Update distopt API for coalesced NCCL calls

[timmoon10]

What does this PR do ?

The Pytorch API for coalescing NCCL calls has changed in 2.1.0 (see pytorch/pytorch#98793), which required some changes in the Apex wrapper functions (see NVIDIA/apex#1663).

Collection: NLP

Changelog

• Use updated API for coalescing NCCL calls in distributed optimizer

Usage

Set the optimizer to distributed_fused_adam in the config file.

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

• PyTorch API change in [c10d] Faster coalescing pytorch/pytorch#98793
• Apex API change in Update distributed optimizer with new coalescing manager API in PyTorch apex#1663

[ericharper]

is this commit needed with the dlfw 23.06 pytorch container?