Skip to content

Patch bump to 0.32.1 - #3816

Merged
zcbenz merged 1 commit into
ml-explore:mainfrom
zcbenz:patch-bump
Jul 8, 2026
Merged

zcbenz merged 1 commit into
ml-explore:mainfrom
zcbenz:patch-bump

Conversation

@zcbenz

@zcbenz zcbenz commented Jul 7, 2026

Copy link
Copy Markdown
Member

No description provided.

@zcbenz
zcbenz merged commit 57c66ca into ml-explore:main Jul 8, 2026
28 checks passed
@zcbenz
zcbenz deleted the patch-bump branch July 8, 2026 00:18
inureyes added a commit to lablup/mlxcel that referenced this pull request Jul 9, 2026
## Summary
- Bump all three synchronized MLX pin locations to upstream `57c66cac7cb3e5b1eb350488a61f1506b40d39f8` (`Patch bump to 0.32.1`, ml-explore/mlx#3816).
- Fix mlxcel's Metal compiled patch overlay for the latest `compiled_collapse_contiguous_dims` return shape by binding `negative_strides`, selecting the large-index path for negative strides, and generating the rank-1 large strided Metal kernel that upstream MLX now needs.
- Note the relevant upstream range includes ml-explore/mlx#3804 (`Fix fp quantized matvec for output dim < 8`), ml-explore/mlx#3728 (`Add math mode option for custom Metal kernels`), and ml-explore/mlx#3720 (`Fix compiled kernel correctness for negative-strided inputs`).

## Test plan
- [x] `cargo build --release --bin mlxcel-bench-decode`
- [x] `cargo test --release -p mlxcel-core sparse_v_kernel_threshold_zero_matches_graph`
- [x] `cargo test --release -p mlxcel-core delegated_fused_kernel_matches_reference_over_200_steps`
- [x] `cargo test --release -p mlxcel-core delegated_steel_envelope_matches_cold_only_fused_over_200_steps`
- [x] `python3 scripts/ci/check_cross_repo_refs.py`

Closes #703
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants