`train_autoencoder`

Trains (or validates etc.) an AutoEncoder model against images using the PyTorch Lightning Command Line Interface.

Please see cnn.AutoEncoder for details of the auto-encoder architecture.

Input Arguments

Please see PyTorch Lightning CLI for details on how to use the command line interface.

The help page prints all options:

python -m anchor_python_training.train_autoencoder -h

As a suggestion, first create a configuration file by saving the contents to e.g. config.yaml:

python -m anchor_python_training.train_autoencoder fit --print_config

and then use these configuration file to perform the training:

python -m anchor_python_training.train_autoencoder fit --config config.yaml

The outputted model is saved incrementally via checkpoints to a directory lightning_logs in the working directory (unless otherwise configured).

The predict subcommand is not currently recommended for productive use.