Train CIFAR-10 and CIFAR-100 with PyTorch

Train deep networks with PyTorch on the CIFAR-10 and CIFAR-100 datasets.

Usage

Train ResNet56 on CIFAR-10

• with batch size 128
• for182 epochs
• with initial learning rate 0.1
• and a piecewise constant learning rate decay function
• with a decay factor of 0.1 (default) at epochs 91 and 136
• using the first two GPUs
• storing 10 state checkpoints
• and printing a progress bar

# setup options
MODEL=resnet56
BATCH_SIZE=128
NUM_EPOCHS=182
NUM_CKPTS=10
LR=0.1
DECAY_POLICY=pconst
LR_MILESTONES="91 136"
export CUDA_VISIBLE_DEVICES=0,1

# run 
SCRIPT=main.py

python $SCRIPT \
  --model ${MODEL} \
  --batch_size ${BATCH_SIZE} \
  --num_epochs ${NUM_EPOCHS} \
  --num_ckpts ${NUM_CKPTS} \
  --progress_bar \
  --lr ${LR} \
  --lr_decay_policy ${DECAY_POLICY} \
  --lr_milestones ${LR_MILESTONES}

==> Preparing data..
Files already downloaded and verified
Files already downloaded and verified
==> Building resnet56 model..

Epoch: 0
 [==>........................... 19/391 ..............................]  Step: 1s392ms | Tot: 24s295ms | lr: 1.000e-01 | Loss: 2.173 | Acc: 17.393% (423/2432)

Accuracy (as reported by @kuangliu)

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%