Skip to content
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

Fix the retention for existing gradients in the Grad Acc API #8658

Merged
merged 1 commit into from
Jan 31, 2025
Merged

Fix the retention for existing gradients in the Grad Acc API #8658

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
7 changes: 5 additions & 2 deletions torch_xla/experimental/gradient_accumulation.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -39,10 +39,13 @@ def gradient_accumulation(

Notes:

The model tracing will happen entirely within the loop. Hence, it is
* The model tracing will happen entirely within the loop. Hence, it is
assumed that `train_step` is purposefully encapsulated inside of the
loop. Hence, it is not recommended to have any operation involving the
model parameters outside of `train_step`.
* Note that zeroing the gradients to zero instead of None, (e.g.
`.zero_grad(set_to_none=False)) will avoid the device transfer of the
initial gradients in every call.

Args:
train_step: Training function that takes iterable tensors and carried
Expand Down Expand Up @@ -380,7 +383,7 @@ def body_fn(iteri: torch.Tensor, _: torch.Tensor,
for param in model_parameters:
if not param.requires_grad:
continue
if param.grad:
if param.grad is not None:
grad = param.grad
else:
grad = torch.zeros(param.size()).to(param.device).requires_grad_(False)
Expand Down