update mutligpu readme and MllamaForConditionalGeneration import

[AAndersn]

Fixes an incorrect parameter in documentation for multigpu that says int4 instead of 4bit for quantization type.

Fixes incorrect import of MllamaForConditionalGeneration from transformers instead of transformers.models.mllama.modeling_mllama

Fixes #680

Before submitting

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

Thanks for contributing 🎉!

[wukaixingxp]

Hi! Can you check your transformer version? I think the latest transformer can do from transformers import MllamaForConditionalGeneration

[AAndersn]

@wukaixingxp You are right.

I think I might have had the 4.44.0 when I created that bug ticket and PR last night. Re-running with transformers 4.45.0 this morning and from transformers import MllamaForConditionalGeneration works fine

[wukaixingxp]

Thanks for your help! We want to bump the transformer version to 4.45.0 but it has a bug as stated in my PR. Now people must pip install from source for transformers to avoid this bug. We are waiting for a new release of transformers pip package.

[init27]

@AAndersn thanks again for the PR-since this is fixed in the latest HF version can you take another parse please?

[AAndersn]

Sorry for the delay. Yes, I will fix my branch to resolve the conflict with the other portion that has already been fixed

[init27]

Many Thanks!