Search results

Future plans for MXFP8 development

Status: Open.

#3333 In pytorch/ao;

· avizon-aws opened on Nov 12, 2025

KV Cache quantization

Status: Open.

#3330 In pytorch/ao;

· cjm715 opened on Nov 12, 2025

Align <code>_choose_scale_float8</code>'s use of <code>block_size</code> with other code

good first issueGood for newcomersGood for newcomers

Status: Open.

#3324 In pytorch/ao;

· jerryzh168 opened on Nov 11, 2025

Loading 8bit optimizer state from checkpoint causes dtype mismatch

Status: Open.

#3314 In pytorch/ao;

· yz-ppl opened on Nov 8, 2025

[MXFP8 MoE] What's the expected inference solution on H100s, after training with TorchAO MXFP8 MoE?

moe

mx

question

Status: Open.

#3305 In pytorch/ao;

· goldhuang opened on Nov 6, 2025

AttributeError: 'TorchAoConfig' object has no attribute 'include_input_output_embeddings'

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.

Status: Open.

#3301 In pytorch/ao;

· TumCCC opened on Nov 6, 2025

Examples of using llms with PT2E workflow?

Status: Open.

#3295 In pytorch/ao;

· cjm715 opened on Nov 5, 2025

NVFP4 Training Tracker

moe

mx

tracker

training

Status: Open.

#3293 In pytorch/ao;

· danielvegamyhre opened on Nov 5, 2025

FP8 Blockwise Training Tracker

float8

good first issueGood for newcomersGood for newcomers

moe

training

Status: Open.

#3290 In pytorch/ao;

· danielvegamyhre opened on Nov 4, 2025

Unify the torchao evaluation framework

Status: Open.

#3289 In pytorch/ao;

· jainapurva opened on Nov 4, 2025

float8 inference weight-only quant should map to a fused kernel or explain why not

Status: Open.

#3288 In pytorch/ao;

· vkuzo opened on Nov 4, 2025

TorchAO multi-gpu support tracker

tracker

Status: Open.

#3284 In pytorch/ao;

· jcaip opened on Nov 3, 2025