Update docs for fill_last_batch parameter to match the real behavior

[bonlime]

Why we need this PR?

• It fixes a bug
• It adds new feature needed because why we need this feature
• Refactoring to improve what

What happened in this PR?

Fixed fill_last_batch parameter docstring to match the real behavior

[JanuszL]

I think it is still not fully accurate. There is also a last_batch_padded in the iterator and pad_last_batch from the reader which combined together impacts the overall behaviour. You can check the table with an example at the end of the function docstring you have updated.

[bonlime]

would changing

Whether to fill the last batch with data from next epoch

to

Whether to fill the last batch with data up to 'self.batch_size'

make it more accurate? I do understand that a combination of these 3 parameters may produce unexpected behavior but I think that it's enough to say what one parameter changes and everything else is explained in the table at the end of the docstring

[JanuszL]

would changing

Whether to fill the last batch with data from next epoch

to

Whether to fill the last batch with data up to 'self.batch_size'

make it more accurate? I do understand that a combination of these 3 parameters may produce unexpected behavior but I think that it's enough to say what one parameter changes and everything else is explained in the table at the end of the docstring

I agree with the new wording. Please adjust accordingly.

[JanuszL]

@bonlime - can you add a sign-off to your commit so we can merge it?

[bonlime]

Okay. I have to admit I don't know how to sign previous commits :c

[JanuszL]

That is enough. Thanks for the contribution!