Skip to content

Pull requests: ml-explore/mlx

New pull request New
47 Open 2,174 Closed
47 Open 2,174 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

Enable fused SDPA vector kernel for asymmetric Q/V head dims (192, 128)
#3637 opened Jun 6, 2026 by yohann-bearzi Loading…
2
Fix JVPs of power, divmod, and slice_update with partly traced inputs
#3636 opened Jun 6, 2026 by qflen Contributor Loading…
4 tasks done
1
Roll back compile cache entry when the first trace throws
#3635 opened Jun 6, 2026 by adityasingh2400 Contributor Loading…
3 of 4 tasks
1
2
Fix JVPs of select, arctan2, masked_scatter, and bitwise ops
#3633 opened Jun 5, 2026 by qflen Contributor Loading…
4 tasks done
1
Fix gather_qmm NAX kernel name mismatch
#3632 opened Jun 5, 2026 by scyyh11 Loading…
4 tasks done
Fix int16 overflow in NAX qmm edge-tile bounds
#3631 opened Jun 5, 2026 by scyyh11 Loading…
4 tasks done
Fix intermittent wrong bias gradient in fast::layer_norm VJP (Metal WAR hazard)
#3630 opened Jun 5, 2026 by tillahoffmann Loading…
NAX requires setting MACOSX_DEPLOYMENT_TARGET=26.2
#3622 opened Jun 4, 2026 by zcbenz Collaborator Loading…
1
2
Fix flaky test_siblings_without_eval
#3621 opened Jun 4, 2026 by zcbenz Collaborator Loading…
Return zero VJPs for scatter indices
#3612 opened Jun 1, 2026 by devin-lai Loading…
1
1
Add reflect and symmetric padding modes to mx.pad
#3608 opened May 30, 2026 by katlun-lgtm Loading…
1
Add VJP for cumulative max and min scans
#3599 opened May 28, 2026 by devin-lai Loading…
3
Add metal::set_metallib_path() to override the metallib search path
#3597 opened May 28, 2026 by olilarkin Loading…
3
Cache JIT-compiled CUDA kernels
#3587 opened May 24, 2026 by zcbenz Collaborator Loading…
Implement batched matmul for large 1D dot products
#3580 opened May 22, 2026 by Ved235 Loading…
2 of 4 tasks
8
Add new_thread_unsafe_stream API
#3578 opened May 22, 2026 by zcbenz Collaborator Loading…
[CUDA] JIT-compile qmm_naive
#3576 opened May 21, 2026 by zcbenz Collaborator Loading…
Fix CPU GatherMM one-row NaNs
#3563 opened May 19, 2026 by TyceHerrman Loading…
1
6
feat(fast): add sliding-window SDPA kernel path
#3552 opened May 15, 2026 by rabbitson87 Loading…
9
Add Metal DLPack zero-copy sharing
#3531 opened May 11, 2026 by XXXXRT666 Contributor Loading…
4 tasks done
140
Catch error in CommandBuffer and poison the events
#3523 opened May 11, 2026 by zcbenz Collaborator Loading…
1
10
[CPU - Linux] AVX SIMD backend for fp16 and bf16 matmul
#3502 opened May 9, 2026 by acsweet Loading…
3 of 4 tasks
20
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
Reduce peak memory of large half-precision random.uniform/normal
#3439 opened Apr 22, 2026 by dogukanveziroglu Contributor • Draft
4 tasks done
5
fsdp: improve state saving determinism and tree reconstruction
#3407 opened Apr 13, 2026 by coder-jayp Loading…
1 of 4 tasks
3
ProTip! Follow long discussions with comments:>50.