Skip to content

Pull requests: ml-explore/mlx

New pull request New
51 Open 2,188 Closed
51 Open 2,188 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

Fix use-after-free when a custom Metal kernel is called with different dtypes in one graph
#3662 opened Jun 11, 2026 by discobot Loading…
4 tasks done
1
Fused SDPA full-attention for head_dim=192/256 (revival of #3293)
#3660 opened Jun 11, 2026 by hojin12312 Loading…
Ignore numpy matmul warnings in test_blas
#3657 opened Jun 11, 2026 by zcbenz Collaborator Loading…
1
Fix multi-wire recv prefill deadlock in jaccl ring backend
#3654 opened Jun 10, 2026 by jasonpaulso Loading…
2 of 4 tasks
1
[fix] allow exporting custom Metal kernels without a Metal backend
#3650 opened Jun 10, 2026 by xthomaswang Loading…
4 tasks done
1
fix(shapeless): infer dynamic dim in Unflatten::output_shapes (#2607)
#3649 opened Jun 9, 2026 by nac7 Loading…
1
1
Fix CPU gather transposing column-contiguous slices
#3647 opened Jun 9, 2026 by tillahoffmann Contributor Loading…
1
Add stft and istft to mlx.core.fft
#3639 opened Jun 8, 2026 by eyupcanakman Contributor Loading…
4 tasks done
2
Enable fused SDPA vector kernel for asymmetric Q/V head dims (192, 128)
#3637 opened Jun 6, 2026 by yohann-bearzi Loading…
3
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
1
Fix JVPs of select, arctan2, masked_scatter, and bitwise ops
#3633 opened Jun 5, 2026 by qflen Contributor Loading…
4 tasks done
1
1
Fix gather_qmm NAX kernel name mismatch
#3632 opened Jun 5, 2026 by scyyh11 Loading…
4 tasks done
1
Fix int16 overflow in NAX qmm edge-tile bounds
#3631 opened Jun 5, 2026 by scyyh11 Loading…
4 tasks done
5
Fix intermittent wrong bias gradient in fast::layer_norm VJP (Metal WAR hazard)
#3630 opened Jun 5, 2026 by tillahoffmann Contributor Loading…
4
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
7
ProTip! Follow long discussions with comments:>50.