Skip to content

Fix weight dictionary lookup to handle nn.share_scope. #772

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
copybara-service wants to merge 1 commit into from
Closed

Fix weight dictionary lookup to handle nn.share_scope. #772

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
14 changes: 6 additions & 8 deletions gemma/gm/nn/gemma4/_layers.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -37,11 +37,10 @@ def __call__(self, eqn: str, x: jax.Array) -> jax.Array:
self.dtype if self.dtype is not None else None,
)
# Workaround for behavior with nn.share_scope in parent modules:
# self.param might return a dict {'w': tensor} instead of the bare tensor.
# The key appears to be the default class weight_name 'w'.
if isinstance(w, dict):
if 'w' in w:
w = w['w']
# self.param might return a dict {self.weight_name: tensor} instead of the
# bare tensor.
if isinstance(w, dict) and self.weight_name in w:
w = w[self.weight_name]
if self.w_scale is not None:
w *= self.w_scale
return jnp.einsum(eqn, x, w)
Expand All @@ -64,9 +63,8 @@ def __call__(self, eqn: str, x: jax.Array) -> jax.Array:
self.shape,
self.dtype if self.dtype is not None else None,
)
if isinstance(w, dict):
if 'w' in w:
w = w['w']
if isinstance(w, dict) and self.weight_name in w:
w = w[self.weight_name]
if self.w_scale is not None:
w *= self.w_scale

Expand Down
Loading