Search results

Float8DynamicActivationInt4WeightConfig should not expose group_size

Status: Open.

#2763 In pytorch/ao;

· andrewor14 opened on Aug 13, 2025

Float8DynamicActivationFloat8WeightConfig should default to PerRow

Status: Open.

#2760 In pytorch/ao;

· andrewor14 opened on Aug 13, 2025

Migrating from AffineQuantizedTensor + Layouts to new structure of tensor subclasses

Status: Open.

#2752 In pytorch/ao;

· jerryzh168 opened on Aug 12, 2025

Deprecate and remove subclass.py, dynamic_quant.py, and weight_only.py

Status: Open.

#2745 In pytorch/ao;

· andrewor14 opened on Aug 12, 2025

Update notify path

Status: Open.

#2741 In pytorch/ao;

· drisspg opened on Aug 12, 2025

<code>torch_version_at_least</code> semantics are incorrect

good first issueGood for newcomersGood for newcomers

Status: Open.

#2722 In pytorch/ao;

· andrewor14 opened on Aug 8, 2025

[fp8 blockwise training] try using torch._scaled_mm instead of Triton kernels for fp8 gemms

float8

good first issueGood for newcomersGood for newcomers

Status: Open.

#2713 In pytorch/ao;

· danielvegamyhre opened on Aug 7, 2025

[not user facing] Split <code>implements</code> and <code>implements_torch_function</code>

good first issueGood for newcomersGood for newcomers

Status: Open.

#2707 In pytorch/ao;

· jerryzh168 opened on Aug 6, 2025

[moe training] fsdp2 bug for llama4 shared experts where num_experts=1

bugSomething isn't workingSomething isn't working

Status: Open.

#2673 In pytorch/ao;

· danielvegamyhre opened on Aug 4, 2025

Deprecation for Float8DynamicActivationFloat8WeightConfig and Float8WeightOnlyConfig and the models

Status: Open.

#2649 In pytorch/ao;

· jerryzh168 opened on Jul 31, 2025

Refresh "Quantization Overview" docs page

Status: Open.

#2643 In pytorch/ao;

· andrewor14 opened on Jul 30, 2025

Confusing behavior of <code>register_quantize_module_handler</code>

Status: Open.

#2640 In pytorch/ao;

· jerryzh168 opened on Jul 30, 2025