Fix/dsv4 cp relax assert offline packed thd - #6158
Merged
thomasdhc merged 2 commits intoJul 31, 2026
Merged
Conversation
…ked THD DSv4 hybrid + CP > 1 previously required sequence_packing_scheduler, but offline-packed SFT supplies pre-packed THD data externally without the scheduler. Allow cp_partition_mode=contiguous as an alternative signal that THD inputs are externally provided, so Bridge offline-packed SFT does not need to set sequence_packing_scheduler to bypass the validation. Co-Authored-By: Claude Sonnet 4.6 (1M context) <noreply@anthropic.com>
…d + CP
The assert required sequence_packing_scheduler to be set for DSv4 hybrid
with CP > 1, assuming THD inputs always come from MCore online packing.
This blocked offline-packed SFT (externally pre-packed THD data) without
a semantically correct workaround.
Remove the assert entirely — the forward path already enforces THD format
via ValueError("DSv4 Hybrid with CP requires qkv_format=thd") if THD data
is not actually provided. Legitimate misconfigurations are caught at runtime.
Co-Authored-By: Claude Sonnet 4.6 (1M context) <noreply@anthropic.com>
cuichenx
approved these changes
Jul 30, 2026
hxbai
approved these changes
Jul 31, 2026
Contributor
|
/ok to test efc4009 |
Contributor
|
🔄 Merge queue validation started! You can track the progress here: https://github.com/NVIDIA/Megatron-LM/actions/runs/30649507989 |
Contributor
|
🔄 Merge queue validation started! You can track the progress here: https://github.com/NVIDIA/Megatron-LM/actions/runs/30654314227 |
github-merge-queue
Bot
removed this pull request from the merge queue due to failed status checks
Jul 31, 2026
github-merge-queue
Bot
removed this pull request from the merge queue due to failed status checks
Jul 31, 2026
github-merge-queue
Bot
removed this pull request from the merge queue due to failed status checks
Jul 31, 2026
github-merge-queue
Bot
removed this pull request from the merge queue due to failed status checks
Jul 31, 2026
Contributor
|
🔄 Merge queue validation started! You can track the progress here: https://github.com/NVIDIA/Megatron-LM/actions/runs/30664843680 |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
5 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
What does this PR do?
Remove the overly strict sequence_packing_scheduler is not None assertion for DSv4 hybrid + CP, which blocked offline-packed SFT. the forward path already enforces THD format correctness at runtime via ValueError("DSv4 Hybrid with CP requires qkv_format='thd'").
Issue tracking
For PRs from open-source community contributors:
Linked issue:
Contribution process
Pre-checks
Code review
Feel free to message or comment @NVIDIA/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"
.github/CODEOWNERS.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, theFinal Reviewlabel 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
Approvedlabel is applied automatically.Merge
Any member of mcore-engineers will be able to merge your PR.