mlx 0.30.4

[BrewTestBot]

Created by brew bump

---

Created with brew bump-formula-pr.

• resource blocks have been checked for updates.

release notes

## Highlights
- Metal: Much faster vector fused grouped-query attention for long context
- CUDA: Several improvements to speed up LLM inference for CUDA  backend
- CUDA: Support for dense MoEs
- CUDA: Better support for consumer GPUs (4090, 5090, RTX 6000, ...)
What's Changed

patch bump for next release by @awni in patch bump for next release ml-explore/mlx#2991
Fix fence by @awni in Fix fence ml-explore/mlx#2998
Reverts changing the MLX_IBV_DEVICES to MLX_JACCL_DEVICES by @angeloskath in Reverts changing the MLX_IBV_DEVICES to MLX_JACCL_DEVICES ml-explore/mlx#2999
fix distributed all_to_sharded bias shard axis from -2 to -1 by @gufengc in fix distributed all_to_sharded bias shard axis from -2 to -1 ml-explore/mlx#2987
Fix sharding of quantized models with non-power-of-2 bits  by @kernelpool in Fix sharding of quantized models with non-power-of-2 bits  ml-explore/mlx#3006
Update CCCL to v3.1.3 by @zcbenz in Update CCCL to v3.1.3 ml-explore/mlx#3012
Fix python package install path in stubgen by @zcbenz in Fix python package install path in stubgen ml-explore/mlx#3009
Type Enhancement for Func Transforms and Bug Fix by @XXXXRT666 in Type Enhancement for Func Transforms and Bug Fix ml-explore/mlx#3003
Do not clear disk space in setup-linux by @zcbenz in Do not clear disk space in setup-linux ml-explore/mlx#3013
Do not give workflow boolean inputs default values by @zcbenz in Do not give workflow boolean inputs default values ml-explore/mlx#3014
Fix negative dim indexing by @MillaFleurs in Fix negative dim indexing ml-explore/mlx#2994
Windows CI by @zcbenz in Windows CI ml-explore/mlx#3021
Optimize erf function with expm1f in Metal backend by @bjornefisk in Optimize erf function with expm1f in Metal backend ml-explore/mlx#3025
[CUDA] Faster grouped mm by @zcbenz in [CUDA] Faster grouped mm ml-explore/mlx#3011
PR 3007 Fix Seg Fault by @MillaFleurs in PR 3007 Fix Seg Fault ml-explore/mlx#3008
Use higher precision for linspace with double by @awni in Use higher precision for linspace with double ml-explore/mlx#3029
Handle data smaller than BUFFER_SIZE in jaccl recv by @rltakashige in Handle data smaller than BUFFER_SIZE in jaccl recv ml-explore/mlx#3033
build 26.0 release in actions by @awni in build 26.0 release in actions ml-explore/mlx#3035
Remove xmlrunner from macOS CI by @zcbenz in Remove xmlrunner from macOS CI ml-explore/mlx#3032
Columnwise quantize by @nastya236 in Columnwise quantize ml-explore/mlx#2989
Turn nccl_stub into a normal target by @zcbenz in Turn nccl_stub into a normal target ml-explore/mlx#3037
Use cuda::std for math ops by @zcbenz in Use cuda::std for math ops ml-explore/mlx#3041
win: symbol exports and minor fixes by @dhiltgen in win: symbol exports and minor fixes ml-explore/mlx#3024
CUDA gather mv by @angeloskath in CUDA gather mv ml-explore/mlx#3039
Link with prebuilt OpenBLAS and fix shared libs build on Windows by @zcbenz in Link with prebuilt OpenBLAS and fix shared libs build on Windows ml-explore/mlx#3036
Allow take on empty array when it makes sense by @awni in Allow take on empty array when it makes sense ml-explore/mlx#3046
Add missing  include to buffer_cache.h by @Anri-Lombard in Add missing <algorithm> include to buffer_cache.h ml-explore/mlx#3053
Build and test python package on Windows CI by @zcbenz in Build and test python package on Windows CI ml-explore/mlx#3049
Fix some MSVC compilation errors by @zcbenz in Fix some MSVC compilation errors ml-explore/mlx#3048
Use C++20 by @zcbenz in Use C++20 ml-explore/mlx#3050
Faster two pass sdpa by @awni in Faster two pass sdpa ml-explore/mlx#3023
Find system-installed cuDNN on Windows by @zcbenz in Find system-installed cuDNN on Windows ml-explore/mlx#3052
Fix some NVCC warnings when building CUDA backend with MSVC by @zcbenz in Fix some NVCC warnings when building CUDA backend with MSVC ml-explore/mlx#3038
Hide symbols by default for mac/linux by @zcbenz in Hide symbols by default for mac/linux ml-explore/mlx#3057
[CUDA] Fast sorting by @awni in [CUDA] Fast sorting ml-explore/mlx#3060
Fix flaky macOS test by @awni in Fix flaky macOS test ml-explore/mlx#3063
Update pre-commit hooks and versions for clang-format, black, and isort by @NripeshN in Update pre-commit hooks and versions for clang-format, black, and isort ml-explore/mlx#3059
GPU discovery by @dhiltgen in GPU discovery ml-explore/mlx#3055
Add NAX Split-K GEMM for large-K matmuls to improve performance by @hxu296 in Add NAX Split-K GEMM for large-K matmuls to improve performance ml-explore/mlx#3018
Improve CPU discovery by @dhiltgen in Improve CPU discovery ml-explore/mlx#3068
Fix long cache file path on Windows by @zcbenz in Fix long cache file path on Windows ml-explore/mlx#3065
Better support consumer CUDA GPUs by @jessegross in Better support consumer CUDA GPUs ml-explore/mlx#3056
Delay load CUDA libs and resolve DLL paths at runtime by @zcbenz in Delay load CUDA libs and resolve DLL paths at runtime ml-explore/mlx#3061
Do not require ConcurrentManagedAccess when not used by @zcbenz in Do not require ConcurrentManagedAccess when not used ml-explore/mlx#3062
Fp qmv by @awni in Fp qmv ml-explore/mlx#2984
remove thrust by @awni in remove thrust ml-explore/mlx#3067

