mlx 0.30.1

[BrewTestBot]

Created by brew bump

---

Created with brew bump-formula-pr.

• resource blocks have been checked for updates.

release notes

## Highlights
* RDMA over thunderbolt with the JACCL backend (macOS >= 26.2) ([some numbers](https://github.com/ml-explore/mlx/pull/2808))
* NAX with JIT so that they can be used in MLX Swift
* CUDA improvements
  * Many improvements to SDPA (masking, T_q != T_kv)
  * Faster quantize/dequantize
  * QQMM to make use of faster tensor cores
  * Fix in col reduce speeds up training
What's Changed

patch + fix docs build by @awni in patch + fix docs build ml-explore/mlx#2799
Fix macos release target and linux arm release by @awni in Fix macos release target and linux arm release ml-explore/mlx#2802
Fix cuda allocator copy condition by @awni in Fix cuda allocator copy condition ml-explore/mlx#2800
[CUDA] Partly fix random for large sizes by @awni in [CUDA] Partly fix random for large sizes ml-explore/mlx#2798
patch bump for future version by @awni in patch bump for future version ml-explore/mlx#2804
Centralize NAX condition by @awni in Centralize NAX condition ml-explore/mlx#2811
Tolerance for some ops tests on cuda by @awni in Tolerance for some ops tests on cuda ml-explore/mlx#2815
Fix typo: refs/head/main => refs/heads/main by @zcbenz in Fix typo: refs/head/main => refs/heads/main ml-explore/mlx#2818
Add float64 Eig and complex64 SVD/Eig support (Fixes #2708) by @harsh-sutariya in Add float64 Eig and complex64 SVD/Eig support (Fixes #2708) ml-explore/mlx#2737
Fix mx.core.load type annotation by @CC-Yeh in Fix mx.core.load type annotation ml-explore/mlx#2819
Force cudaGraphExec reinstantiation when clusters are used by @andportnoy in Force cudaGraphExec reinstantiation when clusters are used ml-explore/mlx#2813
Bump actions/checkout from 5 to 6 by @dependabot[bot] in Bump actions/checkout from 5 to 6 ml-explore/mlx#2828
Fix mx.core.linspace type annotation by @CC-Yeh in Fix mx.core.linspace type annotation ml-explore/mlx#2820
[CUDA] Exit on crash and more helpful errors by @awni in [CUDA] Exit on crash and more helpful errors ml-explore/mlx#2830
[CUDA] Add debug env to save cuda graphs to dot files by @zcbenz in [CUDA] Add debug env to save cuda graphs to dot files ml-explore/mlx#2825
[CUDA] Output of SDPA should have same layout with inputs by @zcbenz in [CUDA] Output of SDPA should have same layout with inputs ml-explore/mlx#2826
Merge build-cuda and build-linux actions by @zcbenz in Merge build-cuda and build-linux actions ml-explore/mlx#2783
[CUDA] Support array mask in SDPA by @zcbenz in [CUDA] Support array mask in SDPA ml-explore/mlx#2822
[CUDA] Faster rms norm for small dimension by @awni in [CUDA] Faster rms norm for small dimension ml-explore/mlx#2838
Added clarification to apply_fn parameter of apply_to_modules by @yuchaoran2011 in Added clarification to apply_fn parameter of apply_to_modules ml-explore/mlx#2831
[CUDA] Use cuDNN attention when T_q != T_kv by @zcbenz in [CUDA] Use cuDNN attention when T_q != T_kv ml-explore/mlx#2843
[CUDA] Migrate conv code to new cuDNN APIs by @zcbenz in [CUDA] Migrate conv code to new cuDNN APIs ml-explore/mlx#2847
Support more Numpy interfaces for masked_scatter by @CC-Yeh in Support more Numpy interfaces for masked_scatter ml-explore/mlx#2832
use thread local cpature mode by @awni in use thread local cpature mode ml-explore/mlx#2850
Fix export scatters by @awni in Fix export scatters ml-explore/mlx#2852
Reduce JVP by @awni in Reduce JVP ml-explore/mlx#2854
Fix graph updating by @awni in Fix graph updating ml-explore/mlx#2857
Fix init from double by @awni in Fix init from double ml-explore/mlx#2861
Update gumbel function signature parameters by @tianenchong in Update gumbel function signature parameters ml-explore/mlx#2868
Added support for pytree types that inherit from tuple and typing.namedtuple by @romanoneg in Added support for pytree types that inherit from tuple and typing.namedtuple ml-explore/mlx#2845
Layer norm throws on dimension mismatch by @awni in Layer norm throws on dimension mismatch ml-explore/mlx#2870
fix compile copying by @awni in fix compile copying ml-explore/mlx#2871
Do a PyPi release for cuda on arm by @awni in Do a PyPi release for cuda on arm ml-explore/mlx#2866
Add a 2-pass col reduce for CUDA by @angeloskath in Add a 2-pass col reduce for CUDA ml-explore/mlx#2863
