7739K wps
Begin Testing...
[Epoch 1] train avg loss 0.0137993, test acc 0.6146, test avg loss 0.654177, throughput 12.8818K wps
Observed Improvement.
Begin Testing...
[Epoch 2 Batch 30/173] avg loss 0.0131823, throughput 13.1606K wps
[Epoch 2 Batch 60/173] avg loss 0.0133086, throughput 12.7606K wps
[Epoch 2 Batch 90/173] avg loss 0.013133, throughput 12.7856K wps
[Epoch 2 Batch 120/173] avg loss 0.0131993, throughput 12.8106K wps
[Epoch 2 Batch 150/173] avg loss 0.0133309, throughput 12.8034K wps
Begin Testing...
[Epoch 2] train avg loss 0.0132448, test acc 0.6000, test avg loss 0.647682, throughput 12.8739K wps
[Epoch 3 Batch 30/173] avg loss 0.0129001, throughput 13.0937K wps
[Epoch 3 Batch 60/173] avg loss 0.0126516, throughput 12.8076K wps
[Epoch 3 Batch 90/173] avg loss 0.013, throughput 12.9081K wps
[Epoch 3 Batch 120/173] avg loss 0.0124683, throughput 12.8833K wps
[Epoch 3 Batch 150/173] avg loss 0.0125236, throughput 12.7595K wps
Begin Testing...
[Epoch 3] train avg loss 0.0127326, test acc 0.6708, test avg loss 0.629476, throughput 12.8957K wps
Observed Improvement.
Begin Testing...
[Epoch 4 Batch 30/173] avg loss 0.0121846, throughput 13.2621K wps
[Epoch 4 Batch 60/173] avg loss 0.0123009, throughput 12.821K wps
[Epoch 4 Batch 90/173] avg loss 0.0124242, throughput 12.8948K wps
[Epoch 4 Batch 120/173] avg loss 0.0125126, throughput 12.9846K wps
[Epoch 4 Batch 150/173] avg loss 0.0122665, throughput 12.9144K wps
Begin Testing...
[Epoch 4] train avg loss 0.0123501, test acc 0.7125, test avg loss 0.6119, throughput 12.955K wps
Observed Improvement.
Begin Testing...
[Epoch 5 Batch 30/173] avg loss 0.0120304, throughput 13.2815K wps
[Epoch 5 Batch 60/173] avg loss 0.0120373, throughput 12.7864K wps
[Epoch 5 Batch 90/173] avg loss 0.0121305, throughput 12.9171K wps
[Epoch 5 Batch 120/173] avg loss 0.0120413, throughput 12.8008K wps
[Epoch 5 Batch 150/173] avg loss 0.0120095, throughput 12.7711K wps
Begin Testing...
[Epoch 5] train avg loss 0.0120432, test acc 0.7042, test avg loss 0.600617, throughput 12.8997K wps
[Epoch 6 Batch 30/173] avg loss 0.0118331, throughput 13.2543K wps
[Epoch 6 Batch 60/173] avg loss 0.0116418, throughput 12.7903K wps
[Epoch 6 Batch 90/173] avg loss 0.0114478, throughput 12.7673K wps
[Epoch 6 Batch 120/173] avg loss 0.0116841, throughput 12.8712K wps
[Epoch 6 Batch 150/173] avg loss 0.0114796, throughput 12.8597K wps
Begin Testing...
[Epoch 6] train avg loss 0.0116569, test acc 0.7188, test avg loss 0.590604, throughput 12.8941K wps
Observed Improvement.
Begin Testing...
[Epoch 7 Batch 30/173] avg loss 0.011293, throughput 13.2975K wps
[Epoch 7 Batch 60/173] avg loss 0.0116101, throughput 12.7734K wps
[Epoch 7 Batch 90/173] avg loss 0.0116064, throughput 12.8033K wps
[Epoch 7 Batch 120/173] avg loss 0.0113093, throughput 12.7707K wps
[Epoch 7 Batch 150/173] avg loss 0.0112388, throughput 12.8979K wps
Begin Testing...
[Epoch 7] train avg loss 0.0114401, test acc 0.7385, test avg loss 0.57223, throughput 12.9178K wps
Observed Improvement.
Begin Testing...
[Epoch 8 Batch 30/173] avg loss 0.0109634, throughput 13.2005K wps
[Epoch 8 Batch 60/173] avg loss 0.0111787, throughput 12.7743K wps
[Epoch 8 Batch 90/173] avg loss 0.0107449, throughput 12.8172K wps
[Epoch 8 Batch 120/173] avg loss 0.0110937, throughput 12.7838K wps
[Epoch 8 Batch 150/173] avg loss 0.0108394, throughput 12.8174K wps
Begin Testing...
[Epoch 8] train avg loss 0.0110065, test acc 0.7417, test avg loss 0.559466, throughput 12.8907K wps
Observed Improvement.
Begin Testing...
[Epoch 9 Batch 30/173] avg loss 0.0105679, throughput 13.2181K wps
[Epoch 9 Batch 60/173] avg loss 0.0109315, throughput 12.7743K wps
[Epoch 9 Batch 90/173] avg loss 0.0106038, throughput 12.9508K wps
[Epoch 9 Batch 120/173] avg loss 0.0105916, throughput 12.8548K wps
[Epoch 9 Batch 150/173] avg loss 0.0105019, throughput 12.7581K wps
Begin Testing...
[Epoch 9] train avg loss 0.0106386, test acc 0.7490, test avg loss 0.546249, throughput 12.8973K wps
Observed Improvement.
Begin Testing...
[Epoch 10 Batch 30/173] avg loss 0.0102117, throughput 13.242K wps
[Epoch 10 Batch 60/173] avg loss 0.0103004, throughput 12.8063K wps
[Epoch 10 Batch 90/173] avg loss 0.01057, throughput 12.9622K wps
[Epoch 10 Batch 120/173] avg loss 0.0105972, throughput 12.8065K wps
[Epoch 10 Batch 150/173] avg loss 0.010099, throughput 12.8941K wps
Begin Testing...
[Epoch 10] train avg loss 0.0103478, test acc 0.7490, test avg loss 0.535707, throughput 12.929K wps
Observed Improvement.
Begin Testing...
[Epoch 11 Batch 30/173] avg loss 0.00997879, throughput 13.3239K wps
[Epoch 11 Batch 60/173] avg loss 0.0101033, throughput 12.8106K wps
[Epoch 11 Batch 90/173] avg loss 0.00999182, throughput 12.7786K wps
[Epoch 11 Batch 120/173] avg loss 0.00987409, throughput 12.7963K wps
[Epoch 11 Batch 150/173] avg loss 0.0101817, throughput 12.8047K wps
Begin Testing...
[Epoch 11] train avg loss 0.0100593, test acc 0.7615, test avg loss 0.523114, throughput 12.9089K wps
Observed Improvement.
Begin Testing...
[Epoch 12 Batch 30/173] avg loss 0.00971304, throughput 13.197K wps
[Epoch 12 Batch 60/173] avg loss 0.00966823, throughput 12.7979K wps
[Epoch 12 Batch 90/173] avg loss 0.00962999, throughput 12.8188K wps
[Epoch 12 Batch 120/173] avg loss 0.00982177, throughput 12.9531K wps
[Epoch 12 Batch 150/173] avg loss 0.00963292, throughput 12.8156K wps
Begin Testing...
[Epoch 12] train avg loss 0.00973134, test acc 0.7562, test avg loss 0.513513, throughput 12.899K wps
[Epoch 13 Batch 30/173] avg loss 0.00926402, throughput 13.1854K wps
[Epoch 13 Batch 60/173] avg loss 0.00929456, throughput 12.7201K wps
[Epoch 13 Batch 90/173] avg loss 0.00950903, throughput 12.8933K wps
[Epoch 13 Batch 120/173] avg loss 0.00963039, throughput 12.7522K wps
[Epoch 13 Batch 150/173] avg loss 0.00981554, throughput 12.9211K wps
Begin Testing...
[Epoch 13] train avg loss 0.00951672, test acc 0.7646, test avg loss 0.512058, throughput 12.8751K wps
Observed Improvement.
Begin Testing...
[Epoch 14 Batch 30/173] avg loss 0.0094597, throughput 13.2449K wps
[Epoch 14 Batch 60/173] avg loss 0.00914976, throughput 12.6726K wps
[Epoch 14 Batch 90/173] avg loss 0.00934535, throughput 12.8802K wps
[Epoch 14 Batch 120/173] avg loss 0.00920582, throughput 12.8504K wps
[Epoch 14 Batch 150/173] avg loss 0.00888657, throughput 12.8673K wps
Begin Testing...
[Epoch 14] train avg loss 0.00926212, test acc 0.7583, test avg loss 0.509739, throughput 12.8859K wps
[Epoch 15 Batch 30/173] avg loss 0.00873182, throughput 13.1996K wps
[Epoch 15 Batch 60/173] avg loss 0.0090593, throughput 12.7725K wps
[Epoch 15 Batch 90/173] avg loss 0.00915317, throughput 12.9295K wps
[Epoch 15 Batch 120/173] avg loss 0.00912466, throughput 12.8997K wps
[Epoch 15 Batch 150/173] avg loss 0.00875719, throughput 12.7867K wps
Begin Testing...
[Epoch 15] train avg loss 0.00897753, test acc 0.7667, test avg loss 0.491154, throughput 12.913K wps
Observed Improvement.
Begin Testing...
[Epoch 16 Batch 30/173] avg loss 0.00910891, throughput 13.2477K wps
[Epoch 16 Batch 60/173] avg loss 0.008829, throughput 12.7675K wps
[Epoch 16 Batch 90/173] avg loss 0.00887202, throughput 12.8577K wps
[Epoch 16 Batch 120/173] avg loss 0.00876365, throughput 12.9568K wps
[Epoch 16 Batch 150/173] avg loss 0.00873061, throughput 12.8876K wps
Begin Testing...
[Epoch 16] train avg loss 0.00879817, test acc 0.7698, test avg loss 0.488277, throughput 12.9279K wps
Observed Improvement.
Begin Testing...
[Epoch 17 Batch 30/173] avg loss 0.00854362, throughput 13.281K wps
[Epoch 17 Batch 60/173] avg loss 0.00877197, throughput 12.795K wps
[Epoch 17 Batch 90/173] avg loss 0.00843864, throughput 12.9489K wps
[Epoch 17 Batch 120/173] avg loss 0.00844312, throughput 12.9813K wps
[Epoch 17 Batch 150/173] avg loss 0.00859772, throughput 12.9361K wps
Begin Testing...
[Epoch 17] train avg loss 0.00858237, test acc 0.7698, test avg loss 0.48382, throughput 12.9844K wps
Observed Improvement.
Begin Testing...
[Epoch 18 Batch 30/173] avg loss 0.00846001, throughput 13.2549K wps
[Epoch 18 Batch 60/173] avg loss 0.00825919, throughput 12.8265K wps
[Epoch 18 Batch 90/173] avg loss 0.0085591, throughput 12.956K wps
[Epoch 18 Batch 120/173] avg loss 0.00863444, throughput 12.8326K wps
[Epoch 18 Batch 150/173] avg loss 0.00815112, throughput 12.9134K wps
Begin Testing...
[Epoch 18] train avg loss 0.00841119, test acc 0.7677, test avg loss 0.478503, throughput 12.9423K wps
[Epoch 19 Batch 30/173] avg loss 0.0082141, throughput 13.1731K wps
[Epoch 19 Batch 60/173] avg loss 0.00815551, throughput 12.8142K wps
[Epoch 19 Batch 90/173] avg loss 0.00818757, throughput 12.9679K wps
[Epoch 19 Batch 120/173] avg loss 0.00847455, throughput 12.9245K wps
[Epoch 19 Batch 150/173] avg loss 0.00816777, throughput 12.9645K wps
Begin Testing...
[Epoch 19] train avg loss 0.00829664, test acc 0.7729, test avg loss 0.473521, throughput 12.9698K wps
Observed Improvement.
Begin Testing...
[Epoch 20 Batch 30/173] avg loss 0.00808355, throughput 13.2564K wps
[Epoch 20 Batch 60/173] avg loss 0.00807577, throughput 12.7893K wps
[Epoch 20 Batch 90/173] avg loss 0.00809028, throughput 12.8935K wps
[Epoch 20 Batch 120/173] avg loss 0.0079363, throughput 12.9376K wps
[Epoch 20 Batch 150/173] avg loss 0.00817427, throughput 12.9284K wps
Begin Testing...
[Epoch 20] train avg loss 0.00809477, test acc 0.7708, test avg loss 0.471195, throughput 12.9626K wps
[Epoch 21 Batch 30/173] avg loss 0.00804274, throughput 13.1109K wps
[Epoch 21 Batch 60/173] avg loss 0.00803225, throughput 12.8079K wps
[Epoch 21 Batch 90/173] avg loss 0.00819778, throughput 12.9436K wps
[Epoch 21 Batch 120/173] avg loss 0.00813954, throughput 12.966K wps
[Epoch 21 Batch 150/173] avg loss 0.00770431, throughput 12.788K wps
Begin Testing...
[Epoch 21] train avg loss 0.00797562, test acc 0.7677, test avg loss 0.468739, throughput 12.9078K wps
[Epoch 22 Batch 30/173] avg loss 0.00808579, throughput 13.0856K wps
[Epoch 22 Batch 60/173] avg loss 0.00775664, throughput 12.778K wps
[Epoch 22 Batch 90/173] avg loss 0.00774119, throughput 12.8202K wps
[Epoch 22 Batch 120/173] avg loss 0.00733436, throughput 12.9544K wps
[Epoch 22 Batch 150/173] avg loss 0.00773478, throughput 12.9322K wps
Begin Testing...
[Epoch 22] train avg loss 0.00774804, test acc 0.7750, test avg loss 0.468649, throughput 12.9088K wps
Observed Improvement.
Begin Testing...
[Epoch 23 Batch 30/173] avg loss 0.00779132, throughput 13.2167K wps
[Epoch 23 Batch 60/173] avg loss 0.00751588, throughput 12.7613K wps
[Epoch 23 Batch 90/173] avg loss 0.00766665, throughput 12.9615K wps
[Epoch 23 Batch 120/173] avg loss 0.00761337, throughput 12.8661K wps
[Epoch 23 Batch 150/173] avg loss 0.00779441, throughput 12.8075K wps
Begin Testing...
[Epoch 23] train avg loss 0.00765259, test acc 0.7667, test avg loss 0.475136, throughput 12.9183K wps
[Epoch 24 Batch 30/173] avg loss 0.00756684, throughput 13.2323K wps
[Epoch 24 Batch 60/173] avg loss 0.0072531, throughput 12.8036K wps
[Epoch 24 Batch 90/173] avg loss 0.00752812, throughput 12.8757K wps
[Epoch 24 Batch 120/173] avg loss 0.00740524, throughput 12.8193K wps
[Epoch 24 Batch 150/173] avg loss 0.00722131, throughput 12.8935K wps
Begin Testing...
[Epoch 24] train avg loss 0.00741971, test acc 0.7781, test avg loss 0.461488, throughput 12.9311K wps
Observed Improvement.
Begin Testing...
[Epoch 25 Batch 30/173] avg loss 0.00768979, throughput 13.2707K wps
[Epoch 25 Batch 60/173] avg loss 0.00722811, throughput 12.7705K wps
[Epoch 25 Batch 90/173] avg loss 0.00728734, throughput 12.7877K wps
[Epoch 25 Batch 120/173] avg loss 0.0070443, throughput 12.7845K wps
[Epoch 25 Batch 150/173] avg loss 0.00706195, throughput 12.7723K wps
Begin Testing...
[Epoch 25] train avg loss 0.00731599, test acc 0.7781, test avg loss 0.460158, throughput 12.8873K wps
Observed Improvement.
Begin Testing...
[Epoch 26 Batch 30/173] avg loss 0.00702848, throughput 13.1921K wps
[Epoch 26 Batch 60/173] avg loss 0.00736843, throughput 12.7549K wps
[Epoch 26 Batch 90/173] avg loss 0.00724079, throughput 12.9511K wps
[Epoch 26 Batch 120/173] avg loss 0.00717112, throughput 12.9566K wps
[Epoch 26 Batch 150/173] avg loss 0.00705303, throughput 12.842K wps
Begin Testing...
[Epoch 26] train avg loss 0.00722339, test acc 0.7844, test avg loss 0.459371, throughput 12.9268K wps
Observed Improvement.
Begin Testing...
[Epoch 27 Batch 30/173] avg loss 0.00710885, throughput 13.2661K wps
[Epoch 27 Batch 60/173] avg loss 0.00704888, throughput 12.7843K wps
[Epoch 27 Batch 90/173] avg loss 0.0073358, throughput 12.8916K wps
[Epoch 27 Batch 120/173] avg loss 0.00695588, throughput 12.8383K wps
[Epoch 27 Batch 150/173] avg loss 0.00708986, throughput 12.8321K wps
Begin Testing...
[Epoch 27] train avg loss 0.00707327, test acc 0.7781, test avg loss 0.456824, throughput 12.9083K wps
[Epoch 28 Batch 30/173] avg loss 0.00703076, throughput 13.277K wps
[Epoch 28 Batch 60/173] avg loss 0.00711359, throughput 12.8114K wps
[Epoch 28 Batch 90/173] avg loss 0.00673422, throughput 12.9637K wps
[Epoch 28 Batch 120/173] avg loss 0.00711219, throughput 12.9537K wps
[Epoch 28 Batch 150/173] avg loss 0.00706294, throughput 12.9668K wps
Begin Testing...
