python cifar10_main.py [options]
Options --data_dir : path to data directory where dataset is stored --model_dir : path where TF model will be stored --resnet_size : size of resnet --train_epochs : number of training epochs --epochs_per_eval : how many epochs between each evaluation --batch_size : batch size --data_format : 'channels_first' / 'channels_last' - 'channels_first' for GPU performance boost
Command to run:
python python cifar10_main.py --model_dir /tmp/cifar10_resnet14_tanh_50e --data_format channels_first
Compressing data & store in local directory:
cd /tmp && tar -czvf ~/models/official/resnet/cifar10_resnet14_tanh_50e.tar.gz cifar10_resnet14_tanh_50e/ && cd -
Copy compressed package from remote and decompress:
scp scarlettguo@cs.toronto.edu:/h/285/scarlettguo/models/official/resnet/cifar10_resnet14_tanh_50e.tar.gz && tar -xzvf cifar10_resnet14_tanh_50e.tar.gz
Visualizing data on TensorBoard:
tensorboard --logdir cifar10_resnet14_tanh_50e