Skip to content

Fix MoE target_parameters module_count alignment (#3405, #3701) #499

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
GoldenGrapeGentleman wants to merge 2 commits into
base: main
Choose a base branch
from
+12 −3
Open

Fix MoE target_parameters module_count alignment (#3405, #3701) #499

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
42133ed
Fix MoE target_parameters module_count alignment (#3405, #3701)
GoldenGrapeGentleman Feb 14, 2026
8754834
Refine: use stats.module is None instead of string matching
GoldenGrapeGentleman Feb 14, 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
15 changes: 12 additions & 3 deletions unsloth_zoo/saving_utils.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -390,6 +390,18 @@ def create_lora_statistics(model, merge_into_original = False, return_state_dict
remove_keys.add(name)
pass
pass
# PEFT target_parameters (ParamWrapper for nn.Parameter, not nn.Linear)
# have lora_A/B/scaling but no .base_layer, leaving module_count short.
# nn.Linear targets always get .base_layer set, so module=None reliably
# identifies nn.Parameter targets. Count them to align (#3405, #3701).
for _key, _stats in lora_weights.items():
if (
_stats.lora_A is not None
and _stats.lora_B is not None
and _stats.module is None
):
module_count += 1

if not (module_count == lora_A_count == lora_B_count == scaling_count):
print(
f"[Unsloth merge debug] LoRA count mismatch: modules={module_count}, "
Expand All @@ -405,9 +417,6 @@ def create_lora_statistics(model, merge_into_original = False, return_state_dict
print(f" key={k} param={param_name} A={a_shape} B={b_shape}")
except Exception:
pass
# Allow merge to continue; downstream checks will still fail loudly if tensors are missing
# but this avoids silent assertion without context.
# TODO: handle MoE target_parameters to align counts.

# Also return state_dict if needed
if return_state_dict:
Expand Down