[Epoch 28] train avg loss 0.00696044, test acc 0.7802, test avg loss 0.460448, throughput 12.9831K wps
[Epoch 29 Batch 30/173] avg loss 0.00654768, throughput 13.2558K wps
[Epoch 29 Batch 60/173] avg loss 0.00677464, throughput 12.8025K wps
[Epoch 29 Batch 90/173] avg loss 0.00687439, throughput 12.824K wps
[Epoch 29 Batch 120/173] avg loss 0.00665672, throughput 12.9377K wps
[Epoch 29 Batch 150/173] avg loss 0.0067034, throughput 12.7853K wps
Begin Testing...
[Epoch 29] train avg loss 0.0068117, test acc 0.7792, test avg loss 0.455833, throughput 12.9305K wps
[Epoch 30 Batch 30/173] avg loss 0.00678316, throughput 13.1901K wps
[Epoch 30 Batch 60/173] avg loss 0.00686538, throughput 12.7812K wps
[Epoch 30 Batch 90/173] avg loss 0.00665113, throughput 12.7385K wps
[Epoch 30 Batch 120/173] avg loss 0.00630923, throughput 12.7899K wps
[Epoch 30 Batch 150/173] avg loss 0.00680799, throughput 12.8886K wps
Begin Testing...
[Epoch 30] train avg loss 0.0066733, test acc 0.7875, test avg loss 0.451096, throughput 12.8776K wps
Observed Improvement.
Begin Testing...
[Epoch 31 Batch 30/173] avg loss 0.00659975, throughput 13.3034K wps
[Epoch 31 Batch 60/173] avg loss 0.00638544, throughput 12.8071K wps
[Epoch 31 Batch 90/173] avg loss 0.00656025, throughput 12.9398K wps
[Epoch 31 Batch 120/173] avg loss 0.00707676, throughput 12.8713K wps
[Epoch 31 Batch 150/173] avg loss 0.00593394, throughput 12.7762K wps
Begin Testing...
[Epoch 31] train avg loss 0.00651455, test acc 0.7812, test avg loss 0.451682, throughput 12.9323K wps
[Epoch 32 Batch 30/173] avg loss 0.00627199, throughput 13.2076K wps
[Epoch 32 Batch 60/173] avg loss 0.00621114, throughput 12.7965K wps
[Epoch 32 Batch 90/173] avg loss 0.00646408, throughput 12.9075K wps
[Epoch 32 Batch 120/173] avg loss 0.00634481, throughput 12.9679K wps
[Epoch 32 Batch 150/173] avg loss 0.0064678, throughput 12.8086K wps
Begin Testing...
[Epoch 32] train avg loss 0.00637467, test acc 0.7844, test avg loss 0.450416, throughput 12.9276K wps
[Epoch 33 Batch 30/173] avg loss 0.00603403, throughput 13.2698K wps
[Epoch 33 Batch 60/173] avg loss 0.00627953, throughput 12.7781K wps
[Epoch 33 Batch 90/173] avg loss 0.006188, throughput 12.9178K wps
[Epoch 33 Batch 120/173] avg loss 0.00632797, throughput 12.8895K wps
[Epoch 33 Batch 150/173] avg loss 0.00641736, throughput 12.8957K wps
Begin Testing...
[Epoch 33] train avg loss 0.00623177, test acc 0.7896, test avg loss 0.450955, throughput 12.9525K wps
Observed Improvement.
Begin Testing...
[Epoch 34 Batch 30/173] avg loss 0.00597752, throughput 13.2132K wps
[Epoch 34 Batch 60/173] avg loss 0.00652459, throughput 12.786K wps
[Epoch 34 Batch 90/173] avg loss 0.006288, throughput 12.9418K wps
[Epoch 34 Batch 120/173] avg loss 0.00637783, throughput 12.9047K wps
[Epoch 34 Batch 150/173] avg loss 0.00600182, throughput 12.7902K wps
Begin Testing...
[Epoch 34] train avg loss 0.00622015, test acc 0.7802, test avg loss 0.454028, throughput 12.9315K wps
[Epoch 35 Batch 30/173] avg loss 0.00623032, throughput 13.2642K wps
[Epoch 35 Batch 60/173] avg loss 0.00612549, throughput 12.8102K wps
[Epoch 35 Batch 90/173] avg loss 0.00624006, throughput 12.9566K wps
[Epoch 35 Batch 120/173] avg loss 0.00604131, throughput 12.9712K wps
[Epoch 35 Batch 150/173] avg loss 0.00575501, throughput 12.9626K wps
Begin Testing...
[Epoch 35] train avg loss 0.00609595, test acc 0.7875, test avg loss 0.452284, throughput 12.9932K wps
[Epoch 36 Batch 30/173] avg loss 0.00583603, throughput 13.3231K wps
[Epoch 36 Batch 60/173] avg loss 0.00615194, throughput 12.8394K wps
[Epoch 36 Batch 90/173] avg loss 0.00593535, throughput 12.7762K wps
[Epoch 36 Batch 120/173] avg loss 0.00590272, throughput 12.8726K wps
[Epoch 36 Batch 150/173] avg loss 0.00589625, throughput 12.9586K wps
Begin Testing...
[Epoch 36] train avg loss 0.00595036, test acc 0.7885, test avg loss 0.45124, throughput 12.942K wps
[Epoch 37 Batch 30/173] avg loss 0.00624004, throughput 13.2174K wps
[Epoch 37 Batch 60/173] avg loss 0.00588385, throughput 12.7963K wps
[Epoch 37 Batch 90/173] avg loss 0.00597014, throughput 12.963K wps
[Epoch 37 Batch 120/173] avg loss 0.00565909, throughput 12.9481K wps
[Epoch 37 Batch 150/173] avg loss 0.00588683, throughput 12.9183K wps
Begin Testing...
[Epoch 37] train avg loss 0.00590012, test acc 0.7854, test avg loss 0.44948, throughput 12.9521K wps
[Epoch 38 Batch 30/173] avg loss 0.00567332, throughput 13.276K wps
[Epoch 38 Batch 60/173] avg loss 0.00544514, throughput 12.7811K wps
[Epoch 38 Batch 90/173] avg loss 0.00571777, throughput 12.8614K wps
[Epoch 38 Batch 120/173] avg loss 0.00579337, throughput 12.968K wps
[Epoch 38 Batch 150/173] avg loss 0.00591231, throughput 12.9441K wps
Begin Testing...
[Epoch 38] train avg loss 0.00571841, test acc 0.7927, test avg loss 0.449601, throughput 12.9465K wps
Observed Improvement.
Begin Testing...
[Epoch 39 Batch 30/173] avg loss 0.00555381, throughput 13.2636K wps
[Epoch 39 Batch 60/173] avg loss 0.0056748, throughput 12.7832K wps
[Epoch 39 Batch 90/173] avg loss 0.00543016, throughput 12.8533K wps
[Epoch 39 Batch 120/173] avg loss 0.00592483, throughput 12.891K wps
[Epoch 39 Batch 150/173] avg loss 0.00554589, throughput 12.8066K wps
Begin Testing...
[Epoch 39] train avg loss 0.0056262, test acc 0.7896, test avg loss 0.452468, throughput 12.9095K wps
[Epoch 40 Batch 30/173] avg loss 0.00555143, throughput 13.2743K wps
[Epoch 40 Batch 60/173] avg loss 0.00555345, throughput 12.7916K wps
[Epoch 40 Batch 90/173] avg loss 0.00558856, throughput 12.9035K wps
[Epoch 40 Batch 120/173] avg loss 0.00535745, throughput 12.9057K wps
[Epoch 40 Batch 150/173] avg loss 0.00578999, throughput 12.7759K wps
Begin Testing...
[Epoch 40] train avg loss 0.00552272, test acc 0.7875, test avg loss 0.456279, throughput 12.9349K wps
[Epoch 41 Batch 30/173] avg loss 0.00547088, throughput 13.2669K wps
[Epoch 41 Batch 60/173] avg loss 0.00511363, throughput 12.8884K wps
[Epoch 41 Batch 90/173] avg loss 0.00518816, throughput 12.9591K wps
[Epoch 41 Batch 120/173] avg loss 0.00562839, throughput 12.9486K wps
[Epoch 41 Batch 150/173] avg loss 0.00521179, throughput 12.8346K wps
Begin Testing...
[Epoch 41] train avg loss 0.00533954, test acc 0.7948, test avg loss 0.449729, throughput 12.9772K wps
Observed Improvement.
Begin Testing...
[Epoch 42 Batch 30/173] avg loss 0.00516374, throughput 13.2734K wps
[Epoch 42 Batch 60/173] avg loss 0.00515159, throughput 12.7906K wps
[Epoch 42 Batch 90/173] avg loss 0.00525356, throughput 12.8467K wps
[Epoch 42 Batch 120/173] avg loss 0.00556608, throughput 12.9304K wps
[Epoch 42 Batch 150/173] avg loss 0.00531424, throughput 12.9378K wps
Begin Testing...
[Epoch 42] train avg loss 0.00530119, test acc 0.7896, test avg loss 0.450675, throughput 12.938K wps
[Epoch 43 Batch 30/173] avg loss 0.00548464, throughput 13.2325K wps
[Epoch 43 Batch 60/173] avg loss 0.0051081, throughput 12.7455K wps
[Epoch 43 Batch 90/173] avg loss 0.0049617, throughput 12.8879K wps
[Epoch 43 Batch 120/173] avg loss 0.00490707, throughput 12.8256K wps
[Epoch 43 Batch 150/173] avg loss 0.00514412, throughput 12.8863K wps
Begin Testing...
[Epoch 43] train avg loss 0.00516037, test acc 0.7927, test avg loss 0.451054, throughput 12.9224K wps
[Epoch 44 Batch 30/173] avg loss 0.00488593, throughput 13.2096K wps
[Epoch 44 Batch 60/173] avg loss 0.00508555, throughput 12.7757K wps
[Epoch 44 Batch 90/173] avg loss 0.00535462, throughput 12.8551K wps
[Epoch 44 Batch 120/173] avg loss 0.00516862, throughput 12.7984K wps
[Epoch 44 Batch 150/173] avg loss 0.00505829, throughput 12.8478K wps
Begin Testing...
[Epoch 44] train avg loss 0.00514339, test acc 0.7906, test avg loss 0.449089, throughput 12.8781K wps
[Epoch 45 Batch 30/173] avg loss 0.00508845, throughput 13.1365K wps
[Epoch 45 Batch 60/173] avg loss 0.00497883, throughput 12.7755K wps
[Epoch 45 Batch 90/173] avg loss 0.00490851, throughput 12.7874K wps
[Epoch 45 Batch 120/173] avg loss 0.00509275, throughput 12.7816K wps
[Epoch 45 Batch 150/173] avg loss 0.00459775, throughput 12.8816K wps
Begin Testing...
[Epoch 45] train avg loss 0.00495548, test acc 0.7958, test avg loss 0.452437, throughput 12.8673K wps
Observed Improvement.
Begin Testing...
[Epoch 46 Batch 30/173] avg loss 0.00501474, throughput 13.2403K wps
[Epoch 46 Batch 60/173] avg loss 0.00468508, throughput 12.7948K wps
[Epoch 46 Batch 90/173] avg loss 0.00513447, throughput 12.7588K wps
[Epoch 46 Batch 120/173] avg loss 0.00474654, throughput 12.9186K wps
[Epoch 46 Batch 150/173] avg loss 0.00512449, throughput 12.9262K wps
Begin Testing...
[Epoch 46] train avg loss 0.00491758, test acc 0.7885, test avg loss 0.451219, throughput 12.923K wps
[Epoch 47 Batch 30/173] avg loss 0.00470895, throughput 13.26K wps
[Epoch 47 Batch 60/173] avg loss 0.00491853, throughput 12.801K wps
[Epoch 47 Batch 90/173] avg loss 0.00472786, throughput 12.8726K wps
[Epoch 47 Batch 120/173] avg loss 0.00478283, throughput 12.9485K wps
[Epoch 47 Batch 150/173] avg loss 0.00483256, throughput 12.7944K wps
Begin Testing...
[Epoch 47] train avg loss 0.00478862, test acc 0.7865, test avg loss 0.453931, throughput 12.9395K wps
[Epoch 48 Batch 30/173] avg loss 0.00471405, throughput 13.2486K wps
[Epoch 48 Batch 60/173] avg loss 0.00436667, throughput 12.7495K wps
[Epoch 48 Batch 90/173] avg loss 0.00460176, throughput 12.9099K wps
[Epoch 48 Batch 120/173] avg loss 0.0047637, throughput 12.8802K wps
[Epoch 48 Batch 150/173] avg loss 0.00519834, throughput 12.867K wps
Begin Testing...
[Epoch 48] train avg loss 0.00476268, test acc 0.7802, test avg loss 0.45602, throughput 12.9063K wps
[Epoch 49 Batch 30/173] avg loss 0.00440451, throughput 13.1357K wps
[Epoch 49 Batch 60/173] avg loss 0.00445611, throughput 12.806K wps
[Epoch 49 Batch 90/173] avg loss 0.0046065, throughput 12.9583K wps
[Epoch 49 Batch 120/173] avg loss 0.00449033, throughput 12.9267K wps
[Epoch 49 Batch 150/173] avg loss 0.00463133, throughput 12.8944K wps
Begin Testing...
[Epoch 49] train avg loss 0.00453096, test acc 0.7937, test avg loss 0.453874, throughput 12.9375K wps
[Epoch 50 Batch 30/173] avg loss 0.00411116, throughput 13.1805K wps
[Epoch 50 Batch 60/173] avg loss 0.00440719, throughput 12.7744K wps
[Epoch 50 Batch 90/173] avg loss 0.0041443, throughput 12.9341K wps
[Epoch 50 Batch 120/173] avg loss 0.00461916, throughput 12.9391K wps
[Epoch 50 Batch 150/173] avg loss 0.004756, throughput 12.9017K wps
Begin Testing...
[Epoch 50] train avg loss 0.00442167, test acc 0.7937, test avg loss 0.45824, throughput 12.9289K wps
[Epoch 51 Batch 30/173] avg loss 0.00480964, throughput 13.2522K wps
[Epoch 51 Batch 60/173] avg loss 0.00462351, throughput 12.7349K wps
[Epoch 51 Batch 90/173] avg loss 0.00446369, throughput 12.8905K wps
[Epoch 51 Batch 120/173] avg loss 0.00468148, throughput 12.9367K wps
[Epoch 51 Batch 150/173] avg loss 0.00434632, throughput 12.9547K wps
Begin Testing...
[Epoch 51] train avg loss 0.00453697, test acc 0.7833, test avg loss 0.457438, throughput 12.9539K wps
[Epoch 52 Batch 30/173] avg loss 0.00436081, throughput 13.0776K wps
[Epoch 52 Batch 60/173] avg loss 0.00418361, throughput 12.8745K wps
[Epoch 52 Batch 90/173] avg loss 0.0041032, throughput 12.9082K wps
[Epoch 52 Batch 120/173] avg loss 0.00432752, throughput 12.9094K wps
[Epoch 52 Batch 150/173] avg loss 0.00401164, throughput 12.8941K wps
Begin Testing...
[Epoch 52] train avg loss 0.00425834, test acc 0.7969, test avg loss 0.450921, throughput 12.9199K wps
Observed Improvement.
Begin Testing...
[Epoch 53 Batch 30/173] avg loss 0.00390495, throughput 13.3051K wps
[Epoch 53 Batch 60/173] avg loss 0.00449239, throughput 12.8019K wps
[Epoch 53 Batch 90/173] avg loss 0.00419768, throughput 12.8346K wps
[Epoch 53 Batch 120/173] avg loss 0.0041194, throughput 12.92K wps
[Epoch 53 Batch 150/173] avg loss 0.00420822, throughput 12.9639K wps
Begin Testing...
[Epoch 53] train avg loss 0.00419085, test acc 0.7969, test avg loss 0.453645, throughput 12.968K wps
Observed Improvement.
Begin Testing...
[Epoch 54 Batch 30/173] avg loss 0.003999, throughput 13.1739K wps
[Epoch 54 Batch 60/173] avg loss 0.0042399, throughput 12.7765K wps
[Epoch 54 Batch 90/173] avg loss 0.00404003, throughput 12.9295K wps
[Epoch 54 Batch 120/173] avg loss 0.00401723, throughput 13.0108K wps
[Epoch 54 Batch 150/173] avg loss 0.00422406, throughput 12.8145K wps
Begin Testing...
[Epoch 54] train avg loss 0.00407816, test acc 0.7885, test avg loss 0.460251, throughput 12.9393K wps
[Epoch 55 Batch 30/173] avg loss 0.00411006, throughput 13.2351K wps
[Epoch 55 Batch 60/173] avg loss 0.00381366, throughput 12.7981K wps
[Epoch 55 Batch 90/173] avg loss 0.00397656, throughput 12.805K wps
[Epoch 55 Batch 120/173] avg loss 0.00408371, throughput 12.9698K wps
[Epoch 55 Batch 150/173] avg loss 0.00384574, throughput 12.9291K wps
Begin Testing...
[Epoch 55] train avg loss 0.00397705, test acc 0.7875, test avg loss 0.458764, throughput 12.9273K wps
[Epoch 56 Batch 30/173] avg loss 0.00395954, throughput 13.2758K wps
[Epoch 56 Batch 60/173] avg loss 0.00377877, throughput 12.8329K wps
[Epoch 56 Batch 90/173] avg loss 0.00375084, throughput 12.9382K wps
[Epoch 56 Batch 120/173] avg loss 0.00382146, throughput 12.9564K wps
[Epoch 56 Batch 150/173] avg loss 0.00373116, throughput 12.7039K wps
Begin Testing...
