Search results

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

Any plans to support <code>USE_DISTRIBUTED=0</code> pytorch?

Status: Open.

#3452 In pytorch/ao;

· rene-descartes2021 opened on Dec 4, 2025

Delete special handling of per tensor quant in <code>_choose_scale_float8</code>

good first issueGood for newcomersGood for newcomers

triaged

Status: Open.

#3414 In pytorch/ao;

· jerryzh168 opened on Dec 2, 2025

Errors when convert quantized model to onnx

triaged

Status: Open.

#3392 In pytorch/ao;

· qiuzhewei opened on Nov 26, 2025

Next steps for mxfp8 MoE training

moe

mx

tracker

Status: Open.

#3379 In pytorch/ao;

· danielvegamyhre opened on Nov 24, 2025