check for first or last stage

[ericharper]

What does this PR do ?

Fix bug when inference.compute_logprob=True. Only first and last pipeline stages have a response.

Collection: NLP

Changelog

• Add specific line by line info of high level changes in this PR.

Usage

• You can potentially add a usage example below

# Add a code snippet demonstrating how to use this 

Before your PR is "Ready for review"

Pre checks:

• Make sure you read and followed Contributor guidelines
• Did you write any new necessary tests?
• Did you add or update any necessary documentation?
• Does the PR affect components that are optional to install? (Ex: Numba, Pynini, Apex etc)
  • Reviewer: Does the PR have correct import guards for all optional libraries?

PR Type:

• New Feature
• Bugfix
• Documentation

If you haven't finished some of the above items you can still open "Draft" PR.

Who can review?

Anyone in the community is free to review the PR once the checks have passed.
Contributor guidelines contains specific people who can review PRs to various areas.

Additional Information

• Related to # (issue)

[titu1994]

Thanks!

[yidong72]

Looks good. just one comment.

[yidong72]

why we need this condition not parallel_state.model_parallel_is_initialized()?
if the model_parallel is not initialized, it will throw an exception at the very beginning right?

[ericharper]

I was just trying to make it robust. This method will only be called after parallel_state is initialized?

[yidong72]

this parallel_state.model_parallel_is_initialized() method will always be true after parallel init.

def model_parallel_is_initialized():
    """Check if model and data parallel groups are initialized."""
    if _TENSOR_MODEL_PARALLEL_GROUP is None or \
        _PIPELINE_MODEL_PARALLEL_GROUP is None or \
        _DATA_PARALLEL_GROUP is None:
        return False
    return True

[ericharper]

I mean will this method ever be used without model parallel init?

[yidong72]

no. it is only used with model parallel.