[Epoch 56] train avg loss 0.00388636, test acc 0.7927, test avg loss 0.459109, throughput 12.9398K wps
[Epoch 57 Batch 30/173] avg loss 0.00393309, throughput 13.1379K wps
[Epoch 57 Batch 60/173] avg loss 0.00392789, throughput 12.7864K wps
[Epoch 57 Batch 90/173] avg loss 0.00401969, throughput 12.9721K wps
[Epoch 57 Batch 120/173] avg loss 0.00395036, throughput 12.9243K wps
[Epoch 57 Batch 150/173] avg loss 0.00376909, throughput 12.9178K wps
Begin Testing...
[Epoch 57] train avg loss 0.00388577, test acc 0.7896, test avg loss 0.459172, throughput 12.9447K wps
[Epoch 58 Batch 30/173] avg loss 0.00385398, throughput 13.083K wps
[Epoch 58 Batch 60/173] avg loss 0.00348405, throughput 12.7516K wps
[Epoch 58 Batch 90/173] avg loss 0.00394508, throughput 12.7709K wps
[Epoch 58 Batch 120/173] avg loss 0.00357632, throughput 12.7386K wps
[Epoch 58 Batch 150/173] avg loss 0.0038767, throughput 12.7496K wps
Begin Testing...
[Epoch 58] train avg loss 0.00375922, test acc 0.8010, test avg loss 0.45453, throughput 12.8297K wps
Observed Improvement.
Begin Testing...
[Epoch 59 Batch 30/173] avg loss 0.00332288, throughput 13.2487K wps
[Epoch 59 Batch 60/173] avg loss 0.00371909, throughput 12.7411K wps
[Epoch 59 Batch 90/173] avg loss 0.0034172, throughput 12.8212K wps
[Epoch 59 Batch 120/173] avg loss 0.00373705, throughput 12.8898K wps
[Epoch 59 Batch 150/173] avg loss 0.0038282, throughput 12.9419K wps
Begin Testing...
[Epoch 59] train avg loss 0.00363553, test acc 0.7906, test avg loss 0.458758, throughput 12.9331K wps
Test loss 0.43467, test acc 0.8039
Total time cost 177.14s
9596 1066
[Epoch 0 Batch 30/173] avg loss 0.0152416, throughput 11.7765K wps
[Epoch 0 Batch 60/173] avg loss 0.0148996, throughput 12.704K wps
[Epoch 0 Batch 90/173] avg loss 0.0148569, throughput 12.7911K wps
[Epoch 0 Batch 120/173] avg loss 0.014735, throughput 12.8968K wps
[Epoch 0 Batch 150/173] avg loss 0.0145503, throughput 12.8704K wps
Begin Testing...
[Epoch 0] train avg loss 0.0148231, test acc 0.5927, test avg loss 0.6671, throughput 12.6331K wps
Observed Improvement.
Begin Testing...
[Epoch 1 Batch 30/173] avg loss 0.0138203, throughput 13.1799K wps
[Epoch 1 Batch 60/173] avg loss 0.0135717, throughput 12.7912K wps
[Epoch 1 Batch 90/173] avg loss 0.0136764, throughput 12.8099K wps
[Epoch 1 Batch 120/173] avg loss 0.0136713, throughput 12.7908K wps
[Epoch 1 Batch 150/173] avg loss 0.0136461, throughput 12.7919K wps
Begin Testing...
[Epoch 1] train avg loss 0.0137064, test acc 0.6198, test avg loss 0.649249, throughput 12.8593K wps
Observed Improvement.
Begin Testing...
[Epoch 2 Batch 30/173] avg loss 0.0133821, throughput 13.102K wps
[Epoch 2 Batch 60/173] avg loss 0.0131651, throughput 12.7961K wps
[Epoch 2 Batch 90/173] avg loss 0.0130871, throughput 12.905K wps
[Epoch 2 Batch 120/173] avg loss 0.0129358, throughput 12.7685K wps
[Epoch 2 Batch 150/173] avg loss 0.0131214, throughput 12.7697K wps
Begin Testing...
[Epoch 2] train avg loss 0.0131339, test acc 0.6479, test avg loss 0.641355, throughput 12.8583K wps
Observed Improvement.
Begin Testing...
[Epoch 3 Batch 30/173] avg loss 0.0129397, throughput 13.2878K wps
[Epoch 3 Batch 60/173] avg loss 0.0127305, throughput 12.752K wps
[Epoch 3 Batch 90/173] avg loss 0.012653, throughput 12.894K wps
[Epoch 3 Batch 120/173] avg loss 0.012689, throughput 12.8466K wps
[Epoch 3 Batch 150/173] avg loss 0.0124735, throughput 12.8341K wps
Begin Testing...
[Epoch 3] train avg loss 0.0127098, test acc 0.6687, test avg loss 0.633519, throughput 12.9127K wps
Observed Improvement.
Begin Testing...
[Epoch 4 Batch 30/173] avg loss 0.0123589, throughput 13.2307K wps
[Epoch 4 Batch 60/173] avg loss 0.0124437, throughput 12.7992K wps
[Epoch 4 Batch 90/173] avg loss 0.012188, throughput 12.8911K wps
[Epoch 4 Batch 120/173] avg loss 0.0123149, throughput 12.9845K wps
[Epoch 4 Batch 150/173] avg loss 0.0122796, throughput 12.8472K wps
Begin Testing...
[Epoch 4] train avg loss 0.0123279, test acc 0.6906, test avg loss 0.617608, throughput 12.9302K wps
Observed Improvement.
Begin Testing...
[Epoch 5 Batch 30/173] avg loss 0.0115818, throughput 13.3011K wps
[Epoch 5 Batch 60/173] avg loss 0.0119489, throughput 12.8501K wps
[Epoch 5 Batch 90/173] avg loss 0.0120751, throughput 12.8261K wps
[Epoch 5 Batch 120/173] avg loss 0.0120072, throughput 12.7833K wps
[Epoch 5 Batch 150/173] avg loss 0.0119699, throughput 12.773K wps
Begin Testing...
[Epoch 5] train avg loss 0.0119389, test acc 0.7052, test avg loss 0.601421, throughput 12.8923K wps
Observed Improvement.
Begin Testing...
[Epoch 6 Batch 30/173] avg loss 0.0116845, throughput 13.129K wps
[Epoch 6 Batch 60/173] avg loss 0.0117902, throughput 12.8041K wps
[Epoch 6 Batch 90/173] avg loss 0.0117217, throughput 12.8874K wps
[Epoch 6 Batch 120/173] avg loss 0.0113631, throughput 12.7901K wps
[Epoch 6 Batch 150/173] avg loss 0.0114827, throughput 12.8613K wps
Begin Testing...
[Epoch 6] train avg loss 0.0116592, test acc 0.7167, test avg loss 0.587011, throughput 12.8876K wps
Observed Improvement.
Begin Testing...
[Epoch 7 Batch 30/173] avg loss 0.0114576, throughput 13.3063K wps
[Epoch 7 Batch 60/173] avg loss 0.0112933, throughput 12.781K wps
[Epoch 7 Batch 90/173] avg loss 0.0113342, throughput 12.8527K wps
[Epoch 7 Batch 120/173] avg loss 0.011138, throughput 12.7966K wps
[Epoch 7 Batch 150/173] avg loss 0.0111431, throughput 12.8024K wps
Begin Testing...
[Epoch 7] train avg loss 0.011299, test acc 0.7198, test avg loss 0.573181, throughput 12.896K wps
Observed Improvement.
Begin Testing...
[Epoch 8 Batch 30/173] avg loss 0.0111251, throughput 13.2787K wps
[Epoch 8 Batch 60/173] avg loss 0.01077, throughput 12.7801K wps
[Epoch 8 Batch 90/173] avg loss 0.0110053, throughput 12.7995K wps
[Epoch 8 Batch 120/173] avg loss 0.0109568, throughput 12.7918K wps
[Epoch 8 Batch 150/173] avg loss 0.0107107, throughput 12.7724K wps
Begin Testing...
[Epoch 8] train avg loss 0.0109265, test acc 0.7354, test avg loss 0.55834, throughput 12.8826K wps
Observed Improvement.
Begin Testing...
[Epoch 9 Batch 30/173] avg loss 0.0105993, throughput 13.2227K wps
[Epoch 9 Batch 60/173] avg loss 0.0108002, throughput 12.7867K wps
[Epoch 9 Batch 90/173] avg loss 0.0104369, throughput 12.86K wps
[Epoch 9 Batch 120/173] avg loss 0.0104997, throughput 12.7708K wps
[Epoch 9 Batch 150/173] avg loss 0.0107932, throughput 12.8607K wps
Begin Testing...
[Epoch 9] train avg loss 0.010606, test acc 0.7438, test avg loss 0.545519, throughput 12.9044K wps
Observed Improvement.
Begin Testing...
[Epoch 10 Batch 30/173] avg loss 0.0102119, throughput 13.3665K wps
[Epoch 10 Batch 60/173] avg loss 0.0106415, throughput 12.7769K wps
[Epoch 10 Batch 90/173] avg loss 0.0102584, throughput 12.8387K wps
[Epoch 10 Batch 120/173] avg loss 0.0102753, throughput 12.9371K wps
[Epoch 10 Batch 150/173] avg loss 0.0102455, throughput 12.8348K wps
Begin Testing...
[Epoch 10] train avg loss 0.0102896, test acc 0.7542, test avg loss 0.533307, throughput 12.9457K wps
Observed Improvement.
Begin Testing...
[Epoch 11 Batch 30/173] avg loss 0.0102244, throughput 13.2609K wps
[Epoch 11 Batch 60/173] avg loss 0.0100969, throughput 12.8025K wps
[Epoch 11 Batch 90/173] avg loss 0.00991189, throughput 12.7889K wps
[Epoch 11 Batch 120/173] avg loss 0.00969664, throughput 12.7736K wps
[Epoch 11 Batch 150/173] avg loss 0.0100708, throughput 12.9024K wps
Begin Testing...
[Epoch 11] train avg loss 0.00993718, test acc 0.7615, test avg loss 0.520505, throughput 12.9012K wps
Observed Improvement.
Begin Testing...
[Epoch 12 Batch 30/173] avg loss 0.00975588, throughput 13.2495K wps
[Epoch 12 Batch 60/173] avg loss 0.00960302, throughput 12.7828K wps
[Epoch 12 Batch 90/173] avg loss 0.00944251, throughput 12.797K wps
[Epoch 12 Batch 120/173] avg loss 0.00955897, throughput 12.8577K wps
[Epoch 12 Batch 150/173] avg loss 0.0095756, throughput 12.8437K wps
Begin Testing...
[Epoch 12] train avg loss 0.00960407, test acc 0.7688, test avg loss 0.507553, throughput 12.89K wps
Observed Improvement.
Begin Testing...
[Epoch 13 Batch 30/173] avg loss 0.0094935, throughput 13.241K wps
[Epoch 13 Batch 60/173] avg loss 0.00924733, throughput 12.8505K wps
[Epoch 13 Batch 90/173] avg loss 0.00921195, throughput 12.8646K wps
[Epoch 13 Batch 120/173] avg loss 0.00951387, throughput 12.8868K wps
[Epoch 13 Batch 150/173] avg loss 0.0092447, throughput 12.8087K wps
Begin Testing...
[Epoch 13] train avg loss 0.00933972, test acc 0.7750, test avg loss 0.49731, throughput 12.9266K wps
Observed Improvement.
Begin Testing...
[Epoch 14 Batch 30/173] avg loss 0.00891477, throughput 13.1992K wps
[Epoch 14 Batch 60/173] avg loss 0.00914441, throughput 12.782K wps
[Epoch 14 Batch 90/173] avg loss 0.00891138, throughput 12.8659K wps
[Epoch 14 Batch 120/173] avg loss 0.00908219, throughput 12.8426K wps
[Epoch 14 Batch 150/173] avg loss 0.00930132, throughput 12.8101K wps
Begin Testing...
[Epoch 14] train avg loss 0.00913082, test acc 0.7750, test avg loss 0.490761, throughput 12.8849K wps
Observed Improvement.
Begin Testing...
[Epoch 15 Batch 30/173] avg loss 0.00881125, throughput 13.2566K wps
[Epoch 15 Batch 60/173] avg loss 0.00883639, throughput 12.7673K wps
[Epoch 15 Batch 90/173] avg loss 0.00893154, throughput 12.8381K wps
[Epoch 15 Batch 120/173] avg loss 0.00889379, throughput 12.7897K wps
[Epoch 15 Batch 150/173] avg loss 0.00883732, throughput 12.7684K wps
Begin Testing...
[Epoch 15] train avg loss 0.00888747, test acc 0.7729, test avg loss 0.483315, throughput 12.8837K wps
[Epoch 16 Batch 30/173] avg loss 0.00874249, throughput 13.2476K wps
[Epoch 16 Batch 60/173] avg loss 0.00859106, throughput 12.7881K wps
[Epoch 16 Batch 90/173] avg loss 0.00905275, throughput 12.7956K wps
[Epoch 16 Batch 120/173] avg loss 0.00862084, throughput 12.8623K wps
[Epoch 16 Batch 150/173] avg loss 0.00843465, throughput 12.8232K wps
Begin Testing...
[Epoch 16] train avg loss 0.00870667, test acc 0.7740, test avg loss 0.477901, throughput 12.8912K wps
[Epoch 17 Batch 30/173] avg loss 0.0084612, throughput 13.1192K wps
[Epoch 17 Batch 60/173] avg loss 0.0084476, throughput 12.7808K wps
[Epoch 17 Batch 90/173] avg loss 0.00825636, throughput 12.853K wps
[Epoch 17 Batch 120/173] avg loss 0.00858457, throughput 12.7644K wps
[Epoch 17 Batch 150/173] avg loss 0.00870676, throughput 12.877K wps
Begin Testing...
[Epoch 17] train avg loss 0.00850357, test acc 0.7760, test avg loss 0.469897, throughput 12.8921K wps
Observed Improvement.
Begin Testing...
[Epoch 18 Batch 30/173] avg loss 0.00842613, throughput 13.257K wps
[Epoch 18 Batch 60/173] avg loss 0.00809084, throughput 12.8286K wps
[Epoch 18 Batch 90/173] avg loss 0.00809428, throughput 12.911K wps
[Epoch 18 Batch 120/173] avg loss 0.00818979, throughput 12.8064K wps
[Epoch 18 Batch 150/173] avg loss 0.00843263, throughput 12.8202K wps
Begin Testing...
[Epoch 18] train avg loss 0.00826717, test acc 0.7885, test avg loss 0.466459, throughput 12.9332K wps
Observed Improvement.
Begin Testing...
[Epoch 19 Batch 30/173] avg loss 0.0082462, throughput 13.1921K wps
[Epoch 19 Batch 60/173] avg loss 0.00836823, throughput 12.8465K wps
[Epoch 19 Batch 90/173] avg loss 0.00834733, throughput 12.8269K wps
[Epoch 19 Batch 120/173] avg loss 0.00789745, throughput 12.7469K wps
[Epoch 19 Batch 150/173] avg loss 0.00813174, throughput 12.8828K wps
Begin Testing...
[Epoch 19] train avg loss 0.00818217, test acc 0.7792, test avg loss 0.464603, throughput 12.909K wps
[Epoch 20 Batch 30/173] avg loss 0.00819581, throughput 13.2526K wps
[Epoch 20 Batch 60/173] avg loss 0.00817462, throughput 12.77K wps
[Epoch 20 Batch 90/173] avg loss 0.00794985, throughput 12.8588K wps
[Epoch 20 Batch 120/173] avg loss 0.00788836, throughput 12.8715K wps
[Epoch 20 Batch 150/173] avg loss 0.00776954, throughput 12.797K wps
Begin Testing...
[Epoch 20] train avg loss 0.00795095, test acc 0.7865, test avg loss 0.459692, throughput 12.9229K wps
[Epoch 21 Batch 30/173] avg loss 0.00784389, throughput 13.2092K wps
[Epoch 21 Batch 60/173] avg loss 0.00772576, throughput 12.8512K wps
[Epoch 21 Batch 90/173] avg loss 0.0079284, throughput 12.9657K wps
[Epoch 21 Batch 120/173] avg loss 0.00804779, throughput 12.8675K wps
[Epoch 21 Batch 150/173] avg loss 0.00749301, throughput 12.8087K wps
Begin Testing...
[Epoch 21] train avg loss 0.00782943, test acc 0.7833, test avg loss 0.457368, throughput 12.9181K wps
[Epoch 22 Batch 30/173] avg loss 0.00732515, throughput 13.2289K wps
[Epoch 22 Batch 60/173] avg loss 0.00821643, throughput 12.8023K wps
[Epoch 22 Batch 90/173] avg loss 0.00753794, throughput 12.9605K wps
[Epoch 22 Batch 120/173] avg loss 0.00740541, throughput 12.9412K wps
[Epoch 22 Batch 150/173] avg loss 0.00763724, throughput 12.9718K wps
Begin Testing...
[Epoch 22] train avg loss 0.00763567, test acc 0.7812, test avg loss 0.458522, throughput 12.9648K wps
[Epoch 23 Batch 30/173] avg loss 0.00724203, throughput 13.1796K wps
[Epoch 23 Batch 60/173] avg loss 0.00752829, throughput 12.8022K wps
[Epoch 23 Batch 90/173] avg loss 0.00746502, throughput 12.7782K wps
[Epoch 23 Batch 120/173] avg loss 0.0073297, throughput 12.9354K wps
[Epoch 23 Batch 150/173] avg loss 0.00759891, throughput 12.9356K wps
Begin Testing...
