Skip to content

return total grad norm in torchrec grad clipping #2507

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

Closed
wants to merge 1 commit into from
Closed

return total grad norm in torchrec grad clipping #2507

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
3 changes: 2 additions & 1 deletion torchrec/optim/clipping.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -136,7 +136,7 @@ def step(self, closure: Any = None) -> None:
self._step_num += 1

@torch.no_grad()
def clip_grad_norm_(self) -> None:
def clip_grad_norm_(self) -> Optional[Union[float, torch.Tensor]]:
"""Clip the gradient norm of all parameters."""
max_norm = self._max_gradient
norm_type = float(self._norm_type)
Expand Down Expand Up @@ -224,6 +224,7 @@ def clip_grad_norm_(self) -> None:
clip_coef = cast(torch.Tensor, max_norm / (total_grad_norm + 1e-6))
clip_coef_clamped = torch.clamp(clip_coef, max=1.0)
torch._foreach_mul_(all_grads, clip_coef_clamped)
return total_grad_norm


def _batch_cal_norm(
Expand Down
Loading