/black

@vfdev-5 , could you please help review this PR?

Thanks in advance.

/black

The current error message is ambiguous and does not really help the user debug without a dive in the MONAI codebase. I would change the error message to be more descriptive from:

raise ValueError("current max epochs setting is smaller than epoch count in the checkpoint.")

to

raise ValueError(f"The epoch count found in the checkpoint ({engine.state.epoch}) is larger than the engine.state.max_epochs of the constructed engine prior loading the checkpoint ({prior_max_epochs}). If you want to further train this checkpoint please construct the trainer with a max_epochs bigger than the epoch count of the checkpoint and bigger or equal with its max_epochs. If you want to use it for inference please use the same max_epochs you used during training.")

It might be too stuffy, please have a go at trimming it down if you feel it needs it.

raise ValueError(f"The epoch count found in the checkpoint ({engine.state.epoch}) is larger than the engine.state.max_epochs of the constructed engine prior loading the checkpoint ({prior_max_epochs}). If you want to further train this checkpoint please construct the trainer with a max_epochs bigger than the epoch count of the checkpoint and bigger or equal with its max_epochs. If you want to use it for inference please use the same max_epochs you used during training.")

A little more concise:

ValueError(f"Epoch count in checkpoint ({engine.state.epoch}) is larger than the `engine.state.max_epochs` of engine ({prior_max_epochs}). To further train from checkpoint construct trainer with `max_epochs` larger than checkpoint's epoch count. To use for inference `max_epochs` must be the same as checkpoint.")

Hi @ericspod ,

I updated the PR for the error message, slightly modified the inference part.
Could you please help review it again?

Thanks.

/black

/black