Add a basic Meta Llama SUT

[wpietri]

This uses Meta's client to add support for their Meta LLama hosting service so that we can run some of their beta models.

It's my first time writing a SUT and it wasn't clear to me which bits are necessary and which aren't, so this is pretty minimal. What else does it need?

I certainly should add more retrying. What's our current best practice?

Also, I'm doing this directly in the main code base, like the Together SUT. Is there a reason to make it a plugin?

[github-actions]

MLCommons CLA bot All contributors have signed the MLCommons CLA ✍️ ✅

[bkorycki]

Yay congrats on your first SUT!

It's my first time writing a SUT and it wasn't clear to me which bits are necessary and which aren't, so this is pretty minimal. What else does it need?

This SUT has everything it needs!

I certainly should add more retrying. What's our current best practice?

Most SUTs that use python clients use the @retry decorator from retry_decorator.py. You can look at the OpenAI SUT for an example.

Also, I'm doing this directly in the main code base, like the Together SUT. Is there a reason to make it a plugin?

Like we discussed, the original reason for making things plugins isn't really important anymore (i.e. making it easier/more lightweight for developers using the library).
I think the only reason to make it a plugin now is to be consistent with the rest of the codebase. And I think it makes it easier to find. But it's up to you.