All issues

Search results

finegrained-fp8: fused MX MoE kernels write past intermediate and read past hidden/intermediate; host_descriptor mode never compiles

Status: Open.

#1158 In huggingface/kernels-community;

· malaiwah opened on Sep 16, 2026

finegrained-fp8: grouped matmul writes past N and reads past N/K (wrong results on ordinary expert shapes)

Status: Open.

#1153 In huggingface/kernels-community;

· malaiwah opened on Sep 16, 2026

[Kernel Contribution] AMD Strix Halo (gfx1151 / RDNA 3.5) ROCm Kernels

Status: Open.

#1149 In huggingface/kernels-community;

· AzeezIsh opened on Sep 15, 2026

Upgrade kernel-bot to a real GitHub App

Status: Open.

#1127 In huggingface/kernels-community;

· moon-bot-app opened on Aug 31, 2026

Benchmark important kernels and report results in the system card (/kernel-bot benchmark)

Status: Open.

#1111 In huggingface/kernels-community;

· moon-bot-app opened on Aug 27, 2026

flash-attn2 v3 (stable ABI): backward fails for GQA/MQA

Status: Open.

#1085 In huggingface/kernels-community;

· qgallouedec opened on Aug 18, 2026

flash-attn2: no build variant for torch 2.13

Status: Open.

#1082 In huggingface/kernels-community;

· qgallouedec opened on Aug 14, 2026

Add Sol-Attn sparse attention kernel

Status: Open.

#1064 In huggingface/kernels-community;

· hp-l33 opened on Aug 6, 2026

mamba-ssm: replace deprecated torch.get_autocast_gpu_dtype() calls

Status: Open.

#1052 In huggingface/kernels-community;

· albertvillanova opened on Jul 31, 2026

sgl-flash-attn3: build against the Torch stable ABI

Status: Open.

#1043 In huggingface/kernels-community;

· moon-bot-app opened on Jul 28, 2026

Propagate FA3 build fix: rename #include-d .cu files to .cuh to avoid double-compilation

Status: Open.

#1032 In huggingface/kernels-community;

· moon-bot-app opened on Jul 16, 2026

Improve API-change check: gate on version bump and allow additive-only changes

Status: Open.

#1030 In huggingface/kernels-community;

· moon-bot-app opened on Jul 16, 2026