Skip to content

Pull requests: ml-explore/mlx

New pull request New
41 Open 2,098 Closed
41 Open 2,098 Closed
Author
Filter by author
Loading
Label
Filter by label
Loading
Use alt + click/return to exclude labels
or ⇧ + click/return for logical OR
Projects
Filter by project
Loading
Milestones
Filter by milestone
Loading
Reviews
Filter by reviews
No reviews Review required Approved review Changes requested
Assignee
Filter by who’s assigned
Assigned to nobody Loading
Sort
Sort by
Newest Oldest Most commented Least commented Recently updated Least recently updated Best match
Most reactions
👍 👎 😄 🎉 😕 ❤️ 🚀 👀

Pull requests list

Use uv in macOS CI
#3491 opened May 7, 2026 by zcbenz Collaborator Loading…
Accumulate CPU half-precision sums in float32
#3488 opened May 6, 2026 by sofinvalery Loading…
4 tasks done
1
2
Make device_count() return 0 when there is no GPU
#3486 opened May 6, 2026 by zcbenz Collaborator Loading…
1
1
feat(compile): CustomKernel and GatherQMM implement output_shapes for shapeless compile
#3485 opened May 5, 2026 by dexwritescode Loading…
8
[Python] Fix mx.array DLPack dispatch
#3476 opened May 4, 2026 by apstenku123 Loading…
9
Compute contiguity from the actual occupied data
#3475 opened May 3, 2026 by sofinvalery Loading…
4 tasks done
1
Add MLX_NUMERICAL_STRICT_MODE for shape-independent quantized_matmul
#3473 opened May 2, 2026 by rakshith48 Loading…
1
Fix JACCL GID selection on Apple Thunderbolt RDMA (errno 22 RTR fail)
#3468 opened Apr 29, 2026 by danielkristofik Loading…
3 of 4 tasks
1
Do not use prebuilt cpu compile preamble when headers are installed
#3463 opened Apr 28, 2026 by zcbenz Collaborator Loading…
2
Add MLX_SDPA_BLOCKS env var for 2-pass vector kernel block-count override
#3455 opened Apr 26, 2026 by adurham Loading…
7
[jaccl] Fix race on local_staging in MeshImpl::all_reduce
#3451 opened Apr 25, 2026 by kernelpool Contributor Loading…
3 of 4 tasks
1
Handle non-multiple-of-8 spatial dims in depthwise conv2d Metal path
#3446 opened Apr 24, 2026 by qflen Contributor Loading…
4 tasks done
1
[CUDA] Separate main loop into a function in qmm
#3443 opened Apr 23, 2026 by zcbenz Collaborator Loading…
Reduce peak memory of large half-precision random.uniform/normal
#3439 opened Apr 22, 2026 by dogukanveziroglu Contributor • Draft
4 tasks done
5
fix: avoid clearing in-flight pipeline states in custom kernel cache
#3434 opened Apr 21, 2026 by Ziqiao-git Contributor Loading…
2
fsdp: improve state saving determinism and tree reconstruction
#3407 opened Apr 13, 2026 by coder-jayp Loading…
1 of 4 tasks
3
Fix SDPA vmap with GQA/MQA shapes (n_heads != n_kv_heads)
#3385 opened Apr 8, 2026 by Brooooooklyn Contributor Loading…
4 tasks done
1
Add basic fp8/bfp8 support, future implementations of 8-bit floating …
#3374 opened Apr 4, 2026 by Geramy Loading…
3 tasks done
2
Add TurboQuant KV cache compression with native Metal SDPA kernel
#3328 opened Mar 28, 2026 by arozanov Loading…
4 tasks done
5
[Metal] Fused Flash Attention backward (VJP) kernels
#3241 opened Mar 11, 2026 by Brooooooklyn Contributor Loading…
1
Fix shapeless compile eliding reductions on size-1 dimensions
#3202 opened Mar 4, 2026 by ghstrider Loading…
1
2
Add bessel_i0e and bessel_i1e ops
#3193 opened Mar 3, 2026 by robert-johansson Contributor Loading…
2 of 3 tasks
Fix command buffer memory tracking to use bytes instead of elements
#3192 opened Mar 3, 2026 by hxu296 Contributor Loading…
3 of 4 tasks
2
Add lgamma and digamma ops
#3181 opened Feb 27, 2026 by robert-johansson Contributor Loading…
3 of 4 tasks
1
Add all_to_all collective primitive
#3164 opened Feb 24, 2026 by 0xDaizz Loading…
2 of 4 tasks
1
ProTip! Updated in the last three days: updated:>2026-05-04.