Skip to content

Fix UnboundLocalError in shard_and_distribute_module for replicated parameters #45675

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
3outeille merged 2 commits into from
Apr 28, 2026
+3 −1
Merged

Fix UnboundLocalError in shard_and_distribute_module for replicated parameters #45675

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
edf31a4
fix
Abdennacer-Badaoui Apr 28, 2026
5f588b5
Merge branch 'main' into small-fix-tp
3outeille Apr 28, 2026
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
4 changes: 3 additions & 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 @@ -1467,6 +1467,7 @@ def shard_and_distribute_module(
else:
logger.info(f"Tensor sharding plan for {param_name}: {current_shard_plan}")

tp_layer = None
if current_shard_plan is not None:
try:
tp_layer = ALL_PARALLEL_STYLES[current_shard_plan]
Expand All @@ -1488,7 +1489,8 @@ def shard_and_distribute_module(
if not isinstance(param, torch.nn.Parameter):
param = torch.nn.Parameter(param, requires_grad=empty_param.is_floating_point())
setattr(module_to_tp, param_type, param)
tp_layer.update_module_attributes(module_to_tp)
if tp_layer is not None:
tp_layer.update_module_attributes(module_to_tp)
return param


Expand Down
Loading