Skip to content

fix: Remove obsolete MoE aux loss validation assertion #1852

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
mingmao-inflection wants to merge 2 commits into
base: main
Choose a base branch
from
+0 −9
Open

fix: Remove obsolete MoE aux loss validation assertion #1852

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
ea5a49f
Remove MoE aux loss validation assertion
mingmao-inflection Jan 30, 2026
946b502
Merge branch 'main' into main
mingmao-inflection Jan 30, 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
9 changes: 0 additions & 9 deletions nemo_rl/models/megatron/setup.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -527,15 +527,6 @@ def _validate_training_config(config: PolicyConfig, model_cfg: Any) -> None:
model_cfg.calculate_per_token_loss = True
model_cfg.perform_initialization = True

# MoE aux loss validation
assert (
"aux_loss" not in model_cfg.moe_router_load_balancing_type
or model_cfg.moe_aux_loss_coeff == 0
), (
"MoE aux loss is currently not supported due to a known bug in Megatron-LM. "
"See https://github.com/NVIDIA/Megatron-LM/issues/1984 for more details."
)


def _validate_dtype_config(
dtype: torch.dtype, model_cfg: Any, optimizer_cfg: Any
Expand Down