Skip to content

Fix refer model with RL #1660

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
hhaAndroid merged 1 commit into from
Apr 9, 2026
Merged

Fix refer model with RL #1660

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: 4 additions & 4 deletions xtuner/v1/rl/base/worker.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -37,7 +37,6 @@
from xtuner.v1.ray.config import RolloutConfig
from xtuner.v1.ray.utils import free_object_refs
from xtuner.v1.rl.base.loss import BaseRLLossContext
from xtuner.v1.rl.utils import gather_logprobs
from xtuner.v1.train.trainer import LoadCheckpointConfig
from xtuner.v1.utils import (
XTUNER_DETERMINISTIC,
Expand Down Expand Up @@ -391,9 +390,10 @@ def compute_ref_logprobs(
ref_logprobs_list: list[torch.Tensor] = []
for seq_ctx, shifted_labels in zip(seq_ctx_list, shifted_labels_list):
with torch.no_grad():
ref_output = self._ref_model(seq_ctx=seq_ctx, loss_ctx=None)
ref_logprobs = gather_logprobs(ref_output["logits"], shifted_labels)
ref_logprobs_list.append(ref_logprobs)
loss_ctx = self.logprob_cfg.build(data={"shifted_labels": shifted_labels})
assert loss_ctx is not None
ref_output = self._ref_model(seq_ctx=seq_ctx, loss_ctx={"lm": loss_ctx})
ref_logprobs_list.append(ref_output["loss"])
Comment thread
hhaAndroid marked this conversation as resolved.
self._ref_model.to_device("cpu")
return ref_logprobs_list

Expand Down
Loading