build: bump nvidia-cutlass-dsl to 4.5.0 and nvidia-cudnn-frontend to 1.24.0 - #4466
Merged
Conversation
…1.24.0 Pin nvidia-cutlass-dsl[cu13] to ==4.5.0 (was unpinned, resolving to 4.6.0.dev0) and bump nvidia-cudnn-frontend to ==1.24.0. Regenerate uv.lock. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> Signed-off-by: oliver könig <okoenig@nvidia.com>
Contributor
|
LGTM — clean dependency pin update. Changes reviewed:
No code, config, or API changes — only build metadata. Suggested test cases No perf tests impacted. L0 + L1 CI (already triggered via labels) covers lockfile validity and import smoke tests. |
Contributor
Author
|
/ok to test 80819f2 |
thomasdhc
reviewed
Jun 23, 2026
| "nvidia-cudnn-frontend==1.23.0", | ||
| "nvidia-cutlass-dsl[cu13]; sys_platform == 'never'" # Rely on system site packages install | ||
| "nvidia-cudnn-frontend==1.24.0", | ||
| "nvidia-cutlass-dsl[cu13]==4.5.0; sys_platform == 'never'" # Rely on system site packages install |
Contributor
There was a problem hiding this comment.
did you mean to remove the sys_platform == never for cutlass?
Contributor
Author
There was a problem hiding this comment.
yeah, we want to bump to a specific version so i think this should be fine..
- nvidia-cutlass-dsl is marked `sys_platform == 'never'` in pyproject, so uv never installs it; the effective install is the Dockerfile.ci pip step. Bump that 4.4.1 -> 4.5.0 and revert the no-op pyproject version pin. - flash_mla lives only in Megatron-LM's no_pypi_wheels dependency group, which is not transitive through the editable path dependency. Declare it in this repo's own no_pypi_wheels group (+ uv source + dependency-metadata, pinned to the same FlashMLA rev as Megatron-LM) so the image build's `uv sync --all-groups` installs it. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> Signed-off-by: oliver könig <okoenig@nvidia.com>
Contributor
Author
|
/ok to test c1c3808 |
thomasdhc
previously approved these changes
Jun 24, 2026
The main pyproject bump only reaches the megatron-bridge image; the fw_final stage syncs docker/common/fw_pyproject.toml at /opt/NeMo-FW, which still pinned nvidia-cudnn-frontend==1.23.0 and downgraded it in the final image. Bump it here so 1.24.0 reaches fw-final as well. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> Signed-off-by: oliver könig <okoenig@nvidia.com>
Contributor
Author
|
/ok to test 7ab3881 |
liding-nv
pushed a commit
that referenced
this pull request
Jun 26, 2026
…1.24.0 (#4466) Signed-off-by: oliver könig <okoenig@nvidia.com> Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
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
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.
Claude summary
What
nvidia-cutlass-dsl[cu13]to==4.5.0(was unpinned → resolved to4.6.0.dev0).nvidia-cudnn-frontendto==1.24.0(was1.23.0).uv.lock.Lockfile delta
Test plan
needs-more-testsapplied)Quarantined tests (this bump)
None yet — will be appended as flakes are identified during CI iteration.