Skip to content

Prevent FLCE backward when reduction="none" #1126

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

Draft
nick-knight wants to merge 1 commit into
base: main
Choose a base branch
from
+2 −5
Draft

Prevent FLCE backward when reduction="none" #1126

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: 2 additions & 5 deletions src/liger_kernel/ops/fused_linear_cross_entropy.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -219,11 +219,6 @@ def fused_linear_cross_entropy_forward(
alpha=1.0,
)

# Need extra calculations for backward if reduction=='none'. Not supporting reduction='none' now.
# if reduction == "none":
# loss = loss_1d
# z_loss = z_loss_1d if return_z_loss else None

if reduction == "none":
# Return per-token losses
loss = loss_1d
Expand All @@ -245,6 +240,8 @@ def fused_linear_cross_entropy_forward(


def fused_linear_cross_entropy_backward(grad_output, grad_input, grad_weight, grad_bias):
# Need extra calculations for backward if reduction=='none'. Not supporting reduction='none' now.
assert grad_output.ndim == 0, 'Backward unsupported for reduction="none"'
# If cross entropy is the last layer, grad_output is 1.0. Skip the mul to save time
if not torch.equal(grad_output, torch.tensor(1.0, device=grad_output.device)):
# We use a Triton kernel instead of a PyTorch operation because modifying inputs in-place
Expand Down