[Epoch 23] train avg loss 0.00749352, test acc 0.7937, test avg loss 0.446318, throughput 12.9073K wps
Observed Improvement.
Begin Testing...
[Epoch 24 Batch 30/173] avg loss 0.00736886, throughput 13.1664K wps
[Epoch 24 Batch 60/173] avg loss 0.00746808, throughput 12.751K wps
[Epoch 24 Batch 90/173] avg loss 0.00750499, throughput 12.9079K wps
[Epoch 24 Batch 120/173] avg loss 0.00746229, throughput 12.941K wps
[Epoch 24 Batch 150/173] avg loss 0.007162, throughput 12.9091K wps
Begin Testing...
[Epoch 24] train avg loss 0.00739463, test acc 0.7927, test avg loss 0.449808, throughput 12.9211K wps
[Epoch 25 Batch 30/173] avg loss 0.00690832, throughput 13.1891K wps
[Epoch 25 Batch 60/173] avg loss 0.0071085, throughput 12.8296K wps
[Epoch 25 Batch 90/173] avg loss 0.00732898, throughput 12.9006K wps
[Epoch 25 Batch 120/173] avg loss 0.00743118, throughput 12.7421K wps
[Epoch 25 Batch 150/173] avg loss 0.00744963, throughput 12.941K wps
Begin Testing...
[Epoch 25] train avg loss 0.00725928, test acc 0.7948, test avg loss 0.445162, throughput 12.9038K wps
Observed Improvement.
Begin Testing...
[Epoch 26 Batch 30/173] avg loss 0.0067647, throughput 13.2943K wps
[Epoch 26 Batch 60/173] avg loss 0.00707818, throughput 12.811K wps
[Epoch 26 Batch 90/173] avg loss 0.00727705, throughput 12.8232K wps
[Epoch 26 Batch 120/173] avg loss 0.00718676, throughput 12.8911K wps
[Epoch 26 Batch 150/173] avg loss 0.00675003, throughput 12.8333K wps
Begin Testing...
[Epoch 26] train avg loss 0.00707696, test acc 0.8010, test avg loss 0.443166, throughput 12.91K wps
Observed Improvement.
Begin Testing...
[Epoch 27 Batch 30/173] avg loss 0.00695672, throughput 13.2001K wps
[Epoch 27 Batch 60/173] avg loss 0.00670516, throughput 12.846K wps
[Epoch 27 Batch 90/173] avg loss 0.0068711, throughput 12.8788K wps
[Epoch 27 Batch 120/173] avg loss 0.00671374, throughput 12.7281K wps
[Epoch 27 Batch 150/173] avg loss 0.00707555, throughput 12.8913K wps
Begin Testing...
[Epoch 27] train avg loss 0.00692924, test acc 0.7969, test avg loss 0.442357, throughput 12.9142K wps
[Epoch 28 Batch 30/173] avg loss 0.00674874, throughput 13.1488K wps
[Epoch 28 Batch 60/173] avg loss 0.0069296, throughput 12.8712K wps
[Epoch 28 Batch 90/173] avg loss 0.00651294, throughput 12.9564K wps
[Epoch 28 Batch 120/173] avg loss 0.00672642, throughput 12.9369K wps
[Epoch 28 Batch 150/173] avg loss 0.00701621, throughput 12.8742K wps
Begin Testing...
[Epoch 28] train avg loss 0.00681178, test acc 0.7823, test avg loss 0.456164, throughput 12.9362K wps
[Epoch 29 Batch 30/173] avg loss 0.00673621, throughput 13.2332K wps
[Epoch 29 Batch 60/173] avg loss 0.0066168, throughput 12.8458K wps
[Epoch 29 Batch 90/173] avg loss 0.00655042, throughput 12.9567K wps
[Epoch 29 Batch 120/173] avg loss 0.00688819, throughput 12.9013K wps
[Epoch 29 Batch 150/173] avg loss 0.00641645, throughput 12.8361K wps
Begin Testing...
[Epoch 29] train avg loss 0.00669439, test acc 0.7917, test avg loss 0.442251, throughput 12.9547K wps
[Epoch 30 Batch 30/173] avg loss 0.00658469, throughput 13.2979K wps
[Epoch 30 Batch 60/173] avg loss 0.00646784, throughput 12.787K wps
[Epoch 30 Batch 90/173] avg loss 0.00649723, throughput 12.8719K wps
[Epoch 30 Batch 120/173] avg loss 0.00669683, throughput 12.7803K wps
[Epoch 30 Batch 150/173] avg loss 0.00655144, throughput 12.8617K wps
Begin Testing...
[Epoch 30] train avg loss 0.00656119, test acc 0.8010, test avg loss 0.438503, throughput 12.9151K wps
Observed Improvement.
Begin Testing...
[Epoch 31 Batch 30/173] avg loss 0.00668232, throughput 13.1902K wps
[Epoch 31 Batch 60/173] avg loss 0.00666656, throughput 12.8025K wps
[Epoch 31 Batch 90/173] avg loss 0.00618202, throughput 12.9774K wps
[Epoch 31 Batch 120/173] avg loss 0.00629025, throughput 12.8107K wps
[Epoch 31 Batch 150/173] avg loss 0.00633942, throughput 12.7852K wps
Begin Testing...
[Epoch 31] train avg loss 0.00644435, test acc 0.7958, test avg loss 0.438193, throughput 12.8945K wps
[Epoch 32 Batch 30/173] avg loss 0.00671466, throughput 13.2092K wps
[Epoch 32 Batch 60/173] avg loss 0.00646349, throughput 12.7947K wps
[Epoch 32 Batch 90/173] avg loss 0.00611852, throughput 12.7737K wps
[Epoch 32 Batch 120/173] avg loss 0.00606784, throughput 12.787K wps
[Epoch 32 Batch 150/173] avg loss 0.00619738, throughput 12.7897K wps
Begin Testing...
[Epoch 32] train avg loss 0.00632327, test acc 0.7990, test avg loss 0.438454, throughput 12.8795K wps
[Epoch 33 Batch 30/173] avg loss 0.00619174, throughput 13.2232K wps
[Epoch 33 Batch 60/173] avg loss 0.00605558, throughput 12.7989K wps
[Epoch 33 Batch 90/173] avg loss 0.00607438, throughput 12.8068K wps
[Epoch 33 Batch 120/173] avg loss 0.00619214, throughput 12.8098K wps
[Epoch 33 Batch 150/173] avg loss 0.00646876, throughput 12.8369K wps
Begin Testing...
[Epoch 33] train avg loss 0.00619219, test acc 0.8000, test avg loss 0.437344, throughput 12.8901K wps
[Epoch 34 Batch 30/173] avg loss 0.00596399, throughput 13.2507K wps
[Epoch 34 Batch 60/173] avg loss 0.00614396, throughput 12.6449K wps
[Epoch 34 Batch 90/173] avg loss 0.00629623, throughput 12.8252K wps
[Epoch 34 Batch 120/173] avg loss 0.00624756, throughput 12.8381K wps
[Epoch 34 Batch 150/173] avg loss 0.00607267, throughput 12.7454K wps
Begin Testing...
[Epoch 34] train avg loss 0.0061, test acc 0.8031, test avg loss 0.431753, throughput 12.8559K wps
Observed Improvement.
Begin Testing...
[Epoch 35 Batch 30/173] avg loss 0.00542717, throughput 13.2206K wps
[Epoch 35 Batch 60/173] avg loss 0.0060703, throughput 12.8048K wps
[Epoch 35 Batch 90/173] avg loss 0.00604735, throughput 12.7872K wps
[Epoch 35 Batch 120/173] avg loss 0.00594056, throughput 12.9249K wps
[Epoch 35 Batch 150/173] avg loss 0.00587245, throughput 12.9328K wps
Begin Testing...
[Epoch 35] train avg loss 0.00591462, test acc 0.7927, test avg loss 0.441933, throughput 12.9183K wps
[Epoch 36 Batch 30/173] avg loss 0.00573208, throughput 13.2362K wps
[Epoch 36 Batch 60/173] avg loss 0.00580836, throughput 12.8546K wps
[Epoch 36 Batch 90/173] avg loss 0.00548542, throughput 12.8766K wps
[Epoch 36 Batch 120/173] avg loss 0.00594546, throughput 12.7951K wps
[Epoch 36 Batch 150/173] avg loss 0.00611948, throughput 12.763K wps
Begin Testing...
[Epoch 36] train avg loss 0.00583143, test acc 0.8021, test avg loss 0.432012, throughput 12.8963K wps
[Epoch 37 Batch 30/173] avg loss 0.00551835, throughput 13.2163K wps
[Epoch 37 Batch 60/173] avg loss 0.0056495, throughput 12.8429K wps
[Epoch 37 Batch 90/173] avg loss 0.00585885, throughput 12.9465K wps
[Epoch 37 Batch 120/173] avg loss 0.00578237, throughput 12.8794K wps
[Epoch 37 Batch 150/173] avg loss 0.00581533, throughput 12.8183K wps
Begin Testing...
[Epoch 37] train avg loss 0.00579698, test acc 0.7948, test avg loss 0.434717, throughput 12.9423K wps
[Epoch 38 Batch 30/173] avg loss 0.00629989, throughput 13.2001K wps
[Epoch 38 Batch 60/173] avg loss 0.00584316, throughput 12.878K wps
[Epoch 38 Batch 90/173] avg loss 0.00520546, throughput 12.8297K wps
[Epoch 38 Batch 120/173] avg loss 0.00591032, throughput 12.9642K wps
[Epoch 38 Batch 150/173] avg loss 0.00549944, throughput 12.9578K wps
Begin Testing...
[Epoch 38] train avg loss 0.0057264, test acc 0.8031, test avg loss 0.4339, throughput 12.9648K wps
Observed Improvement.
Begin Testing...
[Epoch 39 Batch 30/173] avg loss 0.00535252, throughput 13.2542K wps
[Epoch 39 Batch 60/173] avg loss 0.00529099, throughput 12.7922K wps
[Epoch 39 Batch 90/173] avg loss 0.00537102, throughput 12.8733K wps
[Epoch 39 Batch 120/173] avg loss 0.00558117, throughput 12.9395K wps
[Epoch 39 Batch 150/173] avg loss 0.00555919, throughput 12.8415K wps
Begin Testing...
[Epoch 39] train avg loss 0.00545116, test acc 0.8021, test avg loss 0.43363, throughput 12.9445K wps
[Epoch 40 Batch 30/173] avg loss 0.00527682, throughput 13.1891K wps
[Epoch 40 Batch 60/173] avg loss 0.00547957, throughput 12.8622K wps
[Epoch 40 Batch 90/173] avg loss 0.00543197, throughput 12.9507K wps
[Epoch 40 Batch 120/173] avg loss 0.00532066, throughput 12.9433K wps
[Epoch 40 Batch 150/173] avg loss 0.00533701, throughput 12.9359K wps
Begin Testing...
[Epoch 40] train avg loss 0.00541403, test acc 0.7958, test avg loss 0.438027, throughput 12.9742K wps
[Epoch 41 Batch 30/173] avg loss 0.00544074, throughput 13.1703K wps
[Epoch 41 Batch 60/173] avg loss 0.0052864, throughput 12.8283K wps
[Epoch 41 Batch 90/173] avg loss 0.00554881, throughput 12.9607K wps
[Epoch 41 Batch 120/173] avg loss 0.00532747, throughput 12.906K wps
[Epoch 41 Batch 150/173] avg loss 0.00527841, throughput 12.7856K wps
Begin Testing...
[Epoch 41] train avg loss 0.00535267, test acc 0.7979, test avg loss 0.436956, throughput 12.9312K wps
[Epoch 42 Batch 30/173] avg loss 0.00515894, throughput 13.2974K wps
[Epoch 42 Batch 60/173] avg loss 0.00506767, throughput 12.8282K wps
[Epoch 42 Batch 90/173] avg loss 0.00514371, throughput 12.9049K wps
[Epoch 42 Batch 120/173] avg loss 0.0051765, throughput 12.8272K wps
[Epoch 42 Batch 150/173] avg loss 0.00507152, throughput 12.9305K wps
Begin Testing...
[Epoch 42] train avg loss 0.00511513, test acc 0.8052, test avg loss 0.428231, throughput 12.9549K wps
Observed Improvement.
Begin Testing...
[Epoch 43 Batch 30/173] avg loss 0.00519548, throughput 13.3026K wps
[Epoch 43 Batch 60/173] avg loss 0.00495265, throughput 12.7964K wps
[Epoch 43 Batch 90/173] avg loss 0.00496262, throughput 12.9241K wps
[Epoch 43 Batch 120/173] avg loss 0.00482726, throughput 12.8095K wps
[Epoch 43 Batch 150/173] avg loss 0.00515752, throughput 12.9232K wps
Begin Testing...
[Epoch 43] train avg loss 0.00503927, test acc 0.7937, test avg loss 0.436095, throughput 12.9528K wps
[Epoch 44 Batch 30/173] avg loss 0.00479312, throughput 13.2029K wps
[Epoch 44 Batch 60/173] avg loss 0.00487277, throughput 12.8469K wps
[Epoch 44 Batch 90/173] avg loss 0.00484501, throughput 12.9534K wps
[Epoch 44 Batch 120/173] avg loss 0.00515334, throughput 12.8613K wps
[Epoch 44 Batch 150/173] avg loss 0.00483487, throughput 12.8719K wps
Begin Testing...
[Epoch 44] train avg loss 0.00496156, test acc 0.8031, test avg loss 0.436472, throughput 12.9555K wps
[Epoch 45 Batch 30/173] avg loss 0.0048268, throughput 13.2432K wps
[Epoch 45 Batch 60/173] avg loss 0.00494457, throughput 12.7797K wps
[Epoch 45 Batch 90/173] avg loss 0.00455517, throughput 12.9643K wps
[Epoch 45 Batch 120/173] avg loss 0.00476862, throughput 12.9683K wps
[Epoch 45 Batch 150/173] avg loss 0.00510327, throughput 12.9444K wps
Begin Testing...
[Epoch 45] train avg loss 0.00486886, test acc 0.8031, test avg loss 0.429862, throughput 12.9735K wps
[Epoch 46 Batch 30/173] avg loss 0.00460195, throughput 13.2372K wps
[Epoch 46 Batch 60/173] avg loss 0.00446383, throughput 12.7749K wps
[Epoch 46 Batch 90/173] avg loss 0.00474233, throughput 12.9546K wps
[Epoch 46 Batch 120/173] avg loss 0.00507957, throughput 12.9269K wps
[Epoch 46 Batch 150/173] avg loss 0.00487477, throughput 12.9476K wps
Begin Testing...
[Epoch 46] train avg loss 0.00481401, test acc 0.8031, test avg loss 0.426978, throughput 12.9703K wps
[Epoch 47 Batch 30/173] avg loss 0.00507767, throughput 13.2058K wps
[Epoch 47 Batch 60/173] avg loss 0.00446577, throughput 12.8095K wps
[Epoch 47 Batch 90/173] avg loss 0.00456267, throughput 12.8627K wps
[Epoch 47 Batch 120/173] avg loss 0.00476079, throughput 12.9222K wps
[Epoch 47 Batch 150/173] avg loss 0.0045675, throughput 12.9391K wps
Begin Testing...
[Epoch 47] train avg loss 0.00470043, test acc 0.7990, test avg loss 0.433743, throughput 12.9491K wps
[Epoch 48 Batch 30/173] avg loss 0.00453391, throughput 13.2912K wps
[Epoch 48 Batch 60/173] avg loss 0.00468547, throughput 12.8072K wps
[Epoch 48 Batch 90/173] avg loss 0.00420732, throughput 12.8812K wps
[Epoch 48 Batch 120/173] avg loss 0.00479361, throughput 12.9093K wps
[Epoch 48 Batch 150/173] avg loss 0.00466335, throughput 12.7867K wps
Begin Testing...
[Epoch 48] train avg loss 0.00460989, test acc 0.8031, test avg loss 0.433934, throughput 12.9237K wps
[Epoch 49 Batch 30/173] avg loss 0.00419339, throughput 13.2216K wps
[Epoch 49 Batch 60/173] avg loss 0.00440915, throughput 12.7709K wps
[Epoch 49 Batch 90/173] avg loss 0.00472706, throughput 12.8858K wps
[Epoch 49 Batch 120/173] avg loss 0.00478357, throughput 12.8039K wps
[Epoch 49 Batch 150/173] avg loss 0.00423194, throughput 12.7797K wps
Begin Testing...
[Epoch 49] train avg loss 0.00449599, test acc 0.8000, test avg loss 0.433178, throughput 12.8756K wps
[Epoch 50 Batch 30/173] avg loss 0.00436899, throughput 13.2405K wps
[Epoch 50 Batch 60/173] avg loss 0.00444467, throughput 12.8719K wps
[Epoch 50 Batch 90/173] avg loss 0.00430084, throughput 12.943K wps
[Epoch 50 Batch 120/173] avg loss 0.00461714, throughput 12.7565K wps
[Epoch 50 Batch 150/173] avg loss 0.00486936, throughput 12.9494K wps
Begin Testing...
