Model saving bug

[WenjieDu]

1. System Info

PyPOTS v0.6

2. Information

• The official example scripts
• My own created scripts

3. Reproduction

When model_saving_strategy is set as better, PyPOTS doesn't save models when loss gets decreased.

4. Expected behavior

Models should be properly saved

[vemuribv]

Hi @WenjieDu! I think this may have had an unintended effect--now (in v0.7 and later) even when model_saving_strategy is set as "best", PyPOTS saves the model every time the loss is decreased. In v0.6, setting to "best" had the intended behavior (saving only the best model at the very end after training is finished).

[WenjieDu]

Bug confirmed. Thank you, @vemuribv! I should have taken model_saving_strategy into consideration within confirm_saving when saving, but I forgot to do so when global replacing.

I'm reopening this issue and going to submit a PR to fix this. Could you please help review it?

[WenjieDu]

@vemuribv PR submitted. You are also very welcome to propose a better way to optimize the current saving strategy ;-)