Skip to content

fix(autogram): Reset even when exception is raised #417

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

Merged
ValerianRey merged 1 commit into from
Sep 11, 2025
Merged

fix(autogram): Reset even when exception is raised #417

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
18 changes: 11 additions & 7 deletions src/torchjd/autogram/_engine.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -169,11 +169,20 @@ def compute_gramian(self, output: Tensor) -> Tensor:
"""

reshaped_output = output.reshape([-1])
return self._compute_square_gramian(reshaped_output)

def _compute_square_gramian(self, output: Tensor) -> Tensor:
self._module_hook_manager.gramian_accumulation_phase = True

try:
square_gramian = self._compute_square_gramian(reshaped_output)
finally:
# Reset everything that has a state, even if the previous call raised an exception
self._module_hook_manager.gramian_accumulation_phase = False
self._gramian_accumulator.reset()
self._target_edges.reset()

return square_gramian

def _compute_square_gramian(self, output: Tensor) -> Tensor:
leaf_targets = list(self._target_edges.get_leaf_edges({get_gradient_edge(output)}))

def differentiation(_grad_output: Tensor) -> tuple[Tensor, ...]:
Expand All @@ -190,9 +199,4 @@ def differentiation(_grad_output: Tensor) -> tuple[Tensor, ...]:
# have failed. So gramian is necessarily a valid Tensor here.
gramian = cast(Tensor, self._gramian_accumulator.gramian)

# Reset everything that has a state
self._module_hook_manager.gramian_accumulation_phase = False
self._gramian_accumulator.reset()
self._target_edges.reset()

return gramian
Loading