Skip to content

Update keras3 Softmax mask handling to be more numerically robust. #21850

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
merged 2 commits into from
Nov 14, 2025
+9 −4
Merged

Update keras3 Softmax mask handling to be more numerically robust. #21850

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
2823496
Update keras3 Softmax mask handling to be more numerically robust.
yangliyl-g Nov 13, 2025
6c87c58
Fix formatting
yangliyl-g Nov 14, 2025
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
13 changes: 9 additions & 4 deletions keras/src/layers/activations/softmax.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -52,10 +52,15 @@ def __init__(self, axis=-1, **kwargs):

def call(self, inputs, mask=None):
if mask is not None:
adder = (
1.0 - backend.cast(mask, inputs.dtype)
) * _large_negative_number(inputs.dtype)
inputs += adder
# We keep the positions where the mask is True or > 0.5, and set the
# other (masked) positions to -1e.9.
Copy link
Contributor

@gemini-code-assist gemini-code-assist bot Nov 14, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

medium

There's a small typo in this comment. -1e.9 appears to be a typo for -1e9. To improve clarity and accuracy, especially since _large_negative_number can return different values based on dtype, I suggest making the comment more general.

Suggested change
# other (masked) positions to -1e.9.
# other (masked) positions to a large negative number.

if backend.standardize_dtype(mask.dtype) != "bool":
mask = backend.numpy.greater(
mask, backend.cast(0.5, dtype=mask.dtype)
)
inputs = backend.numpy.where(
mask, inputs, _large_negative_number(inputs.dtype)
)
if isinstance(self.axis, (tuple, list)):
if len(self.axis) > 1:
outputs = backend.numpy.exp(
Expand Down