ModelCheckpoint save_best_only not working about pytorch-widedeep HOT 2 CLOSED

hey @egmaminta

if you have a look to the examples folder, for example here:

or in the docs:

https://pytorch-widedeep.readthedocs.io/en/latest/trainer.html#pytorch_widedeep.training.Trainer

you will see that optimizers, callbacks, etc, must be passed when you instantiate the Trainer, not to the fit method

In your case:

trainer = Trainer(
        model,
        objective="multiclass_focal_loss",
        optimizers=deep_optimizer, # no need of dict, you can pass it directly
        lr_schedulers=deep_sch, # no need of dict, you can pass it directly
        callbacks=[model_checkpoint],
        metrics=[Accuracy],
    )

Hope this helps!

Ciao!

from pytorch-widedeep.

@jrzaurin thank you, good sir! I must've missed that, oops.

from pytorch-widedeep.