Adding ViT Autoencoder

[finalelement]

Description

This PR is for adding a modified version of the ViT architecture for the reason because the ViT does not give the same sized output as the input. The proposed architecture allows for the flexibility to reconstruct the same size output as that of the input size

Status

This is work in progress, do not merge yet.

Integration tests need to be commented upon
Doc strings & Documentation also needs to be updated

Types of changes

• Non-breaking change (fix or new feature that would not break existing functionality).
• New tests added to cover the changes.
• Integration tests passed locally by running ./runtests.sh -f -u --net --coverage.
• Quick tests passed locally by running ./runtests.sh --quick --unittests.
• In-line docstrings updated.
• Documentation updated, tested make html command in the docs/ folder.

[Nic-Ma]

Thanks for the quick PR.
I put some minor comments in it.
@wyli @ericspod Could you please help also review it?

Thanks.

[wyli]

Hi @ahatamiz, could you please help review the implementation? thanks!

[ahatamiz]

Hi @ahatamiz, could you please help review the implementation? thanks!

Hi @wyli

I have reviewed and added my comments. The implementation is mostly good but need to remove anything about classification.

Thanks,

[finalelement]

@wyli @Nic-Ma all comments have been addressed apart from the variable naming. All tests pass, there is something going on in the croppad transform which to which I made no changes at all. I also tried pulling the latest, but this one ./runtests.sh --codeformat fails for the array.py in croppad transforms.

Please let me know if there are any more things to address from my side, otherwise lets merge and close this off

[wyli]

/build

[wyli]

Thanks, it looks good to me.