List view

Enable blockwise FP8 dense training kernels on ROCm

CLA SignedThis label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed.This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed.

float8

topic: rocm

Status: Open (in progress).

pytorch/aonumber 4036

#4036 In pytorch/ao;

· brucechanglongxu opened on Mar 10, 2026

Blockwise FP8 training seems to have much lower peak memory usage than others

float8

module: trainingquantize_ api training flowquantize_ api training flow

Status: Open.

#4101 In pytorch/ao;

· CFC87 opened on Mar 17, 2026

FP8 Blockwise Training Tracker

float8

good first issueGood for newcomersGood for newcomers

module: trainingquantize_ api training flowquantize_ api training flow

moe

tracker

Status: Open.

#3290 In pytorch/ao;

· danielvegamyhre opened on Nov 4, 2025

Add blockwise FP8 in roofline_utils and seperate benchmark script

CLA SignedThis label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed.This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed.

module: trainingquantize_ api training flowquantize_ api training flow

Status: Open (in progress).

pytorch/aonumber 4344

#4344 In pytorch/ao;

· iamzainhuda opened on Apr 27, 2026