Update Seq2Seq QA example script to use SQuAD metric.

[karthikrangasai]

What does this PR do?

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. Discussed in this thread
• 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.

@sgugger @NielsRogge

[sgugger]

This is a good switch to the SQUAD metric! I realize however that the processing is not completely adapted to what we do: the answer is generated from the first feature created by each example, so there is really no point keeping all the features.

We either need to figure out a score for a generated answer, so that we can pick the best answer corresponding to all the features generated by one example, or we can just simplify the preprocessing and only keep one feature per example.

[karthikrangasai]

In the extractive method, the logits are used to score the outputs but for generative there isn't any value we can define to rank it because the sentence is already generated. So I suppose we can just use one feature per example.

[sgugger]

It looks like GitHub does not like your rebase and is showing a diff with ~200 files touched. Could you close this PR and create a fresh one?

[karthikrangasai]

Sorry for the late reply.
I didn't notice the error while pushing the code.
I have created the new PR here at #14335 .
I will close this one now.