[Megatron-FSDP] Fix incorrect gradient scaling target.

[cspades]

What does this PR do ?

• When not using per-token loss or collective averaging, we weren't actually applying a 1/DP scaling to the gradient for Megatron-FSDP. This was experienced outside of Megatron-LM, e.g. fully_shard and explains why we don't have gradient scaling parity with FSDP2. Will add golden value or FSDP2 comparison tests later.

Details

• gbuf.data != bucket.data when the gradient is sharded. So we were applying the scaling to the accumulated gradient shard, then adding in a non-scaled reduced gradient. So we end up with something like:

((((0 / DP + grad_step_1) / DP + grad_step_2) / DP + grad_step_3) / DP + grad_step_4 ... )

When gradient accumulation period > 1, we are scaling accumulated gradients by DP**(i-grad_acc_steps) where i is the i-th accumulated gradient, and when not accumulating, then we just don't scale by DP. 😓

⚠️ For major changes (either in lines of code or in its impact), please make sure to first share a design doc with the team. If you're unsure what's the best way to do so, contact the @mcore-oncall.

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

Feel free to message or comment the @mcore-oncall to help accelerate your merge into main. The less complex your PR is, the faster it will be approved and merged!

(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.

[shjwudp]

/ok to test e0ac855