Skip to content

Latest commit

 

History

History
52 lines (45 loc) · 952 Bytes

README.md

File metadata and controls

52 lines (45 loc) · 952 Bytes
Raw

Training convolution network on CIFAR100 dataset

./samples/mnist/run_cifar100_train.sh

Possible output:

Epoch: 22
Loss: 2.39293
Loss: 2.37879
Loss: 2.37729
Loss: 2.38294
Loss: 2.38229
Loss: 2.38789
Loss: 2.38709
Loss: 2.38916
Loss: 2.38848
Loss: 2.38968
Loss: 2.39475
Loss: 2.39076
Loss: 2.38803
Loss: 2.38514
Loss: 2.38353
Loss: 2.38105
Accuracy: 0.344551
Model saved: build/cifar100_model.bin
  • Run test:
./samples/mnist/run_cifar100_test.sh

Output:

Accuracy: 0.344551

You can also run samples using binary target directly:

  • Run train:
./bin/sample_cifar100_train train <path-to-deepworks>/datasets/CIFAR100 <batch_size> <num_epochs> <dump-frequency> <path-to-dump>
  • Run test:
./bin/sample_cifar100_train test <path-to-deepworks>/datasets/CIFAR100 <batch_size> <path-to-model>