Skip to content

fix(attention): validate dropout_rate range and require dropout_rng w… #5510

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
Vitaliy-Pikalo wants to merge 1 commit into
base: main
Choose a base branch
from
+10 −0
Open

fix(attention): validate dropout_rate range and require dropout_rng w… #5510

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
10 changes: 10 additions & 0 deletions flax/nnx/nn/attention.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -104,6 +104,16 @@ def dot_product_attention_weights(
Returns:
Output of shape `[batch..., num_heads, q_length, kv_length]`.
"""
if not (0.0 <= dropout_rate < 1.0):
raise ValueError(
f'dropout_rate must be in the range [0, 1), got {dropout_rate!r}. '
'A negative or NaN value silently disables dropout without regularizing.'
)
if not deterministic and dropout_rate > 0.0 and dropout_rng is None:
raise ValueError(
'dropout_rng must be provided when dropout_rate > 0.0 and '
'deterministic=False. Pass a JAX PRNGKey as dropout_rng.'
)
query, key = promote_dtype((query, key), dtype=dtype) # type: ignore[bad-unpacking]
dtype = query.dtype

Expand Down
Loading