Skip to content

[training_utils] fix: RM extra scaling in KL/PG losses #4711

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
yyDing1 merged 4 commits into from
Dec 30, 2025
Merged

[training_utils] fix: RM extra scaling in KL/PG losses #4711

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
874ff53
Fix metric scaling
JacobHelwig Dec 29, 2025
e1ed63a
Sum reduction
JacobHelwig Dec 29, 2025
ffd5813
Revert "Sum reduction"
JacobHelwig Dec 29, 2025
e43dadb
Reduce in update fn
JacobHelwig Dec 29, 2025
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: 6 additions & 3 deletions verl/workers/actor/dp_actor.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -431,7 +431,10 @@ def update_policy(self, data: DataProto):

on_policy = len(mini_batches) == 1 and self.config.ppo_epochs == 1

metrics = {}
metrics = {
"actor/pg_loss": 0.0,
"actor/kl_loss": 0.0,
}
for _ in range(self.config.ppo_epochs):
for batch_idx, mini_batch in enumerate(mini_batches):
if self.config.use_dynamic_bsz:
Expand Down Expand Up @@ -530,7 +533,7 @@ def update_policy(self, data: DataProto):
kl_loss = agg_loss(loss_mat=kld, loss_mask=response_mask, loss_agg_mode=loss_agg_mode)

policy_loss = policy_loss + kl_loss * self.config.kl_loss_coef
micro_batch_metrics["actor/kl_loss"] = kl_loss.detach().item() * loss_scale_factor
metrics["actor/kl_loss"] += kl_loss.detach().item() * loss_scale_factor
micro_batch_metrics["actor/kl_coef"] = self.config.kl_loss_coef

if self.config.use_dynamic_bsz:
Expand All @@ -543,7 +546,7 @@ def update_policy(self, data: DataProto):
else:
loss.backward()

micro_batch_metrics["actor/pg_loss"] = pg_loss.detach().item() * loss_scale_factor
metrics["actor/pg_loss"] += pg_loss.detach().item() * loss_scale_factor
append_to_dict(metrics, micro_batch_metrics)

grad_norm = self._optimizer_step()
Expand Down
6 changes: 4 additions & 2 deletions verl/workers/critic/dp_critic.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -192,7 +192,9 @@ def compute_values(self, data: DataProto) -> torch.Tensor:
def update_critic(self, data: DataProto):
# make sure we are in training mode
self.critic_module.train()
metrics = {}
metrics = {
"critic/vf_loss": 0.0,
}

select_keys = ["input_ids", "responses", "response_mask", "attention_mask", "position_ids", "values", "returns"]
has_multi_modal_inputs = "multi_modal_inputs" in data.non_tensor_batch.keys()
Expand Down Expand Up @@ -246,12 +248,12 @@ def update_critic(self, data: DataProto):

micro_batch_metrics.update(
{
"critic/vf_loss": vf_loss.detach().item() * loss_scale_factor,
"critic/vf_clipfrac": vf_clipfrac.detach().item(),
"critic/vpred_mean": masked_mean(vpreds, response_mask).detach().item(),
}
)

metrics["critic/vf_loss"] += vf_loss.detach().item() * loss_scale_factor
append_to_dict(metrics, micro_batch_metrics)

grad_norm = self._optimizer_step()
Expand Down
Loading