All issues

Search results

matmul_fp8_block: trailing k scale block dropped for many K, results off by a per-block scale

Status: Open.

#6159 In pytorch/FBGEMM;

· truong-v opened on Aug 14, 2026

scale_fp8_row: unmasked column-scale load reads past the end of w_scale

Status: Open.

#6157 In pytorch/FBGEMM;

· truong-v opened on Aug 14, 2026

[Question] Would you be open to mentioning fbgemm-ascend in the README?

Status: Open.

#6149 In pytorch/FBGEMM;

· lihao7212148 opened on Aug 13, 2026

When can we have a torch 2.13 based and py312 based package for both cpu and gpu?

Status: Open.

#6107 In pytorch/FBGEMM;

· purifier1990 opened on Aug 4, 2026

[MSVC] Different linkage

Status: Open.

#6030 In pytorch/FBGEMM;

· SunBlack opened on Jul 20, 2026

_fbgemm_torch_compat_table is outdated and emits false compatibility warnings

Status: Open.

#5953 In pytorch/FBGEMM;

· bwnotfound opened on Jun 25, 2026

AttributeError: '_OpNamespace' 'fbgemm' object has no attribute 'get_bucket_sorted_indices_and_bucket_tensor'

Status: Open.

#5666 In pytorch/FBGEMM;

· YINUO1221 opened on Apr 21, 2026

FBGEMM-GPU v1.5.0 Cannot Run on SM70 and SM75 Architectures

Status: Open.

#5481 In pytorch/FBGEMM;

· tiankongdeguiji opened on Mar 16, 2026

Incorrect embedding_lookup results (zeroed out) when embedding_dim > 1024, despite FBGEMM limit being 2048

Status: Open.

#5326 In pytorch/FBGEMM;

· tiankongdeguiji opened on Jan 20, 2026

CUDA 13 builds not pushed to pypi index

Status: Open.

#5321 In pytorch/FBGEMM;

· winglian opened on Jan 16, 2026

vllm serve pytorch/gemma-3-12b-it-FP8 failed with fbgemm-gpu-genai 1.4.2 on B200

Status: Open.

#5221 In pytorch/FBGEMM;

· huydhn opened on Dec 13, 2025

How to apply gradient clip in fused optimizer?

Status: Open.

#5176 In pytorch/FBGEMM;

· acmilannesta opened on Nov 28, 2025