Checkpoint folder

Could you please remove the “checkpoint folder” created in the previous training at the beginning of the next training? In the current environment, the “checkpoint files” of the previous training and the next training are put in one “checkpoint folder”, and it is difficult to handle them. Or, to prevent that from happening, I would like you to create a “checkpoint folder” with another name for each training.

Hi @DDPG7 ,

You can use the experiment argument to set a different name for each of your runs. This should generate a checkpoint folder using the value specified in experiment.

1 Like


I have been using it since last week. I am doing parallel processing with multiple parameter sets, but that is