Skip to content

float8: snapshot FSDP precomputed scale #4555

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
vkuzo merged 1 commit into from
Jul 6, 2026
+7 −2
Merged

float8: snapshot FSDP precomputed scale #4555

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
9 changes: 7 additions & 2 deletions torchao/float8/fsdp_utils.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -226,10 +226,15 @@ def __repr__(self):
return f"WeightWithDynamicFloat8CastTensor(tensor={self._tensor}, linear_mm_config={self._linear_mm_config}, dtype={self._dtype})"

def fsdp_pre_all_gather(self, mesh):
if self._precomputed_scale is not None:
# Snapshot once: `_precomputed_scale` may be re-assigned by
# `precompute_float8_dynamic_scale_for_fsdp` between iterations, and
# under free-threaded CPython we must not read the attribute twice in
# the if/use pair.
precomputed_scale = self._precomputed_scale
if precomputed_scale is not None:
float8_training_tensor = hp_tensor_and_scale_to_float8(
self._tensor,
self._precomputed_scale,
precomputed_scale,
self._dtype,
self._linear_mm_config,
GemmInputRole.WEIGHT,
Expand Down
Loading