Patch EOD out of inference results

[tdene]

What does this PR do ?

⚠️ 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

Pre-checks

• 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

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!

All PRs start as draft. If you open a non-draft PR, it will be automatically converted to draft.

Step 1: Mark PR as "Ready for Review"

1. When your PR is ready, click Ready for Review.
2. An oncall reviewer is auto-assigned and expert reviewers are notified based on your changes.
   • Some PRs may jump straight to step 2. This is determined by .github/CODEOWNERS.

⚠️ Only mark as ready once merge-conflicts are resolved and the CI is passing.
Final Review might get declined if these requirements are not fulfilled.

Step 2: Final Review

For PRs that change megatron/core, once all expert reviewers have approved, the Final Review label is applied automatically and final reviewers are assigned.

For PRs outside megatron/core, this step is skipped.

Step 3: Approved

Once all required reviewers have approved, the Approved label is applied automatically.

Merge

Any member of mcore-engineers will be able to merge your PR.

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.

[copy-pr-bot]

Auto-sync is disabled for draft pull requests in this repository. Workflows must be run manually.

Contributors can view more details about this message here.

[tdene]

/claude review

[santhnm2]

LGTM but just 1 question - how does this work for stop sequences as opposed to single stop tokens? Does any additional functionality need to be added for these cases?

[tdene]

LGTM but just 1 question - how does this work for stop sequences as opposed to single stop tokens? Does any additional functionality need to be added for these cases?

Beyond a doubt, additional functionality needs to be added for these cases.

However, there's two reasons I'm not including that in this PR:

• Last I heard, stop strings had some sort of bug (sometimes the whole stop string is emitted, sometimes it's trimmed?)
• We don't know what stop string behavior the customers (i.e. Gym) expect. We do know that customers expect the EOD token to not be detokenized. We'll have to research what they want for stop strings.

[santhnm2]

LGTM but just 1 question - how does this work for stop sequences as opposed to single stop tokens? Does any additional functionality need to be added for these cases?

Beyond a doubt, additional functionality needs to be added for these cases.

However, there's two reasons I'm not including that in this PR:

• Last I heard, stop strings had some sort of bug (sometimes the whole stop string is emitted, sometimes it's trimmed?)
• We don't know what stop string behavior the customers (i.e. Gym) expect. We do know that customers expect the EOD token to not be detokenized. We'll have to research what they want for stop strings.

The bug in stop words processing seems concerning, @shanmugamr1992 do you have more details here? But otherwise maybe for now we can just add a TODO to this PR to handle stop words eventually.

[santhnm2]

LGTM but can we add a unit test for the stop sequence case also?

[svcnvidia-nemo-ci]

🔄 Merge queue validation started!

You can track the progress here: https://github.com/NVIDIA/Megatron-LM/actions/runs/23262285801