Skip to content

[fix] handle FSDP DTensor in broadcast_from_megatron_pp #113

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
yxs wants to merge 1 commit into
base: main
Choose a base branch
from
Open

[fix] handle FSDP DTensor in broadcast_from_megatron_pp #113

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
8 changes: 8 additions & 0 deletions mbridge/core/util.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -268,6 +268,14 @@ def unwrap_model(model, module_instances=ALL_MODULE_WRAPPER_CLASSNAMES):
def broadcast_from_megatron_pp(tensor: torch.Tensor):
# tensor is not None only in one of the pp ranks
if tensor is not None:
# FSDP DTensor: gather full tensor before PP broadcast.
try:
from torch.distributed.tensor import DTensor

if isinstance(tensor, DTensor):
tensor = tensor.full_tensor()
except ImportError:
pass
shape = tensor.shape
dtype = tensor.dtype
tensor_parallel = getattr(tensor, "tensor_model_parallel", None)
Expand Down