Skip to content

[TP] Fix parameter detection issue and some invalid TP-plans #42129

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 14 commits into from
Nov 13, 2025
+124 −93
Merged

[TP] Fix parameter detection issue and some invalid TP-plans #42129

Changes from 1 commit
Commits
Show all changes
14 commits
Select commit Hold shift + click to select a range
be9d7e7
fix
Cyrilvallez Nov 10, 2025
01281a0
add test
Cyrilvallez Nov 10, 2025
0b59a20
fix test
Cyrilvallez Nov 10, 2025
12a4f59
fix the obvious
Cyrilvallez Nov 10, 2025
3d7769b
more fix
Cyrilvallez Nov 10, 2025
f7a41c1
fix
Cyrilvallez Nov 10, 2025
fd0a656
continue to improve
Cyrilvallez Nov 11, 2025
ea300e4
more fix
Cyrilvallez Nov 11, 2025
aef7d5b
more
Cyrilvallez Nov 11, 2025
3f1ab90
Merge branch 'main' into fix-tp
Cyrilvallez Nov 11, 2025
2b36e03
fix
Cyrilvallez Nov 11, 2025
6f8942a
fix
Cyrilvallez Nov 11, 2025
d5247c1
finally
Cyrilvallez Nov 11, 2025
107d71f
CI
Cyrilvallez Nov 11, 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 src/transformers/integrations/tensor_parallel.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -150,7 +150,7 @@ def _get_parameter_tp_plan(parameter_name: str, tp_plan: dict[str, str], is_weig
The `is_weight` is important because for weights, we want to support `.weights` and `.bias` cases seamlessly! but
not parent classes for `post_init` calls
"""
generic_param_name = re.sub(r"\d+", "*", parameter_name)
generic_param_name = re.sub(r"\.\d+\.", ".*.", parameter_name)
if generic_param_name in tp_plan:
return tp_plan[generic_param_name]
elif "." in generic_param_name and generic_param_name.rsplit(".", 1)[0] in tp_plan and is_weight:
Expand Down