[Epoch 50] train avg loss 0.0044834, test acc 0.7927, test avg loss 0.444595, throughput 12.9412K wps
[Epoch 51 Batch 30/173] avg loss 0.00429903, throughput 13.2195K wps
[Epoch 51 Batch 60/173] avg loss 0.00419708, throughput 12.7957K wps
[Epoch 51 Batch 90/173] avg loss 0.0041067, throughput 12.7902K wps
[Epoch 51 Batch 120/173] avg loss 0.00409712, throughput 12.8547K wps
[Epoch 51 Batch 150/173] avg loss 0.00425446, throughput 12.9258K wps
Begin Testing...
[Epoch 51] train avg loss 0.00423827, test acc 0.8021, test avg loss 0.432822, throughput 12.9062K wps
[Epoch 52 Batch 30/173] avg loss 0.00433417, throughput 13.1878K wps
[Epoch 52 Batch 60/173] avg loss 0.00419981, throughput 12.8539K wps
[Epoch 52 Batch 90/173] avg loss 0.00419504, throughput 12.165K wps
[Epoch 52 Batch 120/173] avg loss 0.00454505, throughput 12.9654K wps
[Epoch 52 Batch 150/173] avg loss 0.00419409, throughput 12.9621K wps
Begin Testing...
[Epoch 52] train avg loss 0.00427043, test acc 0.8104, test avg loss 0.432239, throughput 12.8364K wps
Observed Improvement.
Begin Testing...
[Epoch 53 Batch 30/173] avg loss 0.0039017, throughput 13.2215K wps
[Epoch 53 Batch 60/173] avg loss 0.00393939, throughput 12.8013K wps
[Epoch 53 Batch 90/173] avg loss 0.00434096, throughput 12.8201K wps
[Epoch 53 Batch 120/173] avg loss 0.00418256, throughput 12.9583K wps
[Epoch 53 Batch 150/173] avg loss 0.00417906, throughput 12.7956K wps
Begin Testing...
[Epoch 53] train avg loss 0.00412211, test acc 0.7979, test avg loss 0.436419, throughput 12.9105K wps
[Epoch 54 Batch 30/173] avg loss 0.00411843, throughput 13.084K wps
[Epoch 54 Batch 60/173] avg loss 0.00417001, throughput 12.7998K wps
[Epoch 54 Batch 90/173] avg loss 0.00417052, throughput 12.7845K wps
[Epoch 54 Batch 120/173] avg loss 0.00385326, throughput 12.945K wps
[Epoch 54 Batch 150/173] avg loss 0.00391115, throughput 12.9282K wps
Begin Testing...
[Epoch 54] train avg loss 0.0040716, test acc 0.7896, test avg loss 0.448476, throughput 12.9206K wps
[Epoch 55 Batch 30/173] avg loss 0.00377975, throughput 13.2262K wps
[Epoch 55 Batch 60/173] avg loss 0.00411056, throughput 12.7979K wps
[Epoch 55 Batch 90/173] avg loss 0.00396902, throughput 12.9268K wps
[Epoch 55 Batch 120/173] avg loss 0.00409063, throughput 12.7839K wps
[Epoch 55 Batch 150/173] avg loss 0.00393601, throughput 12.7709K wps
Begin Testing...
[Epoch 55] train avg loss 0.00397917, test acc 0.7979, test avg loss 0.433851, throughput 12.9073K wps
[Epoch 56 Batch 30/173] avg loss 0.00367848, throughput 13.2168K wps
[Epoch 56 Batch 60/173] avg loss 0.00407114, throughput 12.8333K wps
[Epoch 56 Batch 90/173] avg loss 0.00385651, throughput 12.9223K wps
[Epoch 56 Batch 120/173] avg loss 0.00377286, throughput 12.8131K wps
[Epoch 56 Batch 150/173] avg loss 0.00384284, throughput 12.904K wps
Begin Testing...
[Epoch 56] train avg loss 0.00386122, test acc 0.7927, test avg loss 0.445353, throughput 12.9428K wps
[Epoch 57 Batch 30/173] avg loss 0.00391635, throughput 13.2527K wps
[Epoch 57 Batch 60/173] avg loss 0.00355428, throughput 12.8548K wps
[Epoch 57 Batch 90/173] avg loss 0.00378404, throughput 12.9419K wps
[Epoch 57 Batch 120/173] avg loss 0.00375473, throughput 12.9264K wps
[Epoch 57 Batch 150/173] avg loss 0.00364835, throughput 12.9629K wps
Begin Testing...
[Epoch 57] train avg loss 0.00374816, test acc 0.8031, test avg loss 0.432026, throughput 12.9882K wps
[Epoch 58 Batch 30/173] avg loss 0.00366866, throughput 13.1868K wps
[Epoch 58 Batch 60/173] avg loss 0.00374618, throughput 12.7918K wps
[Epoch 58 Batch 90/173] avg loss 0.00382135, throughput 12.8193K wps
[Epoch 58 Batch 120/173] avg loss 0.00390214, throughput 12.898K wps
[Epoch 58 Batch 150/173] avg loss 0.00366225, throughput 12.9276K wps
Begin Testing...
[Epoch 58] train avg loss 0.00372073, test acc 0.7958, test avg loss 0.456673, throughput 12.9103K wps
[Epoch 59 Batch 30/173] avg loss 0.00380383, throughput 13.213K wps
[Epoch 59 Batch 60/173] avg loss 0.00342784, throughput 12.7584K wps
[Epoch 59 Batch 90/173] avg loss 0.00356042, throughput 12.93K wps
[Epoch 59 Batch 120/173] avg loss 0.00375755, throughput 12.9497K wps
[Epoch 59 Batch 150/173] avg loss 0.00364487, throughput 12.9282K wps
Begin Testing...
[Epoch 59] train avg loss 0.00364991, test acc 0.7927, test avg loss 0.447884, throughput 12.9553K wps
Test loss 0.480164, test acc 0.7730
Total time cost 176.83s
9596 1066
[Epoch 0 Batch 30/173] avg loss 0.015568, throughput 11.8124K wps
[Epoch 0 Batch 60/173] avg loss 0.0154179, throughput 12.6626K wps
[Epoch 0 Batch 90/173] avg loss 0.0146213, throughput 12.971K wps
[Epoch 0 Batch 120/173] avg loss 0.0144986, throughput 12.7859K wps
[Epoch 0 Batch 150/173] avg loss 0.0140873, throughput 12.9188K wps
Begin Testing...
[Epoch 0] train avg loss 0.0147921, test acc 0.5750, test avg loss 0.67557, throughput 12.6516K wps
Observed Improvement.
Begin Testing...
[Epoch 1 Batch 30/173] avg loss 0.0142139, throughput 13.237K wps
[Epoch 1 Batch 60/173] avg loss 0.0139249, throughput 12.792K wps
[Epoch 1 Batch 90/173] avg loss 0.0136179, throughput 12.9437K wps
[Epoch 1 Batch 120/173] avg loss 0.0136397, throughput 12.7899K wps
[Epoch 1 Batch 150/173] avg loss 0.0137021, throughput 12.8648K wps
Begin Testing...
[Epoch 1] train avg loss 0.0137824, test acc 0.6146, test avg loss 0.657156, throughput 12.9205K wps
Observed Improvement.
Begin Testing...
[Epoch 2 Batch 30/173] avg loss 0.0132588, throughput 13.2863K wps
[Epoch 2 Batch 60/173] avg loss 0.0130712, throughput 12.7665K wps
[Epoch 2 Batch 90/173] avg loss 0.0131879, throughput 12.8972K wps
[Epoch 2 Batch 120/173] avg loss 0.0129878, throughput 12.7377K wps
[Epoch 2 Batch 150/173] avg loss 0.0129248, throughput 12.8898K wps
Begin Testing...
[Epoch 2] train avg loss 0.0131213, test acc 0.6562, test avg loss 0.639103, throughput 12.9027K wps
Observed Improvement.
Begin Testing...
[Epoch 3 Batch 30/173] avg loss 0.0127159, throughput 13.2695K wps
[Epoch 3 Batch 60/173] avg loss 0.0127142, throughput 12.727K wps
[Epoch 3 Batch 90/173] avg loss 0.0128645, throughput 12.9621K wps
[Epoch 3 Batch 120/173] avg loss 0.0127518, throughput 12.8539K wps
[Epoch 3 Batch 150/173] avg loss 0.0124808, throughput 12.8686K wps
Begin Testing...
[Epoch 3] train avg loss 0.012727, test acc 0.6698, test avg loss 0.624911, throughput 12.9408K wps
Observed Improvement.
Begin Testing...
[Epoch 4 Batch 30/173] avg loss 0.0122881, throughput 13.2603K wps
[Epoch 4 Batch 60/173] avg loss 0.0121754, throughput 12.7791K wps
[Epoch 4 Batch 90/173] avg loss 0.0122991, throughput 12.7437K wps
[Epoch 4 Batch 120/173] avg loss 0.0122476, throughput 12.9466K wps
[Epoch 4 Batch 150/173] avg loss 0.0122734, throughput 12.8111K wps
Begin Testing...
[Epoch 4] train avg loss 0.0122923, test acc 0.6813, test avg loss 0.614971, throughput 12.9052K wps
Observed Improvement.
Begin Testing...
[Epoch 5 Batch 30/173] avg loss 0.0120534, throughput 13.2189K wps
[Epoch 5 Batch 60/173] avg loss 0.0119434, throughput 12.7386K wps
[Epoch 5 Batch 90/173] avg loss 0.012034, throughput 12.8216K wps
[Epoch 5 Batch 120/173] avg loss 0.0118711, throughput 12.7756K wps
[Epoch 5 Batch 150/173] avg loss 0.0118984, throughput 12.8454K wps
Begin Testing...
[Epoch 5] train avg loss 0.0119203, test acc 0.7094, test avg loss 0.597464, throughput 12.8937K wps
Observed Improvement.
Begin Testing...
[Epoch 6 Batch 30/173] avg loss 0.0116353, throughput 13.1978K wps
[Epoch 6 Batch 60/173] avg loss 0.0116754, throughput 12.8142K wps
[Epoch 6 Batch 90/173] avg loss 0.0114964, throughput 12.8208K wps
[Epoch 6 Batch 120/173] avg loss 0.0116588, throughput 12.7813K wps
[Epoch 6 Batch 150/173] avg loss 0.0115865, throughput 12.8888K wps
Begin Testing...
[Epoch 6] train avg loss 0.0116197, test acc 0.7271, test avg loss 0.585437, throughput 12.8986K wps
Observed Improvement.
Begin Testing...
[Epoch 7 Batch 30/173] avg loss 0.011108, throughput 13.3594K wps
[Epoch 7 Batch 60/173] avg loss 0.011653, throughput 12.7731K wps
[Epoch 7 Batch 90/173] avg loss 0.011034, throughput 12.7883K wps
[Epoch 7 Batch 120/173] avg loss 0.0114094, throughput 12.789K wps
[Epoch 7 Batch 150/173] avg loss 0.0111911, throughput 12.9062K wps
Begin Testing...
[Epoch 7] train avg loss 0.0113208, test acc 0.7177, test avg loss 0.575228, throughput 12.9178K wps
[Epoch 8 Batch 30/173] avg loss 0.0108471, throughput 13.255K wps
[Epoch 8 Batch 60/173] avg loss 0.0109442, throughput 12.8047K wps
[Epoch 8 Batch 90/173] avg loss 0.010881, throughput 12.8322K wps
[Epoch 8 Batch 120/173] avg loss 0.0109898, throughput 12.917K wps
[Epoch 8 Batch 150/173] avg loss 0.0107222, throughput 12.8668K wps
Begin Testing...
[Epoch 8] train avg loss 0.0108469, test acc 0.7271, test avg loss 0.5597, throughput 12.937K wps
Observed Improvement.
Begin Testing...
[Epoch 9 Batch 30/173] avg loss 0.0104442, throughput 13.306K wps
[Epoch 9 Batch 60/173] avg loss 0.0106108, throughput 12.7738K wps
[Epoch 9 Batch 90/173] avg loss 0.0105647, throughput 12.7524K wps
[Epoch 9 Batch 120/173] avg loss 0.0106258, throughput 12.8469K wps
[Epoch 9 Batch 150/173] avg loss 0.0103248, throughput 12.7769K wps
Begin Testing...
[Epoch 9] train avg loss 0.0105178, test acc 0.7448, test avg loss 0.545338, throughput 12.8858K wps
Observed Improvement.
Begin Testing...
[Epoch 10 Batch 30/173] avg loss 0.0101145, throughput 13.2234K wps
[Epoch 10 Batch 60/173] avg loss 0.0105014, throughput 12.7369K wps
[Epoch 10 Batch 90/173] avg loss 0.0101976, throughput 12.7876K wps
[Epoch 10 Batch 120/173] avg loss 0.0103223, throughput 12.7894K wps
[Epoch 10 Batch 150/173] avg loss 0.0101678, throughput 12.7401K wps
Begin Testing...
[Epoch 10] train avg loss 0.0103176, test acc 0.7438, test avg loss 0.535372, throughput 12.8636K wps
[Epoch 11 Batch 30/173] avg loss 0.00987714, throughput 13.2035K wps
[Epoch 11 Batch 60/173] avg loss 0.0101583, throughput 12.7751K wps
[Epoch 11 Batch 90/173] avg loss 0.0101916, throughput 12.7662K wps
[Epoch 11 Batch 120/173] avg loss 0.00979101, throughput 12.8442K wps
[Epoch 11 Batch 150/173] avg loss 0.00969369, throughput 12.7778K wps
Begin Testing...
[Epoch 11] train avg loss 0.00991257, test acc 0.7490, test avg loss 0.523382, throughput 12.8688K wps
Observed Improvement.
Begin Testing...
[Epoch 12 Batch 30/173] avg loss 0.00992777, throughput 13.2846K wps
[Epoch 12 Batch 60/173] avg loss 0.00979609, throughput 12.8032K wps
[Epoch 12 Batch 90/173] avg loss 0.00961805, throughput 12.7879K wps
[Epoch 12 Batch 120/173] avg loss 0.0096151, throughput 12.7904K wps
[Epoch 12 Batch 150/173] avg loss 0.00967206, throughput 12.9401K wps
Begin Testing...
[Epoch 12] train avg loss 0.00970304, test acc 0.7562, test avg loss 0.51343, throughput 12.9234K wps
Observed Improvement.
Begin Testing...
[Epoch 13 Batch 30/173] avg loss 0.00915886, throughput 13.2217K wps
[Epoch 13 Batch 60/173] avg loss 0.00980919, throughput 12.7831K wps
[Epoch 13 Batch 90/173] avg loss 0.00915811, throughput 12.7747K wps
[Epoch 13 Batch 120/173] avg loss 0.00928382, throughput 12.7574K wps
[Epoch 13 Batch 150/173] avg loss 0.00899467, throughput 12.7604K wps
Begin Testing...
[Epoch 13] train avg loss 0.00931213, test acc 0.7615, test avg loss 0.504476, throughput 12.8737K wps
Observed Improvement.
Begin Testing...
[Epoch 14 Batch 30/173] avg loss 0.00919732, throughput 13.3008K wps
[Epoch 14 Batch 60/173] avg loss 0.00913886, throughput 12.8143K wps
[Epoch 14 Batch 90/173] avg loss 0.00898902, throughput 12.7885K wps
[Epoch 14 Batch 120/173] avg loss 0.0089403, throughput 12.936K wps
[Epoch 14 Batch 150/173] avg loss 0.00886937, throughput 12.8342K wps
Begin Testing...
[Epoch 14] train avg loss 0.00906424, test acc 0.7646, test avg loss 0.499507, throughput 12.9242K wps
Observed Improvement.
Begin Testing...
[Epoch 15 Batch 30/173] avg loss 0.00892161, throughput 13.2465K wps
[Epoch 15 Batch 60/173] avg loss 0.00889402, throughput 12.8657K wps
[Epoch 15 Batch 90/173] avg loss 0.00859598, throughput 12.9451K wps
[Epoch 15 Batch 120/173] avg loss 0.00901993, throughput 12.7854K wps
[Epoch 15 Batch 150/173] avg loss 0.00888319, throughput 12.8099K wps
Begin Testing...
[Epoch 15] train avg loss 0.00884255, test acc 0.7521, test avg loss 0.497956, throughput 12.9319K wps
[Epoch 16 Batch 30/173] avg loss 0.00880397, throughput 13.2726K wps
[Epoch 16 Batch 60/173] avg loss 0.00878167, throughput 12.7896K wps
[Epoch 16 Batch 90/173] avg loss 0.00838735, throughput 12.7884K wps
[Epoch 16 Batch 120/173] avg loss 0.00866678, throughput 12.8532K wps
[Epoch 16 Batch 150/173] avg loss 0.00853386, throughput 12.8786K wps
Begin Testing...
[Epoch 16] train avg loss 0.00869016, test acc 0.7708, test avg loss 0.487676, throughput 12.9132K wps
Observed Improvement.
Begin Testing...
[Epoch 17 Batch 30/173] avg loss 0.00852934, throughput 13.2508K wps
[Epoch 17 Batch 60/173] avg loss 0.00851377, throughput 12.7761K wps
[Epoch 17 Batch 90/173] avg loss 0.0084949, throughput 12.9003K wps
[Epoch 17 Batch 120/173] avg loss 0.00848435, throughput 12.9197K wps
[Epoch 17 Batch 150/173] avg loss 0.0086597, throughput 12.7955K wps
Begin Testing...
[Epoch 17] train avg loss 0.00856952, test acc 0.7729, test avg loss 0.479232, throughput 12.9286K wps
Observed Improvement.
Begin Testing...
