Closed chadcwilliams closed 1 year ago
imo I would set path_checkpoint to a default value e.g. trained_ae/checkpoint.pt In that manner you can continue training rapidly with less effort since the last training will be saved as checkpoint.pt anyways. Then you only have to set load_checkpoint without any filepath
I didn't realize it had that functionality. I like it this way, then! Closing to leave as is.
To continue training a model, you need to provide the filename to
path_checkpoint
but you also need to setload_checkpoint
to true. I think it would be best to keep path_checkpoint in that if a path is provided then it runs but if not then it doesn't.