Skip to content

release 0.3.11 #34

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
asher merged 1 commit into from
Aug 9, 2026
Merged

release 0.3.11 #34

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
12 changes: 8 additions & 4 deletions CHANGELOG.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -6,6 +6,14 @@ adhere to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).

## [Unreleased]

## [0.3.11]

### Changed
- MoE prefill gather (sorted-rhs NAX path) is 12-28% faster per call
below ~64 rows per expert, biggest at 128-529-token chunks on top-8
256-expert shapes (bit-identical; KQ_GATHER_RHS_NAX_BM forces the
tile height).

## [0.3.10]

### Added
Expand All @@ -20,10 +28,6 @@ adhere to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
fine buffers during deep prefill.

### Changed
- MoE prefill gather (sorted-rhs NAX path) is 12-28% faster per call
below ~64 rows per expert, biggest at 128-529-token chunks on top-8
256-expert shapes (bit-identical; KQ_GATHER_RHS_NAX_BM forces the
tile height).
- iq2_xxs / iq2_xs / iq2_s / iq3_s MoE gather decode is 9-12% faster per
call (hoisted block scale, byte-indexed grids); the ext mat-vec at
verify widths 2..8 gains 7-10% on the same codecs.
Expand Down
2 changes: 1 addition & 1 deletion mlx_kquant/_version.py
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1 +1 @@
__version__ = "0.3.10"
__version__ = "0.3.11"