[Epoch 18 Batch 30/173] avg loss 0.00843753, throughput 13.3455K wps
[Epoch 18 Batch 60/173] avg loss 0.00826147, throughput 12.7842K wps
[Epoch 18 Batch 90/173] avg loss 0.00817228, throughput 12.9037K wps
[Epoch 18 Batch 120/173] avg loss 0.00815231, throughput 12.9578K wps
[Epoch 18 Batch 150/173] avg loss 0.00849792, throughput 12.8418K wps
Begin Testing...
[Epoch 18] train avg loss 0.00832623, test acc 0.7750, test avg loss 0.474627, throughput 12.9445K wps
Observed Improvement.
Begin Testing...
[Epoch 19 Batch 30/173] avg loss 0.00802458, throughput 13.2667K wps
[Epoch 19 Batch 60/173] avg loss 0.00803792, throughput 12.8148K wps
[Epoch 19 Batch 90/173] avg loss 0.00796856, throughput 12.982K wps
[Epoch 19 Batch 120/173] avg loss 0.00803512, throughput 12.7881K wps
[Epoch 19 Batch 150/173] avg loss 0.00827794, throughput 12.9117K wps
Begin Testing...
[Epoch 19] train avg loss 0.00808112, test acc 0.7781, test avg loss 0.473115, throughput 12.9559K wps
Observed Improvement.
Begin Testing...
[Epoch 20 Batch 30/173] avg loss 0.00808451, throughput 13.1997K wps
[Epoch 20 Batch 60/173] avg loss 0.00819946, throughput 12.7599K wps
[Epoch 20 Batch 90/173] avg loss 0.00783452, throughput 12.8053K wps
[Epoch 20 Batch 120/173] avg loss 0.00806712, throughput 12.8064K wps
[Epoch 20 Batch 150/173] avg loss 0.00777137, throughput 12.8026K wps
Begin Testing...
[Epoch 20] train avg loss 0.00796462, test acc 0.7812, test avg loss 0.465892, throughput 12.8726K wps
Observed Improvement.
Begin Testing...
[Epoch 21 Batch 30/173] avg loss 0.00757689, throughput 13.3122K wps
[Epoch 21 Batch 60/173] avg loss 0.00792761, throughput 12.8669K wps
[Epoch 21 Batch 90/173] avg loss 0.00745974, throughput 12.973K wps
[Epoch 21 Batch 120/173] avg loss 0.00765181, throughput 12.8637K wps
[Epoch 21 Batch 150/173] avg loss 0.00785178, throughput 12.8243K wps
Begin Testing...
[Epoch 21] train avg loss 0.00782222, test acc 0.7677, test avg loss 0.465061, throughput 12.9473K wps
[Epoch 22 Batch 30/173] avg loss 0.00754372, throughput 13.1302K wps
[Epoch 22 Batch 60/173] avg loss 0.00800655, throughput 12.728K wps
[Epoch 22 Batch 90/173] avg loss 0.00775662, throughput 12.9269K wps
[Epoch 22 Batch 120/173] avg loss 0.00782142, throughput 12.783K wps
[Epoch 22 Batch 150/173] avg loss 0.00748466, throughput 12.9159K wps
Begin Testing...
[Epoch 22] train avg loss 0.00767522, test acc 0.7823, test avg loss 0.46212, throughput 12.9069K wps
Observed Improvement.
Begin Testing...
[Epoch 23 Batch 30/173] avg loss 0.00751131, throughput 13.1377K wps
[Epoch 23 Batch 60/173] avg loss 0.00753365, throughput 12.7622K wps
[Epoch 23 Batch 90/173] avg loss 0.00729205, throughput 12.9367K wps
[Epoch 23 Batch 120/173] avg loss 0.00756441, throughput 12.9186K wps
[Epoch 23 Batch 150/173] avg loss 0.00722683, throughput 12.7709K wps
Begin Testing...
[Epoch 23] train avg loss 0.00746243, test acc 0.7844, test avg loss 0.461515, throughput 12.9099K wps
Observed Improvement.
Begin Testing...
[Epoch 24 Batch 30/173] avg loss 0.00715753, throughput 13.1736K wps
[Epoch 24 Batch 60/173] avg loss 0.00742298, throughput 12.7837K wps
[Epoch 24 Batch 90/173] avg loss 0.00723497, throughput 12.8063K wps
[Epoch 24 Batch 120/173] avg loss 0.00755671, throughput 12.9181K wps
[Epoch 24 Batch 150/173] avg loss 0.00731343, throughput 12.8031K wps
Begin Testing...
[Epoch 24] train avg loss 0.00734161, test acc 0.7937, test avg loss 0.459025, throughput 12.8831K wps
Observed Improvement.
Begin Testing...
[Epoch 25 Batch 30/173] avg loss 0.00686424, throughput 13.2381K wps
[Epoch 25 Batch 60/173] avg loss 0.00718738, throughput 12.8015K wps
[Epoch 25 Batch 90/173] avg loss 0.00733882, throughput 12.8568K wps
[Epoch 25 Batch 120/173] avg loss 0.00724449, throughput 12.8115K wps
[Epoch 25 Batch 150/173] avg loss 0.00723873, throughput 12.7973K wps
Begin Testing...
[Epoch 25] train avg loss 0.00722787, test acc 0.7812, test avg loss 0.46119, throughput 12.8854K wps
[Epoch 26 Batch 30/173] avg loss 0.00698186, throughput 13.2193K wps
[Epoch 26 Batch 60/173] avg loss 0.00687012, throughput 12.7682K wps
[Epoch 26 Batch 90/173] avg loss 0.00719058, throughput 12.7778K wps
[Epoch 26 Batch 120/173] avg loss 0.00729494, throughput 12.7928K wps
[Epoch 26 Batch 150/173] avg loss 0.00727624, throughput 12.7593K wps
Begin Testing...
[Epoch 26] train avg loss 0.00706582, test acc 0.7740, test avg loss 0.46709, throughput 12.8529K wps
[Epoch 27 Batch 30/173] avg loss 0.00675684, throughput 13.2385K wps
[Epoch 27 Batch 60/173] avg loss 0.00693132, throughput 12.8093K wps
[Epoch 27 Batch 90/173] avg loss 0.00714889, throughput 12.8128K wps
[Epoch 27 Batch 120/173] avg loss 0.00685735, throughput 12.8831K wps
[Epoch 27 Batch 150/173] avg loss 0.00700642, throughput 12.9479K wps
Begin Testing...
[Epoch 27] train avg loss 0.00699916, test acc 0.7958, test avg loss 0.453607, throughput 12.9213K wps
Observed Improvement.
Begin Testing...
[Epoch 28 Batch 30/173] avg loss 0.00701665, throughput 13.169K wps
[Epoch 28 Batch 60/173] avg loss 0.00695914, throughput 12.8655K wps
[Epoch 28 Batch 90/173] avg loss 0.00686772, throughput 12.9321K wps
[Epoch 28 Batch 120/173] avg loss 0.0066705, throughput 12.7809K wps
[Epoch 28 Batch 150/173] avg loss 0.00711786, throughput 12.8119K wps
Begin Testing...
[Epoch 28] train avg loss 0.00688541, test acc 0.7844, test avg loss 0.452536, throughput 12.9234K wps
[Epoch 29 Batch 30/173] avg loss 0.00634844, throughput 13.1114K wps
[Epoch 29 Batch 60/173] avg loss 0.00679103, throughput 12.8096K wps
[Epoch 29 Batch 90/173] avg loss 0.00672367, throughput 12.9457K wps
[Epoch 29 Batch 120/173] avg loss 0.006655, throughput 12.8637K wps
[Epoch 29 Batch 150/173] avg loss 0.00664621, throughput 12.7995K wps
Begin Testing...
[Epoch 29] train avg loss 0.00668756, test acc 0.7896, test avg loss 0.449305, throughput 12.8878K wps
[Epoch 30 Batch 30/173] avg loss 0.00658592, throughput 13.206K wps
[Epoch 30 Batch 60/173] avg loss 0.00685896, throughput 12.7847K wps
[Epoch 30 Batch 90/173] avg loss 0.00649983, throughput 12.8845K wps
[Epoch 30 Batch 120/173] avg loss 0.00652244, throughput 12.9664K wps
[Epoch 30 Batch 150/173] avg loss 0.00640917, throughput 12.9033K wps
Begin Testing...
[Epoch 30] train avg loss 0.00660694, test acc 0.7896, test avg loss 0.452048, throughput 12.946K wps
[Epoch 31 Batch 30/173] avg loss 0.00669576, throughput 13.1968K wps
[Epoch 31 Batch 60/173] avg loss 0.00637157, throughput 12.7608K wps
[Epoch 31 Batch 90/173] avg loss 0.00642048, throughput 12.8215K wps
[Epoch 31 Batch 120/173] avg loss 0.00622527, throughput 12.9855K wps
[Epoch 31 Batch 150/173] avg loss 0.00627988, throughput 12.8261K wps
Begin Testing...
[Epoch 31] train avg loss 0.0064396, test acc 0.7917, test avg loss 0.446397, throughput 12.9049K wps
[Epoch 32 Batch 30/173] avg loss 0.00630414, throughput 13.2314K wps
[Epoch 32 Batch 60/173] avg loss 0.00616621, throughput 12.801K wps
[Epoch 32 Batch 90/173] avg loss 0.00638788, throughput 12.809K wps
[Epoch 32 Batch 120/173] avg loss 0.00578685, throughput 12.9285K wps
[Epoch 32 Batch 150/173] avg loss 0.00604578, throughput 12.9828K wps
Begin Testing...
[Epoch 32] train avg loss 0.00624949, test acc 0.7990, test avg loss 0.447526, throughput 12.9525K wps
Observed Improvement.
Begin Testing...
[Epoch 33 Batch 30/173] avg loss 0.0063986, throughput 13.2256K wps
[Epoch 33 Batch 60/173] avg loss 0.00649754, throughput 12.8019K wps
[Epoch 33 Batch 90/173] avg loss 0.00605017, throughput 12.8265K wps
[Epoch 33 Batch 120/173] avg loss 0.00639496, throughput 12.9697K wps
[Epoch 33 Batch 150/173] avg loss 0.00572927, throughput 12.8688K wps
Begin Testing...
[Epoch 33] train avg loss 0.00619082, test acc 0.7875, test avg loss 0.448928, throughput 12.9491K wps
[Epoch 34 Batch 30/173] avg loss 0.00614076, throughput 13.2044K wps
[Epoch 34 Batch 60/173] avg loss 0.00597332, throughput 12.782K wps
[Epoch 34 Batch 90/173] avg loss 0.00603305, throughput 12.9501K wps
[Epoch 34 Batch 120/173] avg loss 0.00565828, throughput 12.8565K wps
[Epoch 34 Batch 150/173] avg loss 0.00629553, throughput 12.8881K wps
Begin Testing...
[Epoch 34] train avg loss 0.0060714, test acc 0.7937, test avg loss 0.44644, throughput 12.9414K wps
[Epoch 35 Batch 30/173] avg loss 0.00599977, throughput 13.2033K wps
[Epoch 35 Batch 60/173] avg loss 0.00594296, throughput 12.65K wps
[Epoch 35 Batch 90/173] avg loss 0.00602217, throughput 12.8848K wps
[Epoch 35 Batch 120/173] avg loss 0.0058541, throughput 12.8598K wps
[Epoch 35 Batch 150/173] avg loss 0.00628601, throughput 12.6221K wps
Begin Testing...
[Epoch 35] train avg loss 0.00598468, test acc 0.7917, test avg loss 0.446275, throughput 12.8574K wps
[Epoch 36 Batch 30/173] avg loss 0.00578252, throughput 13.1022K wps
[Epoch 36 Batch 60/173] avg loss 0.00556803, throughput 12.7594K wps
[Epoch 36 Batch 90/173] avg loss 0.00576368, throughput 12.9419K wps
[Epoch 36 Batch 120/173] avg loss 0.00564173, throughput 12.9241K wps
[Epoch 36 Batch 150/173] avg loss 0.00594902, throughput 12.9351K wps
Begin Testing...
[Epoch 36] train avg loss 0.0057684, test acc 0.7917, test avg loss 0.449714, throughput 12.9343K wps
[Epoch 37 Batch 30/173] avg loss 0.00566123, throughput 13.2076K wps
[Epoch 37 Batch 60/173] avg loss 0.00577354, throughput 12.7887K wps
[Epoch 37 Batch 90/173] avg loss 0.00567072, throughput 12.91K wps
[Epoch 37 Batch 120/173] avg loss 0.00606626, throughput 12.8377K wps
[Epoch 37 Batch 150/173] avg loss 0.00595673, throughput 12.9382K wps
Begin Testing...
[Epoch 37] train avg loss 0.00579637, test acc 0.7979, test avg loss 0.446948, throughput 12.9344K wps
[Epoch 38 Batch 30/173] avg loss 0.00531012, throughput 13.1848K wps
[Epoch 38 Batch 60/173] avg loss 0.00541825, throughput 12.8014K wps
[Epoch 38 Batch 90/173] avg loss 0.005489, throughput 12.9223K wps
[Epoch 38 Batch 120/173] avg loss 0.00563377, throughput 12.9166K wps
[Epoch 38 Batch 150/173] avg loss 0.00547105, throughput 12.8402K wps
Begin Testing...
[Epoch 38] train avg loss 0.00552028, test acc 0.7927, test avg loss 0.448604, throughput 12.9407K wps
[Epoch 39 Batch 30/173] avg loss 0.00538845, throughput 13.2534K wps
[Epoch 39 Batch 60/173] avg loss 0.00534334, throughput 12.8265K wps
[Epoch 39 Batch 90/173] avg loss 0.0053941, throughput 12.9425K wps
[Epoch 39 Batch 120/173] avg loss 0.0055148, throughput 12.7717K wps
[Epoch 39 Batch 150/173] avg loss 0.00555642, throughput 12.7826K wps
Begin Testing...
[Epoch 39] train avg loss 0.00547965, test acc 0.7885, test avg loss 0.447228, throughput 12.9026K wps
[Epoch 40 Batch 30/173] avg loss 0.00521254, throughput 13.2922K wps
[Epoch 40 Batch 60/173] avg loss 0.00523238, throughput 12.7967K wps
[Epoch 40 Batch 90/173] avg loss 0.00525331, throughput 12.8601K wps
[Epoch 40 Batch 120/173] avg loss 0.00547968, throughput 12.8021K wps
[Epoch 40 Batch 150/173] avg loss 0.00561915, throughput 12.798K wps
Begin Testing...
[Epoch 40] train avg loss 0.00537125, test acc 0.7937, test avg loss 0.445861, throughput 12.8919K wps
[Epoch 41 Batch 30/173] avg loss 0.00508013, throughput 13.2683K wps
[Epoch 41 Batch 60/173] avg loss 0.00538801, throughput 12.8028K wps
[Epoch 41 Batch 90/173] avg loss 0.00510374, throughput 12.8848K wps
[Epoch 41 Batch 120/173] avg loss 0.00512197, throughput 12.8788K wps
[Epoch 41 Batch 150/173] avg loss 0.00548251, throughput 12.8234K wps
Begin Testing...
[Epoch 41] train avg loss 0.00521101, test acc 0.7937, test avg loss 0.446787, throughput 12.9138K wps
[Epoch 42 Batch 30/173] avg loss 0.00504147, throughput 13.2263K wps
[Epoch 42 Batch 60/173] avg loss 0.00516582, throughput 12.7524K wps
[Epoch 42 Batch 90/173] avg loss 0.00543353, throughput 12.8109K wps
[Epoch 42 Batch 120/173] avg loss 0.00527924, throughput 12.8285K wps
[Epoch 42 Batch 150/173] avg loss 0.0051421, throughput 12.8422K wps
Begin Testing...
[Epoch 42] train avg loss 0.00518783, test acc 0.7937, test avg loss 0.448824, throughput 12.8881K wps
[Epoch 43 Batch 30/173] avg loss 0.00497428, throughput 13.156K wps
[Epoch 43 Batch 60/173] avg loss 0.0050843, throughput 12.7552K wps
[Epoch 43 Batch 90/173] avg loss 0.00484895, throughput 12.7479K wps
[Epoch 43 Batch 120/173] avg loss 0.00517638, throughput 12.7714K wps
[Epoch 43 Batch 150/173] avg loss 0.00500485, throughput 12.9137K wps
Begin Testing...
[Epoch 43] train avg loss 0.0050974, test acc 0.7885, test avg loss 0.450345, throughput 12.8717K wps
[Epoch 44 Batch 30/173] avg loss 0.00469794, throughput 13.1181K wps
[Epoch 44 Batch 60/173] avg loss 0.00484442, throughput 12.79K wps
[Epoch 44 Batch 90/173] avg loss 0.00538921, throughput 12.789K wps
[Epoch 44 Batch 120/173] avg loss 0.00489721, throughput 12.8619K wps
[Epoch 44 Batch 150/173] avg loss 0.00516881, throughput 12.7166K wps
Begin Testing...
[Epoch 44] train avg loss 0.00501591, test acc 0.7969, test avg loss 0.447967, throughput 12.8426K wps
[Epoch 45 Batch 30/173] avg loss 0.0046972, throughput 13.251K wps
[Epoch 45 Batch 60/173] avg loss 0.00479285, throughput 12.7606K wps
[Epoch 45 Batch 90/173] avg loss 0.00504613, throughput 12.8521K wps
[Epoch 45 Batch 120/173] avg loss 0.0049006, throughput 12.7681K wps
[Epoch 45 Batch 150/173] avg loss 0.00483314, throughput 12.7735K wps
Begin Testing...
