image_classification

Image Classification Training Recipe

Recipe for training image classification models from TorchVision.

Training

Launch Jobs without TorchX

• We can easily launch a job to train a ResNet18 model on the CIFAR10 dataset with the following command:

  python torchrecipes/vision/image_classification/main.py

• Config overrides allow us to swap out different parts of the training job. For example, the following command launches a job to train a ResNet50 model on GPUs:

  python torchrecipes/vision/image_classification/main.py --config-name default_config module/model=resnet50 trainer=gpu

Launch Jobs with TorchX

• We often use TorchX to launch training jobs across different environments. You can install TorchX with

  pip install torchx

• Training jobs can then be launched with the following commands:

  torchx run --scheduler local_cwd utils.python --script torchrecipes/vision/image_classification/main.py