This example downloads the cifar10 dataset using torchvision. A simple CNN is trained using Adam to about 71% accuracy on the test set.
Train a model using default parameters:
python cifar10_example.py
Resume training using a checkpoint:
python cifar10_example.py --resume --path_to_checkpoint model_checkpoint.ckpt