fix: Handle BLIP-2 model output format

[thisisiron]

What does this PR do?

I have modified the code to return logits in addition to loss.

Instead of:

return {"loss": loss}

The code now returns:

return {
    "loss": loss,
    "logits": logits
}

Output:

Done!
First values of original logits: tensor([[ 1.9322,  1.9379,  7.4007],
        [-1.4741, -1.1189,  8.6593],
        [-1.4213, -1.2489,  6.1976]])
First values of HF logits: tensor([[ 1.9322,  1.9379,  7.4007],
        [-1.4741, -1.1189,  8.6593],
        [-1.4213, -1.2489,  6.1976]])
Looks ok!
Generating a caption...
Original generation: ['merlion']
HF generation: ['Question: what object is in this image? Answer: merlion']

Fixes # (issue)

Before submitting

• This PR fixes a typo or improves the docs (you can dismiss the other checks if that's the case).
• Did you read the contributor guideline,
  Pull Request section?
• Was this discussed/approved via a Github issue or the forum? Please add a link
  to it if that's the case.
• Did you make sure to update the documentation with your changes? Here are the
  documentation guidelines, and
  here are tips on formatting docstrings.
• Did you write any new necessary tests?

Who can review?

Anyone in the community is free to review the PR once the tests have passed. Feel free to tag
members/contributors who may be interested in your PR.

@amyeroberts, @qubvel