natten: add gb300 aarch64 wheel to v1.5.0 index#58
Merged
Conversation
PyTorch's prebuilt arch list (torch.cuda.get_arch_list()) does not include sm_103, so the derived TORCH_CUDA_ARCH_LIST omits it and the NATTEN wheel ships no sm_103a kernels. On GB300 the Blackwell FNA/FMHA kernels then fail with "Failed to initialize the CUTLASS kernel" / "no kernel image is available", producing silent all-zero outputs (gray-noise images, gibberish reasoner). Append 10.3 to NATTEN_CUDA_ARCH so the wheel includes sm_103a kernels. Scoped to the natten package only. Relates to cosmos-framework#42. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Add LOCAL_VERSION_SUFFIX build option to tag custom build variants (e.g. gb300) on the wheel's local version, and register the natten 0.21.6.dev6+cu130.torch210.gb300 aarch64 wheel in the index. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
lfengad
approved these changes
Jun 16, 2026
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
2 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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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
LOCAL_VERSION_SUFFIXbuild option to tag custom build variants on the wheel's local version (e.g.LOCAL_VERSION_SUFFIX=gb300→...+cu130.torch210.gb300). Default empty → no change to existing wheel names (backward compatible).natten-0.21.6.dev6+cu130.torch210.gb300-cp313-cp313-linux_aarch64.whl(GB300 / sm_103a aarch64) wheel in the v1.5.0 package index.Files
bin/_build.sh— deriveLOCAL_VERSIONfromLOCAL_VERSION_SUFFIXand pass tofix_wheel.py; documents the manual-build usage in a comment.bin/build.sh— passLOCAL_VERSION_SUFFIXthrough theenv -isandbox.docs/v1.5.0/natten/index.html— add the gb300 wheel link.Notes
LOCAL_VERSION_SUFFIX=gb300 bin/build.sh ... natten) and uploaded to the v1.5.0 GitHub release. Its sha256 (cd63a31…fde9b) was verified to match the release asset and the index entry.bin/_build.shcomment.🤖 Generated with Claude Code