Skip to content

[trainer] fix: data balance dp tokens logic with sequence parallelism #3691

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

[trainer] fix: data balance dp tokens logic with sequence parallelism #3691

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
3 changes: 2 additions & 1 deletion verl/trainer/fsdp_sft_trainer.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -437,7 +437,8 @@ def _compute_loss_and_backward(self, batch, do_backward=True, n_micro_batches=1)

if self.config.data.balance_dp_token:
torch.distributed.all_reduce(valid_token_this_rank)
dp_size = self.ulysses_device_mesh.size("dp") if use_sp else torch.distributed.get_world_size()
# Valid token is reduced over all ranks (including SP ranks), set dp size to world size
dp_size = torch.distributed.get_world_size()
else:
dp_size = 1

Expand Down