Search results

Optimize CPU masked_scatter for contiguous inputs

Status: Open.

#3669 In ml-explore/mlx;

· AK-Khan02 opened on Jun 13, 2026

[BUG] VJP mx.maximum(rho, eps) is NaN at rho=0 even though the forward value is finite

Status: Open.

#3668 In ml-explore/mlx;

· kyrollosyanny opened on Jun 13, 2026

[BUG] MLX doesn't publish iOS-compatible wheels

enhancement

python

Status: Open.

#3665 In ml-explore/mlx;

· freakboy3742 opened on Jun 12, 2026

[BUG] [CUDA] rope cudaGraphAddKernelNode

bug

cuda

Status: Open.

#3659 In ml-explore/mlx;

· nastya236 opened on Jun 11, 2026

head_dim=256 unfused SDPA makes 100K+ context prefill impractical — real-world evidence for reviving #3293

Status: Open.

#3658 In ml-explore/mlx;

· hojin12312 opened on Jun 11, 2026

m5 max 是不是quantized_matmul 这个问题

Status: Open.

#3656 In ml-explore/mlx;

· YorgeJin opened on Jun 11, 2026

[BUG][CUDA] Segmentation Fault with CUDA on Jetson/Tegra

bug

cuda

Status: Open.

#3645 In ml-explore/mlx;

· gmondada opened on Jun 9, 2026

[BUG] JVPs of power, divmod, and slice_update mishandle partly traced inputs

bug

Status: Open.

#3634 In ml-explore/mlx;

· qflen opened on Jun 6, 2026

[BUG] JVP non-deterministic crashes when called in a loop.

bug

Status: Open.

#3629 In ml-explore/mlx;

· kyrollosyanny opened on Jun 4, 2026

[BUG] JVP silently returns a zero tangent when differentiating through mx.where

bug

Status: Open.

#3627 In ml-explore/mlx;

· kyrollosyanny opened on Jun 4, 2026

[Performance] conv_general 5D (3D convolution) is 2-5x slower than equivalent per-frame 2D conv

performance

Status: Open.

#3625 In ml-explore/mlx;

· kimjj81 opened on Jun 4, 2026

[BUG] gemma-4-12b with omlx error loading model

Status: Open.

#3623 In ml-explore/mlx;

· benb0jangles opened on Jun 4, 2026