Skip to content

Fix: Gemma3TextConfig rope scaling assignments #41934

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
Rocketknight1 merged 2 commits into from
Oct 30, 2025
Merged

Fix: Gemma3TextConfig rope scaling assignments #41934

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
479d3f0
Fix: Gemma3TextConfig rope scaling assignments
RyanMullins Oct 29, 2025
13ed5ed
Fix: type annotation for rope_parameters
RyanMullins Oct 29, 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
14 changes: 10 additions & 4 deletions src/transformers/models/gemma3/configuration_gemma3.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -156,7 +156,7 @@ def __init__(
layer_types: Optional[list[str]] = None,
final_logit_softcapping: Optional[float] = None,
attn_logit_softcapping: Optional[float] = None,
rope_parameters: Optional[RopeParameters | dict[RopeParameters]] = None,
rope_parameters: Optional[RopeParameters | dict[str, RopeParameters]] = None,
use_bidirectional_attention: Optional[bool] = False,
**kwargs,
):
Expand Down Expand Up @@ -186,10 +186,16 @@ def __init__(
self.final_logit_softcapping = final_logit_softcapping
self.attn_logit_softcapping = attn_logit_softcapping
self.layer_types = layer_types

# Try to set `rope_scaling` if available, otherwise use `rope_parameters`
rope_scaling = kwargs.pop("rope_scaling", None)
if rope_scaling is not None:
rope_parameters = {"sliding_attention": {"rope_type": "default"}, "full_attention": rope_scaling}
if (rope_scaling := kwargs.pop("rope_scaling", None)) is not None:
if rope_parameters is None:
rope_parameters = {"sliding_attention": {"rope_type": "default"}, "full_attention": rope_scaling}
elif "full_attention" in rope_parameters:
rope_parameters["full_attention"].update(rope_scaling)
else:
rope_parameters.update(rope_scaling)

self.rope_parameters = rope_parameters
self.use_bidirectional_attention = use_bidirectional_attention
if use_bidirectional_attention:
Expand Down
14 changes: 10 additions & 4 deletions src/transformers/models/gemma3/modular_gemma3.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -171,7 +171,7 @@ def __init__(
layer_types: Optional[list[str]] = None,
final_logit_softcapping: Optional[float] = None,
attn_logit_softcapping: Optional[float] = None,
rope_parameters: Optional[RopeParameters | dict[RopeParameters]] = None,
rope_parameters: Optional[RopeParameters | dict[str, RopeParameters]] = None,
use_bidirectional_attention: Optional[bool] = False,
**kwargs,
):
Expand Down Expand Up @@ -201,10 +201,16 @@ def __init__(
self.final_logit_softcapping = final_logit_softcapping
self.attn_logit_softcapping = attn_logit_softcapping
self.layer_types = layer_types

# Try to set `rope_scaling` if available, otherwise use `rope_parameters`
rope_scaling = kwargs.pop("rope_scaling", None)
if rope_scaling is not None:
rope_parameters = {"sliding_attention": {"rope_type": "default"}, "full_attention": rope_scaling}
if (rope_scaling := kwargs.pop("rope_scaling", None)) is not None:
if rope_parameters is None:
rope_parameters = {"sliding_attention": {"rope_type": "default"}, "full_attention": rope_scaling}
elif "full_attention" in rope_parameters:
rope_parameters["full_attention"].update(rope_scaling)
else:
rope_parameters.update(rope_scaling)

self.rope_parameters = rope_parameters
self.use_bidirectional_attention = use_bidirectional_attention
if use_bidirectional_attention:
Expand Down