Search results

mx.compile: assigning an elementwise expression to a negative-strided slice writes only one element

Status: Open.

#3716 In ml-explore/mlx;

· cetagostini opened on Jun 18, 2026

[BUG] jvp of mx.logcumsumexp is not implemented

enhancement

Status: Open.

#3710 In ml-explore/mlx;

· obchain opened on Jun 17, 2026

[BUG] mx.logsumexp jvp does not reduce along the axis (wrong shape and values)

bug

Status: Open.

#3707 In ml-explore/mlx;

· obchain opened on Jun 17, 2026

A19 (iPhone 17 Pro) GPU returns numerically wrong results in float32 workloads — deterministic HF corruption in a Demucs separation pipeline; M2/M3 clean

Status: Open.

#3702 In ml-explore/mlx;

· IARFLOW opened on Jun 16, 2026

[BUG] int64/uint64 put_along_axis / scatter_add_axis crashes the Metal JIT build

Status: Open.

#3690 In ml-explore/mlx;

· programVeins opened on Jun 15, 2026

Use-after-free under memory pressure: buffer-cache trim frees an MTLBuffer still used by an in-flight command buffer (kIOGPUCommandBufferCallbackErrorInvalidResource)

Status: Open.

#3689 In ml-explore/mlx;

· spokvulcan opened on Jun 15, 2026

Optimize CPU masked_scatter for contiguous inputs

enhancement

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

bug

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

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