FAST-CIFAR10-IYO
Fast asynchronous training (time in evaluation is saved) with Iyo on the CIFAR10 dataset. 248
Iyo 19.10
==> Loading training data.. ==> Building training model..
0/10 | Lr: 1.000E-01 | LU: None(None) | Loss: INF(None) | Skipped: - | Acc: 0.000(None): 100%|████████████| 79/79 [00:17<00:00, 4.52it/s] 1/10 | Lr: 1.000E-01 | LU: 0(None) | Loss: 3.00E-03(INF) | Skipped: 0 | Acc: 31.430(0.000): 100%|████████████| 79/79 [00:11<00:00, 6.81it/s] 2/10 | Lr: 1.000E-01 | LU: 1(None) | Loss: 2.51E-03(INF) | Skipped: 0 | Acc: 53.554(0.000): 100%|████████████| 79/79 [00:13<00:00, 6.07it/s] 3/10 | Lr: 1.000E-01 | LU: 2(1) | Loss: 2.21E-03(1.18E-02) | Skipped: 0 | Acc: 63.400(58.400): 100%|████████████| 79/79 [00:13<00:00, 5.98it/s] 4/10 | Lr: 1.000E-01 | LU: 3(2) | Loss: 1.99E-03(1.04E-02) | Skipped: 0 | Acc: 69.571(63.800): 100%|████████████| 79/79 [00:13<00:00, 5.82it/s] 5/10 | Lr: 1.000E-01 | LU: 4(3) | Loss: 1.82E-03(9.12E-03) | Skipped: - | Acc: 74.268(68.890): 100%|████████████| 79/79 [00:13<00:00, 5.74it/s] 6/10 | Lr: 1.000E-01 | LU: 5(4) | Loss: 1.69E-03(7.22E-03) | Skipped: 0 | Acc: 77.512(74.750): 100%|████████████| 79/79 [00:13<00:00, 5.73it/s] 7/10 | Lr: 1.000E-01 | LU: 6(5) | Loss: 1.58E-03(7.03E-03) | Skipped: 0 | Acc: 79.802(75.500): 100%|████████████| 79/79 [00:14<00:00, 5.61it/s] 8/10 | Lr: 1.000E-01 | LU: 7(6) | Loss: 1.49E-03(6.69E-03) | Skipped: 0 | Acc: 81.390(76.770): 100%|████████████| 79/79 [00:13<00:00, 5.90it/s] 9/10 | Lr: 1.000E-01 | LU: 8(6) | Loss: 1.41E-03(6.69E-03) | Skipped: 0 | Acc: 82.826(76.770): 100%|████████████| 79/79 [00:14<00:00, 5.44it/s]
Total over 10 epochs: 0:05:09.459543
==> Preparing data.. ==> Building model..
[==== 391/391 ==========>] Step: 1s87ms | Tot: 36s626ms | Loss: 2.269 | Acc: 28.664% [==== 100/100 ==========>] Step: 43ms | Tot: 4s94ms | Loss: 1.610 | Acc: 40.120% Saving.. Epoch: 1 [==== 391/391 ====>] Step: 118ms | Tot: 35s773ms | Loss: 1.579 | Acc: 42.098% [==== 100/100 ====>] Step: 43ms | Tot: 4s47ms | Loss: 1.401 | Acc: 49.180% Saving.. Epoch: 2 [==== 391/391 ====>] Step: 89ms | Tot: 35s796ms | Loss: 1.397 | Acc: 49.568% [==== 100/100 ====>] Step: 47ms | Tot: 4s45ms | Loss: 1.249 | Acc: 55.580% Saving.. Epoch: 3 [==== 391/391 ====>] Step: 94ms | Tot: 35s29ms | Loss: 1.244 | Acc: 55.548% [==== 100/100 ====>] Step: 43ms | Tot: 4s135ms | Loss: 1.139 | Acc: 58.960% Saving.. Epoch: 4 [==== 391/391 ====>] Step: 101ms | Tot: 35s723ms | Loss: 1.148 | Acc: 59.194% [==== 100/100 ====>] Step: 40ms | Tot: 4s91ms | Loss: 1.034 | Acc: 63.060% Saving.. Epoch: 5 [==== 391/391 ====>] Step: 101ms | Tot: 35s650ms | Loss: 1.084 | Acc: 61.418% [==== 100/100 ====>] Step: 41ms | Tot: 4s67ms | Loss: 1.050 | Acc: 63.380% Saving.. Epoch: 6 [==== 391/391 ====>] Step: 99ms | Tot: 35s613ms | Loss: 1.013 | Acc: 64.106% [==== 100/100 ====>] Step: 36ms | Tot: 4s69ms | Loss: 0.940 | Acc: 66.960% Saving.. Epoch: 7 [==== 391/391 ====>] Step: 97ms | Tot: 35s583ms | Loss: 0.948 | Acc: 66.568% [==== 100/100 ====>] Step: 41ms | Tot: 4s65ms | Loss: 0.925 | Acc: 67.540% Saving.. Epoch: 8 [==== 391/391 ====>] Step: 93ms | Tot: 35s508ms | Loss: 0.909 | Acc: 68.308% [==== 100/100 ====>] Step: 38ms | Tot: 4s11ms | Loss: 0.890 | Acc: 69.220% Saving.. Epoch: 9 [==== 391/391 ====>] Step: 106ms | Tot: 35s602ms | Loss: 0.886 | Acc: 69.102% [==== 100/100 ====>] Step: 36ms | Tot: 4s98ms | Loss: 0.920 | Acc: 67.950% ```
Total over 10 epochs: 0:06:46.550089
conda install -c ninedwlab iyo
Start the training with:
python main.py --arch MobileNetV2 --lr=0.01 --epochs 350
Resume the training with:
python main.py --resume --arch MobileNetV2 --lr=0.01 --epochs 350
Model | Acc. |
---|---|
VGG16 | 92.64% |
ResNet18 | 93.02% |
ResNet50 | 93.62% |
ResNet101 | 93.75% |
MobileNetV2 | 94.43% |
ResNeXt29(32x4d) | 94.73% |
ResNeXt29(2x64d) | 94.82% |
DenseNet121 | 95.04% |
PreActResNet18 | 95.11% |
DPN92 | 95.16% |
For any question please contact me at j.cadic@protonmail.ch
还没有评论,说两句吧!
热门资源
Keras-ResNeXt
Keras ResNeXt Implementation of ResNeXt models...
seetafaceJNI
项目介绍 基于中科院seetaface2进行封装的JAVA...
spark-corenlp
This package wraps Stanford CoreNLP annotators ...
capsnet-with-caps...
CapsNet with capsule-wise convolution Project ...
inferno-boilerplate
This is a very basic boilerplate example for pe...
智能在线
400-630-6780
聆听.建议反馈
E-mail: support@tusaishared.com