Skip to content

Fix bnb fsdp loading for pre-quantized checkpoint #41415

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 4 commits into from
Oct 9, 2025
Merged

Fix bnb fsdp loading for pre-quantized checkpoint #41415

Changes from 1 commit
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
b90ed34
fix
SunMarc Oct 7, 2025
a18fe86
fix
SunMarc Oct 7, 2025
d831ab2
get_param_name
SunMarc Oct 9, 2025
3b45300
fix device name
SunMarc Oct 9, 2025
File filter

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
Prev Previous commit
fix device name
  • Loading branch information
@SunMarc
SunMarc committed Oct 9, 2025
commit 3b453005e9f4dd52fcd189006cff875e2789b0e4
6 changes: 3 additions & 3 deletions src/transformers/modeling_utils.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -770,10 +770,10 @@ def _load_state_dict_into_meta_model(
if value.device.type == "meta":
continue
val_kwargs = value.__dict__
if value.dtype in [torch.uint8, torch.int8]:
if not value.is_floating_point():
val_kwargs["requires_grad"] = False
param_to = "meta" if is_fsdp_enabled() and not is_local_dist_rank_0() else "cpu"
value = type(value)(value.data.to(param_to), **val_kwargs)
device = "meta" if is_fsdp_enabled() and not is_local_dist_rank_0() else "cpu"
value = type(value)(value.data.to(device), **val_kwargs)
setattr(module, param_type, value)

# Remove the param from the state dict if it was not loaded on the fly to avoid wasting memory
Expand Down