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

Jump to bottom

Bugfix: Initialize grad_input, grad_weight to None #68

Merged
merged 2 commits into from
Mar 6, 2024
Merged

Bugfix: Initialize grad_input, grad_weight to None #68

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
ca37aa1
initialize grad_input to None
adityaranjan Mar 6, 2024
a484008
minor
adityaranjan Mar 6, 2024
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
4 changes: 4 additions & 0 deletions axonn/intra_layer/fully_connected.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -106,6 +106,8 @@ def backward(ctx, grad_output):
if dist.get_world_size(ctx.backward_all_reduce_group) > 1 or (
not overlap_reduce_scatter
):
grad_input, grad_weight = None, None

if ctx.needs_input_grad[0]:
grad_input = grad_output.matmul(weight)
handle = dist.all_reduce(
Expand Down Expand Up @@ -135,6 +137,8 @@ def backward(ctx, grad_output):
grad_weight = None # weight gradients are not ready yet
return grad_input, grad_weight, None, None, None, None, None, None, None
else:
grad_input, grad_weight = None, None

if ctx.needs_input_grad[1]:
grad_weight = (
grad_output.reshape(-1, grad_output.shape[-1])
Expand Down
Loading