Search results

int4 weight float8 rowwise activation workflow doesn't work with HF accelerate

Status: Open.

#3550 In pytorch/ao;

· vkuzo opened on Dec 26, 2025

mxfp8 inference does not work with vllm bench throughput

Status: Open.

#3549 In pytorch/ao;

· vkuzo opened on Dec 26, 2025

Support gpt-oss mxfp4 format qat

Status: Open.

#3547 In pytorch/ao;

· shelterwff-byte opened on Dec 26, 2025

[MXLinear]Where is the operator call for implementing MXFP8 in NVD?

Status: Open.

#3543 In pytorch/ao;

· LucaHW opened on Dec 25, 2025

Making torchao ABI compatible and moving closer to python only

tracker

Status: Open.

#3516 In pytorch/ao;

· jerryzh168 opened on Dec 18, 2025

vLLM fails to load TorchAO Int4Opaque weights due to unsupported aten.slice on Int4OpaqueTensor

Status: Open.

#3499 In pytorch/ao;

· Priyjain-amd opened on Dec 17, 2025

W4A16-INT (TinyGEMM+HQQ) is 6-13x slower than FP16 for batched inference

Status: Open.

#3496 In pytorch/ao;

· namgyu-youn opened on Dec 16, 2025

Cannot set version_counter for inference tensor

bugSomething isn't workingSomething isn't working

float8

quantize_quantize_ APIquantize_ API

Status: Open.

#3487 In pytorch/ao;

· Stonepia opened on Dec 15, 2025

per tensor symmetric activation quantization

Status: Open.

#3469 In pytorch/ao;

· jivercx opened on Dec 9, 2025

[bug] PerGroup(1) will throw RuntimeError

Status: Open.

#3458 In pytorch/ao;

· Freed-Wu opened on Dec 6, 2025

safetensors support for diffusers

0.16

integrationIssues related to integrations with other libraries, like huggingface, vllm, sglang, gemlite etc.Issues related to integrations with other libraries, like huggingface, vllm, sglang, gemlite etc.

triaged

Status: Open.

#3456 In pytorch/ao;

· jerryzh168 opened on Dec 6, 2025

Int4WeightOnly torch.bmm semantics

triaged

Status: Open.

#3436 In pytorch/ao;

· jcaip opened on Dec 4, 2025