Skip to content

Fix Torch FSDP2 crash: add force_all_reduce kwarg to base finish_grad_sync #4953

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
cspades merged 3 commits into from
Jul 8, 2026
Merged

Fix Torch FSDP2 crash: add force_all_reduce kwarg to base finish_grad_sync #4953

Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
5242aab
Fix Torch FSDP2 crash: add force_all_reduce kwarg to base finish_grad…
factnn May 23, 2026
6ab3df1
Fix Torch FSDP2 crash: add force_all_reduce kwarg to TorchFullySharde…
factnn Jun 1, 2026
f351ef2
Merge branch 'main' into fix/fsdp2-finish-grad-sync-kwarg
cspades Jul 6, 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
11 changes: 11 additions & 0 deletions megatron/core/distributed/torch_fully_sharded_data_parallel.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -147,6 +147,17 @@ def restore_custom_attrs(module, custom_attrs):

restore_custom_attrs(self.module, attrs)

def finish_grad_sync(self, force_all_reduce=False):
"""
Finishes grad sync (all-reduce or reduce-scatter) communication operations
for all model gradients.

When overlap_grad_reduce is set to True, waits for asynchronous communication
calls to complete. When overlap_grad_reduce is set to False, calls synchronous
communication ops.
"""
super().finish_grad_sync()

def load_state_dict(self, state_dict, strict=True):
"""
No-op because tensors are already loaded in-place by
Expand Down
Loading