[CUDA] Faster general copy by @awni in [CUDA] Faster general copy ml-explore/mlx#2873
[CUDA] Release build for cuda 13 by @awni in [CUDA] Release build for cuda 13 ml-explore/mlx#2872
Make allocator::malloc throw on allocation failure by @zcbenz in Make allocator::malloc throw on allocation failure ml-explore/mlx#2874
[Metal] No copy array init by @awni in [Metal] No copy array init ml-explore/mlx#2875
Try not to fail when there should be memory available by @awni in Try not to fail when there should be memory available ml-explore/mlx#2869
[CUDA] Enable more graphs to be updatable by @awni in [CUDA] Enable more graphs to be updatable ml-explore/mlx#2883
Fix docs: replace nonexistent mx.random.randn with mx.random.normal by @Satyam12singh in Fix docs: replace nonexistent mx.random.randn with mx.random.normal ml-explore/mlx#2890
Allow events in sub graph to be updatable by @awni in Allow events in sub graph to be updatable ml-explore/mlx#2886
bump minimum required Python version by @ngoldbaum in bump minimum required Python version ml-explore/mlx#2891
do not use simd neon intrinsics on x86 by @davidkoski in do not use simd neon intrinsics on x86 ml-explore/mlx#2893
Fix input buffer donation in compile by @CC-Yeh in Fix input buffer donation in compile ml-explore/mlx#2897
Update nanobind pin to most recent version by @ngoldbaum in Update nanobind pin to most recent version ml-explore/mlx#2896
fp quantize by @nastya236 in fp quantize ml-explore/mlx#2892
Fix grad in place updates by @awni in Fix grad in place updates ml-explore/mlx#2899
[CUDA] Add host nodes to subgraph types for graph update by @awni in [CUDA] Add host nodes to subgraph types for graph update ml-explore/mlx#2901
fix: possible heap-buffer-overflow in RandomBits::eval_cpu (follow for new ASAN CI tests) by @incertum in fix: possible heap-buffer-overflow in RandomBits::eval_cpu (follow for new ASAN CI tests) ml-explore/mlx#2877
Fix ccache getting disabled by @zcbenz in Fix ccache getting disabled ml-explore/mlx#2905
Fix attention for large sizes by @awni in Fix attention for large sizes ml-explore/mlx#2903
No VJP for mask or sinks in attention by @awni in No VJP for mask or sinks in attention ml-explore/mlx#2909
Bump actions/upload-artifact from 5 to 6 by @dependabot[bot] in Bump actions/upload-artifact from 5 to 6 ml-explore/mlx#2911
Bump actions/download-artifact from 6 to 7 by @dependabot[bot] in Bump actions/download-artifact from 6 to 7 ml-explore/mlx#2912
Use CUDA runtime headers from local python package by @zcbenz in Use CUDA runtime headers from local python package ml-explore/mlx#2906
DOC : Add compile state example by @Satyam12singh in DOC : Add compile state example ml-explore/mlx#2910
qqmm by @nastya236 in qqmm ml-explore/mlx#2789
Thunderbolt RDMA communications backend by @angeloskath in Thunderbolt RDMA communications backend ml-explore/mlx#2808
Add JIT support for NAX kernels by @jagrit06 in Add JIT support for NAX kernels ml-explore/mlx#2916
Fix warnings for the NAX build by @angeloskath in Fix warnings for the NAX build ml-explore/mlx#2921

New Contributors

@dependabot[bot] made their first contribution in Bump actions/checkout from 5 to 6 ml-explore/mlx#2828
@yuchaoran2011 made their first contribution in Added clarification to apply_fn parameter of apply_to_modules ml-explore/mlx#2831
@tianenchong made their first contribution in Update gumbel function signature parameters ml-explore/mlx#2868
@romanoneg made their first contribution in Added support for pytree types that inherit from tuple and typing.namedtuple ml-explore/mlx#2845
@Satyam12singh made their first contribution in Fix docs: replace nonexistent mx.random.randn with mx.random.normal ml-explore/mlx#2890
@ngoldbaum made their first contribution in bump minimum required Python version ml-explore/mlx#2891

Full Changelog: ml-explore/mlx@v0.30.0...v0.30.1

View the full release notes at https://github.com/ml-explore/mlx/releases/tag/v0.30.1.

---

[github-actions]

🤖 An automated task has requested bottles to be published to this PR.

Caution

Please do not push to this PR branch before the bottle commits have been pushed, as this results in a state that is difficult to recover from. If you need to resolve a merge conflict, please use a merge commit. Do not force-push to this PR branch.