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 axis should be int or None in softmax_cross_entropy_with_integer_labels #1164

Merged
merged 1 commit into from
Jan 6, 2025
Merged

Fix axis should be int or None in softmax_cross_entropy_with_integer_labels #1164

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 optax/losses/_classification.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -273,7 +273,7 @@ def softmax_cross_entropy(
def softmax_cross_entropy_with_integer_labels(
logits: chex.Array,
labels: chex.Array,
axis: Union[int, tuple[int, ...], None] = -1,
axis: Union[int, None] = -1,
where: Union[chex.Array, None] = None,
) -> chex.Array:
r"""Computes softmax cross entropy between the logits and integer labels.
Expand All @@ -297,7 +297,7 @@ def softmax_cross_entropy_with_integer_labels(
labels: Integers specifying the correct class for each input, with shape
``[batch_size]``. Class labels are assumed to be between 0 and
``num_classes - 1`` inclusive.
axis: Axis or axes along which to compute.
axis: Axis along which to compute.
where: Elements to include in the computation.

Returns:
Expand Down Expand Up @@ -329,6 +329,9 @@ def softmax_cross_entropy_with_integer_labels(
"""
chex.assert_type([logits], float)
chex.assert_type([labels], int)
if axis is not None and not isinstance(axis, int):
raise ValueError(f'axis = {axis} is unsupported. Provide an int or None.')

# This is like jnp.take_along_axis(jax.nn.log_softmax(...), ...) except that
# we avoid subtracting the normalizer from all values, just from the values
# for the correct labels.
Expand Down
Loading