[Epoch 45] train avg loss 0.00486259, test acc 0.7875, test avg loss 0.446909, throughput 12.8611K wps
[Epoch 46 Batch 30/173] avg loss 0.00472234, throughput 13.1819K wps
[Epoch 46 Batch 60/173] avg loss 0.00462471, throughput 12.7079K wps
[Epoch 46 Batch 90/173] avg loss 0.00454849, throughput 12.9421K wps
[Epoch 46 Batch 120/173] avg loss 0.00477669, throughput 12.8324K wps
[Epoch 46 Batch 150/173] avg loss 0.00494457, throughput 12.8316K wps
Begin Testing...
[Epoch 46] train avg loss 0.00475252, test acc 0.7990, test avg loss 0.440679, throughput 12.9065K wps
Observed Improvement.
Begin Testing...
[Epoch 47 Batch 30/173] avg loss 0.00462285, throughput 13.2418K wps
[Epoch 47 Batch 60/173] avg loss 0.00443963, throughput 12.6755K wps
[Epoch 47 Batch 90/173] avg loss 0.00477961, throughput 12.9303K wps
[Epoch 47 Batch 120/173] avg loss 0.00452972, throughput 12.8127K wps
[Epoch 47 Batch 150/173] avg loss 0.00449441, throughput 12.9013K wps
Begin Testing...
[Epoch 47] train avg loss 0.00463359, test acc 0.7917, test avg loss 0.444211, throughput 12.8925K wps
[Epoch 48 Batch 30/173] avg loss 0.00476412, throughput 13.2275K wps
[Epoch 48 Batch 60/173] avg loss 0.00503367, throughput 12.7869K wps
[Epoch 48 Batch 90/173] avg loss 0.00457754, throughput 12.9262K wps
[Epoch 48 Batch 120/173] avg loss 0.00454411, throughput 12.8548K wps
[Epoch 48 Batch 150/173] avg loss 0.00445868, throughput 12.846K wps
Begin Testing...
[Epoch 48] train avg loss 0.00467596, test acc 0.8000, test avg loss 0.448245, throughput 12.9344K wps
Observed Improvement.
Begin Testing...
[Epoch 49 Batch 30/173] avg loss 0.00431694, throughput 13.2236K wps
[Epoch 49 Batch 60/173] avg loss 0.00441361, throughput 12.7474K wps
[Epoch 49 Batch 90/173] avg loss 0.00444441, throughput 12.8675K wps
[Epoch 49 Batch 120/173] avg loss 0.00442272, throughput 12.7721K wps
[Epoch 49 Batch 150/173] avg loss 0.00454371, throughput 12.746K wps
Begin Testing...
[Epoch 49] train avg loss 0.00448037, test acc 0.7958, test avg loss 0.447919, throughput 12.8776K wps
[Epoch 50 Batch 30/173] avg loss 0.00442215, throughput 13.2298K wps
[Epoch 50 Batch 60/173] avg loss 0.00422184, throughput 12.7473K wps
[Epoch 50 Batch 90/173] avg loss 0.00444169, throughput 12.922K wps
[Epoch 50 Batch 120/173] avg loss 0.00436218, throughput 12.7692K wps
[Epoch 50 Batch 150/173] avg loss 0.00447274, throughput 12.8993K wps
Begin Testing...
[Epoch 50] train avg loss 0.00442797, test acc 0.8000, test avg loss 0.447016, throughput 12.9157K wps
Observed Improvement.
Begin Testing...
[Epoch 51 Batch 30/173] avg loss 0.00397622, throughput 13.2236K wps
[Epoch 51 Batch 60/173] avg loss 0.00438186, throughput 12.7969K wps
[Epoch 51 Batch 90/173] avg loss 0.00441899, throughput 12.9469K wps
[Epoch 51 Batch 120/173] avg loss 0.00456662, throughput 12.9138K wps
[Epoch 51 Batch 150/173] avg loss 0.00419298, throughput 12.9384K wps
Begin Testing...
[Epoch 51] train avg loss 0.0043607, test acc 0.8000, test avg loss 0.442535, throughput 12.9641K wps
Observed Improvement.
Begin Testing...
[Epoch 52 Batch 30/173] avg loss 0.00418443, throughput 13.3079K wps
[Epoch 52 Batch 60/173] avg loss 0.00432867, throughput 12.8465K wps
[Epoch 52 Batch 90/173] avg loss 0.00427495, throughput 12.9801K wps
[Epoch 52 Batch 120/173] avg loss 0.00382176, throughput 12.9186K wps
[Epoch 52 Batch 150/173] avg loss 0.0043219, throughput 12.929K wps
Begin Testing...
[Epoch 52] train avg loss 0.00421754, test acc 0.7979, test avg loss 0.452721, throughput 12.9904K wps
[Epoch 53 Batch 30/173] avg loss 0.00401491, throughput 13.2735K wps
[Epoch 53 Batch 60/173] avg loss 0.00402641, throughput 12.8006K wps
[Epoch 53 Batch 90/173] avg loss 0.00430501, throughput 12.9466K wps
[Epoch 53 Batch 120/173] avg loss 0.00411614, throughput 12.9623K wps
[Epoch 53 Batch 150/173] avg loss 0.00402696, throughput 12.9564K wps
Begin Testing...
[Epoch 53] train avg loss 0.00410072, test acc 0.7937, test avg loss 0.447603, throughput 12.9837K wps
[Epoch 54 Batch 30/173] avg loss 0.00421959, throughput 13.1712K wps
[Epoch 54 Batch 60/173] avg loss 0.00388437, throughput 12.857K wps
[Epoch 54 Batch 90/173] avg loss 0.00381085, throughput 12.9362K wps
[Epoch 54 Batch 120/173] avg loss 0.00397737, throughput 12.9406K wps
[Epoch 54 Batch 150/173] avg loss 0.00411999, throughput 12.8933K wps
Begin Testing...
[Epoch 54] train avg loss 0.00406817, test acc 0.8000, test avg loss 0.44355, throughput 12.9525K wps
Observed Improvement.
Begin Testing...
[Epoch 55 Batch 30/173] avg loss 0.0039123, throughput 13.2537K wps
[Epoch 55 Batch 60/173] avg loss 0.00393769, throughput 12.7589K wps
[Epoch 55 Batch 90/173] avg loss 0.00394302, throughput 12.9039K wps
[Epoch 55 Batch 120/173] avg loss 0.00397172, throughput 12.7784K wps
[Epoch 55 Batch 150/173] avg loss 0.00395897, throughput 12.9302K wps
Begin Testing...
[Epoch 55] train avg loss 0.0039091, test acc 0.7917, test avg loss 0.449196, throughput 12.9303K wps
[Epoch 56 Batch 30/173] avg loss 0.00387059, throughput 13.2238K wps
[Epoch 56 Batch 60/173] avg loss 0.00384171, throughput 12.8649K wps
[Epoch 56 Batch 90/173] avg loss 0.00398192, throughput 12.9656K wps
[Epoch 56 Batch 120/173] avg loss 0.00401026, throughput 12.974K wps
[Epoch 56 Batch 150/173] avg loss 0.00366688, throughput 12.9231K wps
Begin Testing...
[Epoch 56] train avg loss 0.00388837, test acc 0.7990, test avg loss 0.453669, throughput 12.9857K wps
[Epoch 57 Batch 30/173] avg loss 0.00376151, throughput 13.2209K wps
[Epoch 57 Batch 60/173] avg loss 0.00376851, throughput 12.8524K wps
[Epoch 57 Batch 90/173] avg loss 0.00377802, throughput 12.9025K wps
[Epoch 57 Batch 120/173] avg loss 0.00402671, throughput 12.7979K wps
[Epoch 57 Batch 150/173] avg loss 0.00379276, throughput 12.9652K wps
Begin Testing...
[Epoch 57] train avg loss 0.0038431, test acc 0.8000, test avg loss 0.45027, throughput 12.949K wps
Observed Improvement.
Begin Testing...
[Epoch 58 Batch 30/173] avg loss 0.00359553, throughput 13.2464K wps
[Epoch 58 Batch 60/173] avg loss 0.00361362, throughput 12.7362K wps
[Epoch 58 Batch 90/173] avg loss 0.00359387, throughput 12.9209K wps
[Epoch 58 Batch 120/173] avg loss 0.00363, throughput 12.8574K wps
[Epoch 58 Batch 150/173] avg loss 0.00376666, throughput 12.851K wps
Begin Testing...
[Epoch 58] train avg loss 0.00368884, test acc 0.7948, test avg loss 0.446812, throughput 12.9192K wps
[Epoch 59 Batch 30/173] avg loss 0.00352051, throughput 13.1957K wps
[Epoch 59 Batch 60/173] avg loss 0.00346926, throughput 12.7962K wps
[Epoch 59 Batch 90/173] avg loss 0.00373645, throughput 12.8929K wps
[Epoch 59 Batch 120/173] avg loss 0.00380735, throughput 12.9597K wps
[Epoch 59 Batch 150/173] avg loss 0.00372331, throughput 12.7539K wps
Begin Testing...
[Epoch 59] train avg loss 0.00364218, test acc 0.8031, test avg loss 0.449044, throughput 12.9107K wps
Observed Improvement.
Begin Testing...
Test loss 0.478815, test acc 0.7758
Total time cost 177.84s
9596 1066
[Epoch 0 Batch 30/173] avg loss 0.0149972, throughput 11.8551K wps
[Epoch 0 Batch 60/173] avg loss 0.0150056, throughput 12.7698K wps
[Epoch 0 Batch 90/173] avg loss 0.0146066, throughput 12.8156K wps
[Epoch 0 Batch 120/173] avg loss 0.0144537, throughput 12.8366K wps
[Epoch 0 Batch 150/173] avg loss 0.0143754, throughput 12.7816K wps
Begin Testing...
[Epoch 0] train avg loss 0.0146969, test acc 0.6073, test avg loss 0.661896, throughput 12.6402K wps
Observed Improvement.
Begin Testing...
[Epoch 1 Batch 30/173] avg loss 0.0139817, throughput 13.2709K wps
[Epoch 1 Batch 60/173] avg loss 0.0138876, throughput 12.8054K wps
[Epoch 1 Batch 90/173] avg loss 0.0135303, throughput 12.8454K wps
[Epoch 1 Batch 120/173] avg loss 0.0134583, throughput 12.8356K wps
[Epoch 1 Batch 150/173] avg loss 0.0134412, throughput 12.809K wps
Begin Testing...
[Epoch 1] train avg loss 0.0136866, test acc 0.6396, test avg loss 0.646863, throughput 12.8973K wps
Observed Improvement.
Begin Testing...
[Epoch 2 Batch 30/173] avg loss 0.0129399, throughput 13.254K wps
[Epoch 2 Batch 60/173] avg loss 0.01322, throughput 12.7803K wps
[Epoch 2 Batch 90/173] avg loss 0.0130659, throughput 12.775K wps
[Epoch 2 Batch 120/173] avg loss 0.0132672, throughput 12.7673K wps
[Epoch 2 Batch 150/173] avg loss 0.0132047, throughput 12.7817K wps
Begin Testing...
[Epoch 2] train avg loss 0.0131465, test acc 0.6625, test avg loss 0.639007, throughput 12.8593K wps
Observed Improvement.
Begin Testing...
[Epoch 3 Batch 30/173] avg loss 0.0129918, throughput 13.266K wps
[Epoch 3 Batch 60/173] avg loss 0.0126976, throughput 12.7689K wps
[Epoch 3 Batch 90/173] avg loss 0.01285, throughput 12.7894K wps
[Epoch 3 Batch 120/173] avg loss 0.0128297, throughput 12.7912K wps
[Epoch 3 Batch 150/173] avg loss 0.0127526, throughput 12.742K wps
Begin Testing...
[Epoch 3] train avg loss 0.0128091, test acc 0.6740, test avg loss 0.624293, throughput 12.8575K wps
Observed Improvement.
Begin Testing...
[Epoch 4 Batch 30/173] avg loss 0.0123252, throughput 13.279K wps
[Epoch 4 Batch 60/173] avg loss 0.0125446, throughput 12.7664K wps
[Epoch 4 Batch 90/173] avg loss 0.0125304, throughput 12.854K wps
[Epoch 4 Batch 120/173] avg loss 0.0123434, throughput 12.8034K wps
[Epoch 4 Batch 150/173] avg loss 0.0124823, throughput 12.8139K wps
Begin Testing...
[Epoch 4] train avg loss 0.012441, test acc 0.7042, test avg loss 0.613094, throughput 12.8908K wps
Observed Improvement.
Begin Testing...
[Epoch 5 Batch 30/173] avg loss 0.0122083, throughput 13.2749K wps
[Epoch 5 Batch 60/173] avg loss 0.012176, throughput 12.7981K wps
[Epoch 5 Batch 90/173] avg loss 0.0120177, throughput 12.7785K wps
[Epoch 5 Batch 120/173] avg loss 0.0120825, throughput 12.8019K wps
[Epoch 5 Batch 150/173] avg loss 0.0119788, throughput 12.9427K wps
Begin Testing...
[Epoch 5] train avg loss 0.0120681, test acc 0.7240, test avg loss 0.599184, throughput 12.9052K wps
Observed Improvement.
Begin Testing...
[Epoch 6 Batch 30/173] avg loss 0.0118856, throughput 13.2646K wps
[Epoch 6 Batch 60/173] avg loss 0.011471, throughput 12.8016K wps
[Epoch 6 Batch 90/173] avg loss 0.0117505, throughput 12.8186K wps
[Epoch 6 Batch 120/173] avg loss 0.0116654, throughput 12.7926K wps
[Epoch 6 Batch 150/173] avg loss 0.0117748, throughput 12.7979K wps
Begin Testing...
[Epoch 6] train avg loss 0.0117287, test acc 0.7094, test avg loss 0.587728, throughput 12.8834K wps
[Epoch 7 Batch 30/173] avg loss 0.0117136, throughput 13.199K wps
[Epoch 7 Batch 60/173] avg loss 0.0113475, throughput 12.8158K wps
[Epoch 7 Batch 90/173] avg loss 0.0116163, throughput 12.9237K wps
[Epoch 7 Batch 120/173] avg loss 0.0114219, throughput 12.8161K wps
[Epoch 7 Batch 150/173] avg loss 0.0112365, throughput 12.9451K wps
Begin Testing...
[Epoch 7] train avg loss 0.0114423, test acc 0.7458, test avg loss 0.571633, throughput 12.9201K wps
Observed Improvement.
Begin Testing...
[Epoch 8 Batch 30/173] avg loss 0.0112028, throughput 13.2419K wps
[Epoch 8 Batch 60/173] avg loss 0.0109757, throughput 12.7671K wps
[Epoch 8 Batch 90/173] avg loss 0.0110283, throughput 12.7448K wps
[Epoch 8 Batch 120/173] avg loss 0.0110749, throughput 12.8262K wps
[Epoch 8 Batch 150/173] avg loss 0.0109166, throughput 12.7854K wps
Begin Testing...
[Epoch 8] train avg loss 0.0110664, test acc 0.7500, test avg loss 0.55914, throughput 12.8747K wps
Observed Improvement.
Begin Testing...
[Epoch 9 Batch 30/173] avg loss 0.0104424, throughput 13.1912K wps
[Epoch 9 Batch 60/173] avg loss 0.010581, throughput 12.8146K wps
[Epoch 9 Batch 90/173] avg loss 0.0106991, throughput 12.9428K wps
[Epoch 9 Batch 120/173] avg loss 0.010812, throughput 12.8891K wps
[Epoch 9 Batch 150/173] avg loss 0.0107895, throughput 12.7784K wps
Begin Testing...
[Epoch 9] train avg loss 0.0106851, test acc 0.7583, test avg loss 0.545399, throughput 12.906K wps
Observed Improvement.
Begin Testing...
[Epoch 10 Batch 30/173] avg loss 0.0104519, throughput 13.1785K wps
[Epoch 10 Batch 60/173] avg loss 0.0103744, throughput 12.8138K wps
[Epoch 10 Batch 90/173] avg loss 0.0103966, throughput 12.78K wps
[Epoch 10 Batch 120/173] avg loss 0.0102337, throughput 12.8077K wps
[Epoch 10 Batch 150/173] avg loss 0.0104124, throughput 12.7914K wps
Begin Testing...
[Epoch 10] train avg loss 0.0103846, test acc 0.7510, test avg loss 0.533239, throughput 12.876K wps
[Epoch 11 Batch 30/173] avg loss 0.0102381, throughput 13.2542K wps
[Epoch 11 Batch 60/173] avg loss 0.00982059, throughput 12.7864K wps
[Epoch 11 Batch 90/173] avg loss 0.0103894, throughput 12.8727K wps
[Epoch 11 Batch 120/173] avg loss 0.0101027, throughput 12.8799K wps
[Epoch 11 Batch 150/173] avg loss 0.0100504, throughput 12.7882K wps
Begin Testing...
[Epoch 11] train avg loss 0.0101296, test acc 0.7615, test avg loss 0.521436, throughput 12.8998K wps
Observed Improvement.
Begin Testing...
[Epoch 12 Batch 30/173] avg loss 0.0095221, throughput 13.1929K wps
[Epoch 12 Batch 60/173] avg loss 0.00971827, throughput 12.754K wps
[Epoch 12 Batch 90/173] avg loss 0.00959862, throughput 12.8663K wps
[Epoch 12 Batch 120/173] avg loss 0.00989639, throughput 12.8933K wps
[Epoch 12 Batch 150/173] avg loss 0.00974408, throughput 12.7699K wps
Begin Testing...
[Epoch 12] train avg loss 0.00969489, test acc 0.7677, test avg loss 0.512695, throughput 12.8787K wps
Observed Improvement.
Begin Testing...
