Add MTP support for hybrid models

[rkarimimahab]

What does this PR do ?

(1) supporting to use hybrid mamba models as mtp_model_layer.
(2) splitting the MTP loss calculation in the GPT model’s forward pass into a separate function.
(3) supporting MTP layer repetition.

⚠️ For major changes (either in lines of code or in its impact), please make sure to first share discuss a design-doc with the team.

Contribution process

flowchart LR
    A[Pre-checks] --> B[PR Tests]
    subgraph Code Review/Approval
        C1[Expert Review] --> C2[Final Review]
    end
    B --> C1
    C2 --> D[Merge]

Loading

Pre-checks

• I want this PR in a versioned release and have added the appropriate Milestone (e.g., Core 0.8)
• I have added relevant unit tests
• I have added relevant functional tests
• I have added proper typing to my code Typing guidelines
• I have added relevant documentation
• I have run the autoformatter.sh on my PR

Code review

The following process is enforced via the CODEOWNERS file for changes into megatron/core. For changes outside of megatron/core, it is up to the PR author whether or not to tag the Final Reviewer team.

For MRs into `main` branch

(Step 1): Add PR label Expert Review

(Step 2): Collect the expert reviewers reviews

1. Attach the Expert Review label when your PR is ready for review.
2. GitHub auto-assigns expert reviewers based on your changes. They will get notified and pick up your PR soon.

⚠️ Only proceed to the next step once all reviewers have approved, merge-conflict are resolved and the CI is passing.
Final Review might get declined if these requirements are not fulfilled.

(Step 3): Final Review

1. Add Final Review label
2. GitHub auto-assigns final reviewers based on your changes. They will get notified and pick up your PR soon.

(Optional Step 4): Cherry-pick into release branch

If this PR also needs to be merged into core_r* release branches, after this PR has been merged, select Cherry-pick to open a new PR into the release branch.

For MRs into `dev` branch

The proposed review process for `dev` branch is under active discussion.

MRs are mergable after one approval by either eharper@nvidia.com or zijiey@nvidia.com.

Merging your PR

Any member of core-adlr and core-nemo will be able to merge your PR.

[copy-pr-bot]

This pull request requires additional validation before any workflows can run on NVIDIA's runners.

Pull request vetters can view their responsibilities here.

Contributors can view more details about this message here.

[shifangx]

This pr implements three features:
(1) supporting to use Mamba layer as mtp_model_layer.
(2) splitting the MTP loss calculation in the GPT model’s forward pass into a separate function.
(3) supporting MTP layer repetition.
I have no concerns about the implementation of the first two features.
As for the last feature, we need to consider how to support scenarios where multiple MTP layers are placed on different PP ranks for computation. These MTP layers share the same parameters value, but be placed onto different vpp stage, in order to make sure each vpp stage do not have much computation task.
Maybe we can use assert to guide users can not place mtp layers into different vpp stage currently, and support this feature in future.

[deepakn94]

This pr implements three features: (1) supporting to use Mamba layer as mtp_model_layer. (2) splitting the MTP loss calculation in the GPT model’s forward pass into a separate function. (3) supporting MTP layer repetition. I have no concerns about the implementation of the first two features. As for the last feature, we need to consider how to support scenarios where multiple MTP layers are placed on different PP ranks for computation. These MTP layers share the same parameters value, but be placed onto different vpp stage, in order to make sure each vpp stage do not have much computation task. Maybe we can use assert to guide users can not place mtp layers into different vpp stage currently, and support this feature in future.

This is a good point. Let's go with the assertion for now re: 3.

[rkarimimahab]

This is a good point. Let's go with the assertion for now re: 3.

I added the assert, thanks!

[sancha]

/ok to test 9cc1668

[deepakn94]

For posterity, this PR was re-merged as #3207 with some bugfixes.