Skip to content

Fix attention mask to use float_lowest instead of -inf and add NaN-safe softmax handling #2654

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

Open
wants to merge 2 commits into
base: main
Choose a base branch
from
Open

Fix attention mask to use float_lowest instead of -inf and add NaN-safe softmax handling #2654

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
d6eccaa
Fix attention mask to use float_lowest instead of -inf and add unit t…
Aniketsy Oct 26, 2025
0d7c411
Remove helper function and test
Aniketsy Nov 13, 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
2 changes: 1 addition & 1 deletion onnxscript/function_libs/torch_lib/ops/nn.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2078,7 +2078,7 @@ def _aten_scaled_dot_product_attention_bool_mask_onnx(
key_transposed_scaled = op.Mul(key_transposed, op.Sqrt(scale))
# Turn the Boolean mask to float: attn_mask.masked_fill(not attn_mask, -float('inf'))
zero = op.Constant(value=ir.tensor(0.0, dtype=query.dtype))
neg_inf = op.Constant(value=ir.tensor(-float("inf"), dtype=query.dtype))
neg_inf = op.Constant(value=ir.tensor(query.dtype.min), dtype=query.dtype)
attn_mask = op.Where(attn_mask, zero, neg_inf)
attn_weight = op.Softmax(
op.Add(op.MatMul(query_scaled, key_transposed_scaled), attn_mask),
Expand Down