Search results

torch.export.export a fp8 quantized torch.nn.Linear failed

Status: Open.

#3605 In pytorch/ao;

· lanluo-nvidia opened on Jan 8, 2026

MoE MXFP8 Incompatibility with FusedAdam (llegal memory access error)

moe

mx

Status: Open.

#3601 In pytorch/ao;

· yzhautouskay opened on Jan 8, 2026

MXFP8 MoE training prototype graduation tracker

tracker

Status: Open.

#3599 In pytorch/ao;

· danielvegamyhre opened on Jan 8, 2026

Support emulated mode for mxfp8 moe training to support non-sm100 CI or dev env

moe

mx

Status: Open.

#3598 In pytorch/ao;

· danielvegamyhre opened on Jan 8, 2026

Deprecation for version=1 Int8WeightOnlyConfig and Int8DynamicActivationInt8WeightConfig

Status: Open.

#3596 In pytorch/ao;

· jcaip opened on Jan 7, 2026

Float8DynamicActivationFloat8SemiSparseWeightConfig is being deprecated

Status: Open.

#3594 In pytorch/ao;

· jcaip opened on Jan 7, 2026

Ops supporting FP8 quantization (static quantization / Float8StaticActivationFloat8WeightConfig)

Status: Open.

#3593 In pytorch/ao;

· Alexey234432 opened on Jan 7, 2026

[RFC] XPU Enabling Plan for TorchAO

tracker

xpuIntel XPU related featuresIntel XPU related features

Status: Open.

#3576 In pytorch/ao;

· liangan1 opened on Jan 6, 2026

<code>Int4WeightFakeQuantizer</code> missing <code>enabled</code> attribute

Status: Open.

#3573 In pytorch/ao;

· salmanmohammadi opened on Jan 5, 2026

<code>group_size</code> hardcoded in <code>_infer_fake_quantize_configs</code>.

Status: Open.

#3572 In pytorch/ao;

· salmanmohammadi opened on Jan 5, 2026

Acc mismatch in tutorials

Status: Open.

#3557 In pytorch/ao;

· xiguadong opened on Dec 30, 2025

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

Status: Open.

#3550 In pytorch/ao;

· vkuzo opened on Dec 26, 2025