Skip to content

Pull requests: ml-explore/mlx

New pull request New
57 Open 2,369 Closed
57 Open 2,369 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

docs: document the TF32-class float32 default and MLX_ENABLE_TF32
#3894 opened Jul 23, 2026 by stoyoda0012-cyber Loading…
[CUDA] Fix grid overflow in gemm conv unfold kernels for >= 65,536 output positions
#3893 opened Jul 22, 2026 by AdamDLuz Loading…
Handle packed GGUF metadata without unaligned access
#3892 opened Jul 22, 2026 by markjouh Loading…
Avoid unaligned writes in CPU RandomBits
#3891 opened Jul 22, 2026 by markjouh Loading…
Warn once when float32 ops silently run at TF32 precision
#3883 opened Jul 21, 2026 by pierre427 Contributor Loading…
4
[CUDA][Improvement] RMSNorm backward
#3881 opened Jul 21, 2026 by nastya236 Collaborator • Draft
fix: reject non-canonical integer keys in tree_unflatten
#3878 opened Jul 21, 2026 by Solaris-star Loading…
3 of 4 tasks
1
3
[Metal] Avoid regex in custom kernel name generation
#3869 opened Jul 18, 2026 by aleroot Loading…
2 of 4 tasks
3
Don't reuse a Metal buffer while a command buffer is still using it
#3868 opened Jul 17, 2026 by shubhxho Contributor Loading…
Add inverse-CDF path for large categorical sampling
#3864 opened Jul 17, 2026 by MicroMilo Loading…
1
1
Add a BM=16 qmm tile for the small-M decode dead-zone
#3863 opened Jul 17, 2026 by katlun-lgtm Contributor Loading…
3
Add metadata to exported functions
#3855 opened Jul 16, 2026 by eyupcanakman Contributor Loading…
1
Guard NAX MetalPerformancePrimitives include behind __has_include
#3853 opened Jul 16, 2026 by metascroy Contributor Loading…
Metal Cholesky
#3851 opened Jul 16, 2026 by kapellirohith • Draft
1
Fix installed static MLX package on Windows
#3848 opened Jul 16, 2026 by yingjiacai Loading…
3 of 4 tasks
Fix log_cosh_loss docstring to document the element-wise loss
#3846 opened Jul 15, 2026 by winklemad Loading…
3 of 4 tasks
Fix conv_transpose maxBufferLength failures on Metal via tiled unfold
#3845 opened Jul 14, 2026 by eyupcanakman Contributor Loading…
3 of 4 tasks
Add a fused full-attention path for head_dim 256 on NAX devices
#3842 opened Jul 13, 2026 by wyanzhao Contributor Loading…
4 tasks done
1
Extend SDPA vector 2-pass kernel to query sequence lengths up to 16
#3838 opened Jul 13, 2026 by katlun-lgtm Contributor Loading…
3 of 4 tasks
2
Fix eigh UPLO and zero-size eigh/svd on the CPU
#3834 opened Jul 11, 2026 by kapellirohith Loading…
3 of 4 tasks
Fix custom metal kernel cache collision for same name, different source (#3832)
#3833 opened Jul 11, 2026 by katlun-lgtm Contributor Loading…
1
Fix Glorot/He uniform init docstrings to label the uniform bound, not sigma
#3831 opened Jul 10, 2026 by vineethsaivs Loading…
2 of 4 tasks
1
Fixes an axis mismatch bug in matrix norm for case -1 and 1
#3827 opened Jul 9, 2026 by danlee2002 Contributor Loading…
4 tasks done
3
Make index autodiff errors explicitly recommend stop_gradient
#3820 opened Jul 8, 2026 by dogukanveziroglu Contributor Loading…
5
Refactor wheel building script
#3818 opened Jul 8, 2026 by zcbenz Collaborator Loading…
ProTip! Updated in the last three days: updated:>2026-07-20.