Cannot Properly Resume Training from Loaded Checkpoint

[SaviNimz]

The current Trainer implementation does not correctly support resuming training from a loaded checkpoint. Although load_checkpoint() returns the saved epoch number, the fit() method always starts training from:

for epoch in range(self._epochs):

[spencerirving]

Hey @SaviNimz!

I worked on this implementation of the checkpointing and see your issue.

If @blkdmr is ok with it I can take care of this issue!

[blkdmr]

Sure @spencerirving!