Update support for HumanEval

[mcarbin]

What does this PR do?

Updated the support for HumanEval to compute the pass@k metric with n samples. This specifically, separates n and k, which were previously coupled together to have the same value. Now we can compute pass@k with n > k as is the case for implementations in other work.

What issue(s) does this change relate to?

N/A

Before submitting

• Have you read the contributor guidelines?
• Did you update any related docs and document your change?
• Did you update any related tests and add any new tests related to your change? (see testing)
• Did you run the tests locally to make sure they pass?
• Did you run pre-commit on your change? (see the pre-commit section of prerequisites)

[mvpatel2000]

Can you please add a unit test?

[mcarbin]

Sure, @mvpatel2000. Would it be acceptable to use the previous tests (#2301?), update (set different values of pass_at_k)and post output of tests?

[mvpatel2000]

Sure, @mvpatel2000. Would it be acceptable to use the previous tests (#2301?), update (set different values of pass_at_k)and post output of tests?

Yep, should be fine to update previous tests. No need to paste output -- CI/CD will run it automatically so as long as it passes it should be fine

[mcarbin]

@mvpatel2000 @dakinggg, open to any more input. Otherwise, latest is passing tests and I've resolved other asks.

[mvpatel2000]

LGTM, but will wait for Daniel to approve since he has more context

[dakinggg]

LGTM