All issues

Search results

[BUG] SinusoidalPositionalEncoding: explicit scale=0.0 silently overridden by default

Status: Open.

#4097 In ml-explore/mlx;

· YassineMA03 opened on Aug 9, 2026

[BUG] mx.isinf on complex64 ignores the imaginary part

Status: Open.

#4091 In ml-explore/mlx;

· JasonHonKL opened on Aug 9, 2026

sdpa_vector_2pass_1 loads K/V once per query head instead of once per kv head

Status: Open.

#4076 In ml-explore/mlx;

· dudududukim opened on Aug 8, 2026

[BUG] mx.any returns True for all-zero arrays that mix +0.0 and -0.0

bug

Status: Open.

#4075 In ml-explore/mlx;

· JasonHonKL opened on Aug 8, 2026

[BUG] mx.distributed.all_max / all_min silently drop NaN (inconsistent with mx.max/mx.min)

distributed

Status: Open.

#4072 In ml-explore/mlx;

· JasonHonKL opened on Aug 8, 2026

Metal: set_vector_bytes passes nullptr to setBytes for empty vectors

bug

Status: Open.

#4061 In ml-explore/mlx;

· JamesPriceZV opened on Aug 8, 2026

[BUG] mx.hadamard_transform` crashes on GPU for the documented sizes n ∈ {12, 20, 28}

Status: Open.

#4049 In ml-explore/mlx;

· JasonHonKL opened on Aug 7, 2026

[BUG] mx.cummax / mx.cummin mishandle NaN: device-dependent (CPU ≠ GPU), leak ±∞, and disagree with NumPy

bug

Status: Open.

#4044 In ml-explore/mlx;

· JasonHonKL opened on Aug 7, 2026

Fast quantized GEMV for K dimensions with a non-512 tail

Status: Open.

#4041 In ml-explore/mlx;

· justinchuby opened on Aug 6, 2026

[BUG] astype float→integer overflow is device-dependent: GPU saturates, CPU matches NumPy

Status: Open.

#4038 In ml-explore/mlx;

· JasonHonKL opened on Aug 6, 2026

Returning a tuple instead of list for meshgrid to make it compliant with Array API

Status: Open.

#4034 In ml-explore/mlx;

· aaishwarymishra opened on Aug 6, 2026

[BUG] Metal 2-pass vector SDPA kernel produces garbage decode output on M3 Max (applegpu_g15s) with GQA once KV length reaches 1024

Status: Open.

#4022 In ml-explore/mlx;

· beyondsimulations opened on Aug 5, 2026