[Epoch 13 Batch 30/173] avg loss 0.00942398, throughput 13.2051K wps
[Epoch 13 Batch 60/173] avg loss 0.00973751, throughput 12.903K wps
[Epoch 13 Batch 90/173] avg loss 0.00938134, throughput 12.9293K wps
[Epoch 13 Batch 120/173] avg loss 0.00948338, throughput 12.8223K wps
[Epoch 13 Batch 150/173] avg loss 0.00932546, throughput 12.8708K wps
Begin Testing...
[Epoch 13] train avg loss 0.00948751, test acc 0.7625, test avg loss 0.503199, throughput 12.9462K wps
[Epoch 14 Batch 30/173] avg loss 0.00930472, throughput 13.1556K wps
[Epoch 14 Batch 60/173] avg loss 0.00941448, throughput 12.8088K wps
[Epoch 14 Batch 90/173] avg loss 0.00927721, throughput 12.9474K wps
[Epoch 14 Batch 120/173] avg loss 0.00929613, throughput 12.9124K wps
[Epoch 14 Batch 150/173] avg loss 0.0091535, throughput 12.8098K wps
Begin Testing...
[Epoch 14] train avg loss 0.00924522, test acc 0.7646, test avg loss 0.494787, throughput 12.9105K wps
[Epoch 15 Batch 30/173] avg loss 0.00942151, throughput 13.206K wps
[Epoch 15 Batch 60/173] avg loss 0.0091348, throughput 12.7784K wps
[Epoch 15 Batch 90/173] avg loss 0.00869807, throughput 12.9097K wps
[Epoch 15 Batch 120/173] avg loss 0.00894048, throughput 12.9297K wps
[Epoch 15 Batch 150/173] avg loss 0.00890627, throughput 12.8612K wps
Begin Testing...
[Epoch 15] train avg loss 0.0089968, test acc 0.7688, test avg loss 0.485374, throughput 12.9246K wps
Observed Improvement.
Begin Testing...
[Epoch 16 Batch 30/173] avg loss 0.0086673, throughput 13.246K wps
[Epoch 16 Batch 60/173] avg loss 0.00882975, throughput 12.8594K wps
[Epoch 16 Batch 90/173] avg loss 0.00857426, throughput 12.8906K wps
[Epoch 16 Batch 120/173] avg loss 0.00885554, throughput 12.789K wps
[Epoch 16 Batch 150/173] avg loss 0.00887508, throughput 12.7788K wps
Begin Testing...
[Epoch 16] train avg loss 0.00876719, test acc 0.7688, test avg loss 0.480109, throughput 12.9033K wps
Observed Improvement.
Begin Testing...
[Epoch 17 Batch 30/173] avg loss 0.00872424, throughput 13.3345K wps
[Epoch 17 Batch 60/173] avg loss 0.00861777, throughput 12.7864K wps
[Epoch 17 Batch 90/173] avg loss 0.00848204, throughput 12.9476K wps
[Epoch 17 Batch 120/173] avg loss 0.00874292, throughput 12.9371K wps
[Epoch 17 Batch 150/173] avg loss 0.00859848, throughput 12.8175K wps
Begin Testing...
[Epoch 17] train avg loss 0.00857399, test acc 0.7552, test avg loss 0.483333, throughput 12.9647K wps
[Epoch 18 Batch 30/173] avg loss 0.00828359, throughput 13.2484K wps
[Epoch 18 Batch 60/173] avg loss 0.00809098, throughput 12.7188K wps
[Epoch 18 Batch 90/173] avg loss 0.00856199, throughput 12.8567K wps
[Epoch 18 Batch 120/173] avg loss 0.00840727, throughput 12.8016K wps
[Epoch 18 Batch 150/173] avg loss 0.00862674, throughput 12.9313K wps
Begin Testing...
[Epoch 18] train avg loss 0.00842516, test acc 0.7719, test avg loss 0.477991, throughput 12.8945K wps
Observed Improvement.
Begin Testing...
[Epoch 19 Batch 30/173] avg loss 0.00829075, throughput 13.1355K wps
[Epoch 19 Batch 60/173] avg loss 0.00821984, throughput 12.8152K wps
[Epoch 19 Batch 90/173] avg loss 0.00822049, throughput 12.7776K wps
[Epoch 19 Batch 120/173] avg loss 0.00861467, throughput 12.7905K wps
[Epoch 19 Batch 150/173] avg loss 0.00800706, throughput 12.88K wps
Begin Testing...
[Epoch 19] train avg loss 0.00824135, test acc 0.7771, test avg loss 0.464569, throughput 12.8753K wps
Observed Improvement.
Begin Testing...
[Epoch 20 Batch 30/173] avg loss 0.0081584, throughput 13.2653K wps
[Epoch 20 Batch 60/173] avg loss 0.00815207, throughput 12.7218K wps
[Epoch 20 Batch 90/173] avg loss 0.00784537, throughput 12.844K wps
[Epoch 20 Batch 120/173] avg loss 0.00790399, throughput 12.8006K wps
[Epoch 20 Batch 150/173] avg loss 0.00812737, throughput 12.7618K wps
Begin Testing...
[Epoch 20] train avg loss 0.00809589, test acc 0.7750, test avg loss 0.468458, throughput 12.8947K wps
[Epoch 21 Batch 30/173] avg loss 0.00779702, throughput 13.2364K wps
[Epoch 21 Batch 60/173] avg loss 0.00781294, throughput 12.8006K wps
[Epoch 21 Batch 90/173] avg loss 0.00777838, throughput 12.8314K wps
[Epoch 21 Batch 120/173] avg loss 0.00791734, throughput 12.7605K wps
[Epoch 21 Batch 150/173] avg loss 0.00799753, throughput 12.9297K wps
Begin Testing...
[Epoch 21] train avg loss 0.00789496, test acc 0.7760, test avg loss 0.458149, throughput 12.8967K wps
[Epoch 22 Batch 30/173] avg loss 0.00759846, throughput 13.1433K wps
[Epoch 22 Batch 60/173] avg loss 0.0078858, throughput 12.7584K wps
[Epoch 22 Batch 90/173] avg loss 0.00774612, throughput 12.8742K wps
[Epoch 22 Batch 120/173] avg loss 0.00778187, throughput 12.7288K wps
[Epoch 22 Batch 150/173] avg loss 0.0076115, throughput 12.7855K wps
Begin Testing...
[Epoch 22] train avg loss 0.00775756, test acc 0.7719, test avg loss 0.456439, throughput 12.8701K wps
[Epoch 23 Batch 30/173] avg loss 0.00736455, throughput 13.2749K wps
[Epoch 23 Batch 60/173] avg loss 0.00721671, throughput 12.6801K wps
[Epoch 23 Batch 90/173] avg loss 0.00793493, throughput 12.852K wps
[Epoch 23 Batch 120/173] avg loss 0.00790602, throughput 12.8481K wps
[Epoch 23 Batch 150/173] avg loss 0.00788428, throughput 12.9328K wps
Begin Testing...
[Epoch 23] train avg loss 0.00768659, test acc 0.7760, test avg loss 0.453981, throughput 12.9098K wps
[Epoch 24 Batch 30/173] avg loss 0.00751597, throughput 13.2083K wps
[Epoch 24 Batch 60/173] avg loss 0.00785984, throughput 12.7645K wps
[Epoch 24 Batch 90/173] avg loss 0.00730193, throughput 12.9322K wps
[Epoch 24 Batch 120/173] avg loss 0.00695837, throughput 12.9115K wps
[Epoch 24 Batch 150/173] avg loss 0.00773893, throughput 12.8982K wps
Begin Testing...
[Epoch 24] train avg loss 0.00744972, test acc 0.7812, test avg loss 0.452503, throughput 12.9514K wps
Observed Improvement.
Begin Testing...
[Epoch 25 Batch 30/173] avg loss 0.00723118, throughput 13.2422K wps
[Epoch 25 Batch 60/173] avg loss 0.00753102, throughput 12.7608K wps
[Epoch 25 Batch 90/173] avg loss 0.00723057, throughput 12.9121K wps
[Epoch 25 Batch 120/173] avg loss 0.00724753, throughput 12.9529K wps
[Epoch 25 Batch 150/173] avg loss 0.007408, throughput 12.7421K wps
Begin Testing...
[Epoch 25] train avg loss 0.0072588, test acc 0.7844, test avg loss 0.45135, throughput 12.9212K wps
Observed Improvement.
Begin Testing...
[Epoch 26 Batch 30/173] avg loss 0.00735376, throughput 13.2171K wps
[Epoch 26 Batch 60/173] avg loss 0.00707718, throughput 12.8987K wps
[Epoch 26 Batch 90/173] avg loss 0.00699906, throughput 12.8847K wps
[Epoch 26 Batch 120/173] avg loss 0.00713425, throughput 12.79K wps
[Epoch 26 Batch 150/173] avg loss 0.007155, throughput 12.8598K wps
Begin Testing...
[Epoch 26] train avg loss 0.00716671, test acc 0.7896, test avg loss 0.447353, throughput 12.9401K wps
Observed Improvement.
Begin Testing...
[Epoch 27 Batch 30/173] avg loss 0.00689025, throughput 13.2674K wps
[Epoch 27 Batch 60/173] avg loss 0.00707423, throughput 12.776K wps
[Epoch 27 Batch 90/173] avg loss 0.00677237, throughput 12.9251K wps
[Epoch 27 Batch 120/173] avg loss 0.00703586, throughput 12.859K wps
[Epoch 27 Batch 150/173] avg loss 0.00715387, throughput 12.8322K wps
Begin Testing...
[Epoch 27] train avg loss 0.00704685, test acc 0.7792, test avg loss 0.444587, throughput 12.9422K wps
[Epoch 28 Batch 30/173] avg loss 0.00720382, throughput 13.2353K wps
[Epoch 28 Batch 60/173] avg loss 0.0068496, throughput 12.7595K wps
[Epoch 28 Batch 90/173] avg loss 0.00673173, throughput 12.9838K wps
[Epoch 28 Batch 120/173] avg loss 0.00708668, throughput 12.9598K wps
[Epoch 28 Batch 150/173] avg loss 0.00684273, throughput 12.9005K wps
Begin Testing...
[Epoch 28] train avg loss 0.00698105, test acc 0.7844, test avg loss 0.444811, throughput 12.9497K wps
[Epoch 29 Batch 30/173] avg loss 0.00684844, throughput 13.2766K wps
[Epoch 29 Batch 60/173] avg loss 0.00660666, throughput 12.8071K wps
[Epoch 29 Batch 90/173] avg loss 0.00670538, throughput 12.834K wps
[Epoch 29 Batch 120/173] avg loss 0.0067726, throughput 12.935K wps
[Epoch 29 Batch 150/173] avg loss 0.00695167, throughput 12.9497K wps
Begin Testing...
[Epoch 29] train avg loss 0.00676137, test acc 0.7896, test avg loss 0.441867, throughput 12.9564K wps
Observed Improvement.
Begin Testing...
[Epoch 30 Batch 30/173] avg loss 0.00672372, throughput 13.1704K wps
[Epoch 30 Batch 60/173] avg loss 0.00669215, throughput 12.8206K wps
[Epoch 30 Batch 90/173] avg loss 0.00660241, throughput 12.9422K wps
[Epoch 30 Batch 120/173] avg loss 0.00648701, throughput 12.951K wps
[Epoch 30 Batch 150/173] avg loss 0.00677484, throughput 12.9388K wps
Begin Testing...
[Epoch 30] train avg loss 0.00662659, test acc 0.7875, test avg loss 0.43785, throughput 12.9614K wps
[Epoch 31 Batch 30/173] avg loss 0.00621488, throughput 13.1572K wps
[Epoch 31 Batch 60/173] avg loss 0.00670942, throughput 12.7795K wps
[Epoch 31 Batch 90/173] avg loss 0.0064463, throughput 12.813K wps
[Epoch 31 Batch 120/173] avg loss 0.00635012, throughput 12.9136K wps
[Epoch 31 Batch 150/173] avg loss 0.00654056, throughput 12.9063K wps
Begin Testing...
[Epoch 31] train avg loss 0.00653399, test acc 0.7885, test avg loss 0.439228, throughput 12.917K wps
[Epoch 32 Batch 30/173] avg loss 0.00638736, throughput 13.2567K wps
[Epoch 32 Batch 60/173] avg loss 0.00617443, throughput 12.8378K wps
[Epoch 32 Batch 90/173] avg loss 0.00648698, throughput 12.9178K wps
[Epoch 32 Batch 120/173] avg loss 0.00656556, throughput 12.9291K wps
[Epoch 32 Batch 150/173] avg loss 0.00671644, throughput 12.9698K wps
Begin Testing...
[Epoch 32] train avg loss 0.00643283, test acc 0.7844, test avg loss 0.438368, throughput 12.9638K wps
[Epoch 33 Batch 30/173] avg loss 0.00634586, throughput 13.2627K wps
[Epoch 33 Batch 60/173] avg loss 0.00613299, throughput 12.8091K wps
[Epoch 33 Batch 90/173] avg loss 0.00642516, throughput 12.806K wps
[Epoch 33 Batch 120/173] avg loss 0.00614105, throughput 12.9187K wps
[Epoch 33 Batch 150/173] avg loss 0.00642379, throughput 12.9472K wps
Begin Testing...
[Epoch 33] train avg loss 0.0062961, test acc 0.7896, test avg loss 0.440452, throughput 12.9388K wps
Observed Improvement.
Begin Testing...
[Epoch 34 Batch 30/173] avg loss 0.00602834, throughput 13.2405K wps
[Epoch 34 Batch 60/173] avg loss 0.00636327, throughput 12.7937K wps
[Epoch 34 Batch 90/173] avg loss 0.00646457, throughput 12.7996K wps
[Epoch 34 Batch 120/173] avg loss 0.00573301, throughput 12.8478K wps
[Epoch 34 Batch 150/173] avg loss 0.00648497, throughput 12.9414K wps
Begin Testing...
[Epoch 34] train avg loss 0.00618662, test acc 0.7896, test avg loss 0.436659, throughput 12.9067K wps
Observed Improvement.
Begin Testing...
[Epoch 35 Batch 30/173] avg loss 0.00586867, throughput 13.3048K wps
[Epoch 35 Batch 60/173] avg loss 0.00602855, throughput 12.7949K wps
[Epoch 35 Batch 90/173] avg loss 0.00622632, throughput 12.7817K wps
[Epoch 35 Batch 120/173] avg loss 0.00613479, throughput 12.7639K wps
[Epoch 35 Batch 150/173] avg loss 0.00610751, throughput 12.7767K wps
Begin Testing...
[Epoch 35] train avg loss 0.00605913, test acc 0.7927, test avg loss 0.435755, throughput 12.8952K wps
Observed Improvement.
Begin Testing...
[Epoch 36 Batch 30/173] avg loss 0.0058279, throughput 13.1655K wps
[Epoch 36 Batch 60/173] avg loss 0.00615003, throughput 12.7453K wps
[Epoch 36 Batch 90/173] avg loss 0.00593855, throughput 12.886K wps
[Epoch 36 Batch 120/173] avg loss 0.00564858, throughput 12.8713K wps
[Epoch 36 Batch 150/173] avg loss 0.00618423, throughput 12.9308K wps
Begin Testing...
[Epoch 36] train avg loss 0.00591853, test acc 0.7875, test avg loss 0.436197, throughput 12.9158K wps
[Epoch 37 Batch 30/173] avg loss 0.00602233, throughput 13.2012K wps
[Epoch 37 Batch 60/173] avg loss 0.00559835, throughput 12.7774K wps
[Epoch 37 Batch 90/173] avg loss 0.00585535, throughput 12.8169K wps
[Epoch 37 Batch 120/173] avg loss 0.00576977, throughput 12.8187K wps
[Epoch 37 Batch 150/173] avg loss 0.00611874, throughput 12.8124K wps
Begin Testing...
[Epoch 37] train avg loss 0.00591365, test acc 0.7865, test avg loss 0.435812, throughput 12.8668K wps
[Epoch 38 Batch 30/173] avg loss 0.0057158, throughput 13.2267K wps
[Epoch 38 Batch 60/173] avg loss 0.00548875, throughput 12.7878K wps
[Epoch 38 Batch 90/173] avg loss 0.00594643, throughput 12.8539K wps
[Epoch 38 Batch 120/173] avg loss 0.00537936, throughput 12.7745K wps
[Epoch 38 Batch 150/173] avg loss 0.00568258, throughput 12.7711K wps
Begin Testing...
[Epoch 38] train avg loss 0.00566225, test acc 0.7854, test avg loss 0.439299, throughput 12.8765K wps
[Epoch 39 Batch 30/173] avg loss 0.00591243, throughput 13.2471K wps
[Epoch 39 Batch 60/173] avg loss 0.00551483, throughput 12.7856K wps
[Epoch 39 Batch 90/173] avg loss 0.00556999, throughput 12.8233K wps
[Epoch 39 Batch 120/173] avg loss 0.00560272, throughput 12.7773K wps
[Epoch 39 Batch 150/173] avg loss 0.0055498, throughput 12.9393K wps
Begin Testing...
[Epoch 39] train avg loss 0.00561699, test acc 0.7823, test avg loss 0.439395, throughput 12.9098K wps
[Epoch 40 Batch 30/173] avg loss 0.00550903, throughput 13.2995K wps
[Epoch 40 Batch 60/173] avg loss 0.005518, throughput 12.7912K wps
[Epoch 40 Batch 90/173] avg loss 0.00568395, throughput 12.9541K wps
[Epoch 40 Batch 120/173] avg