-
Notifications
You must be signed in to change notification settings - Fork 57
Bump gram-newton-schulz to 0.1.5 (cutlass-dsl 4.5.2)#95
Merged
Merged
Conversation
Bump the optional [gns] extra's gram-newton-schulz pin 0.1.4 -> 0.1.5. 0.1.5 directly pins quack-kernels==0.5.0 and nvidia-cutlass-dsl==4.5.2 (was quack-kernels>=0.4.1 / cutlass-dsl 4.4.2), so update the quack-kernels pin to 0.5.0 and the transitive-pin comment accordingly. The cutlass-dsl pin now matches the 4.5.2 used by Flash-Attention-4 / Blackwell stacks, resolving the conflict the README/CHANGELOG previously documented; reconcile those docs to the new versions and framing. Verified against the PyPI metadata for gram-newton-schulz 0.1.5 and a pip --dry-run resolve, which both produce gram-newton-schulz 0.1.5 / quack-kernels 0.5.0 / nvidia-cutlass-dsl 4.5.2.
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
What
Bump the optional
[gns]/[gram-newton-schulz]extra's pin fromgram-newton-schulz==0.1.4to0.1.5inrequirements_gns.txt, and reconcilethe transitive pins it controls.
Transitive pins (reconciled, not just one line)
gram-newton-schulz==0.1.5directly pins its dependencies differently from 0.1.4:Requires-Dist)Requires-Dist)quack-kernels>=0.4.1==0.5.00.4.1→0.5.0nvidia-cutlass-dsl==4.4.2==4.5.24.4.2→4.5.2Both transitive versions changed, so this is not a no-op renumber. The explanatory
comment in
requirements_gns.txtwas updated to cite the 0.1.5-correct cutlass version.Docs reconciled
The cutlass-dsl pin is the whole point of these explicit pins — drift has historically
broken FlashAttention-4 downstream. Notably,
4.5.2is the same cutlass version theREADME/CHANGELOG attribute to FA4 / Blackwell stacks, so the previously-documented
4.4.2-vs-4.5.2conflict is now resolved, not merely renumbered. Updated:README.md(Quick Start note +trainextra note):4.4.2→4.5.2, and the "installin a separate environment / conflicts with FA4" wording reworded to reflect that the
extra's cutlass pin now matches FA4 / Blackwell.
CHANGELOG.md(Unreleased): dropped the now-false4.4.2conflict statements from thebreaking-install entry and added a bullet documenting the bump and the resulting
nvidia-cutlass-dsl4.4.2→4.5.2alignment.Verification
gram-newton-schulz==0.1.5exists on PyPI.Requires-Distfor 0.1.5 from the PyPI JSON API:quack-kernels==0.5.0,nvidia-cutlass-dsl==4.5.2. Cross-checkedquack-kernels==0.5.0requiresnvidia-cutlass-dsl>=4.5.2(consistent with the==4.5.2pin).pip install --dry-runresolve ofgram-newton-schulz==0.1.5in a scratch venvresolved to exactly gram-newton-schulz 0.1.5 / quack-kernels 0.5.0 /
nvidia-cutlass-dsl 4.5.2, matching the written pins.