Add mcore full TE transformer layer spec

[jbaczek]

What does this PR do ?

Add spec for model built from te.TransformerLayer

Collection: nlp

Changelog

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

Usage

Use name of the spec either on config level:

[...]
model:
  name: megatron_gpt_full_te_layer_autocast
[...]

or from CLI model.name=megatron_gpt_full_te_layer_autocast

Jenkins CI

To run Jenkins, a NeMo User with write access must comment jenkins on the PR.

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

• To be merged instead of https://github.com/NVIDIA/NeMo/pull/8316/files

[jbaczek]

jenkins

[minitu]

@jbaczek Could you add the changes in this NeMo PR to the AutocastTransformerLayer here as well? We would need this to comply with the changes to TP knobs in this TE commit.

[jbaczek]

jenkins

[jbaczek]

@ericharper Jenskins tests fail because of the fixed version of mcore. What should we do about it?

[jbaczek]

jenkins

[jbaczek]

jenkins

[ShriyaPalsamudram]

jenkins