Skip to content

fix: CPU RAM efficient loading for nd or HSDP parallelisms #3740

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
S1ro1 merged 1 commit into from
Aug 21, 2025
Merged

fix: CPU RAM efficient loading for nd or HSDP parallelisms #3740

Changes from all commits
Commits
Show all changes
1 commit
Select commit
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: 2 additions & 2 deletions src/accelerate/utils/fsdp_utils.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -506,7 +506,7 @@ def _cast_and_contiguous(tensor, to_contiguous, dtype):
for (param_name, full_param), sharded_param in zip(full_sd.items(), meta_sharded_sd.values()):
device_mesh = sharded_param.device_mesh
full_param = full_param.detach().to(device_mesh.device_type)
dist.broadcast(full_param, src=0, group=device_mesh.get_group())
dist.broadcast(full_param, src=0, group=dist.group.WORLD)
sharded_tensor = distribute_tensor(full_param, device_mesh, sharded_param.placements)
to_contiguous, casting_dtype = _infer_parameter_dtype(
model,
Expand All @@ -520,7 +520,7 @@ def _cast_and_contiguous(tensor, to_contiguous, dtype):
for param_name, sharded_param in meta_sharded_sd.items():
device_mesh = sharded_param.device_mesh
full_tensor = torch.empty(sharded_param.size(), device=device_mesh.device_type, dtype=sharded_param.dtype)
dist.broadcast(full_tensor, src=0, group=device_mesh.get_group())
dist.broadcast(full_tensor, src=0, group=dist.group.WORLD)
sharded_tensor = distribute_tensor(full_tensor, device_mesh, sharded_param.placements)
to_contiguous, casting_dtype = _infer_parameter_dtype(
model,
Expand Down