New Contributors

@gufengc made their first contribution in fix distributed all_to_sharded bias shard axis from -2 to -1 ml-explore/mlx#2987
@kernelpool made their first contribution in Fix sharding of quantized models with non-power-of-2 bits  ml-explore/mlx#3006
@bjornefisk made their first contribution in Optimize erf function with expm1f in Metal backend ml-explore/mlx#3025
@rltakashige made their first contribution in Handle data smaller than BUFFER_SIZE in jaccl recv ml-explore/mlx#3033
@dhiltgen made their first contribution in win: symbol exports and minor fixes ml-explore/mlx#3024
@hxu296 made their first contribution in Add NAX Split-K GEMM for large-K matmuls to improve performance ml-explore/mlx#3018
@jessegross made their first contribution in Better support consumer CUDA GPUs ml-explore/mlx#3056

Full Changelog: ml-explore/mlx@v0.30.3...v0.30.4

View the full release notes at https://github.com/ml-explore/mlx/releases/tag/v0.30.4.

---

[chenrui333]

  [100%] Linking CXX shared library libmlxc.dylib
  /opt/homebrew/opt/cmake/bin/cmake -E cmake_link_script CMakeFiles/mlxc.dir/link.txt --verbose=1
  Undefined symbols for architecture arm64:
    "mlx::core::metal::device_info()", referenced from:
        _mlx_metal_device_info in metal.cpp.o
  ld: symbol(s) not found for architecture arm64
  clang++: error: linker command failed with exit code 1 (use -v to see invocation)

[github-actions]

This pull request has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. To keep this pull request open, add a help wanted or in progress label.