Skip to content

fix(trainer): supplement dfed770 by adding missing update_weights in … #469

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
kylemontgomery1 merged 2 commits into from
Apr 4, 2026
+5 −0
Merged

fix(trainer): supplement dfed770 by adding missing update_weights in … #469

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
5b54789
fix(trainer): supplement dfed770 by adding missing update_weights in …
MarkJoson Apr 2, 2026
af297ca
additional fixes of sdk trainer
MarkJoson Apr 4, 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
5 changes: 5 additions & 0 deletions rllm/trainer/verl/agent_sdk_trainer.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -175,6 +175,7 @@ def fit_agent(self):

self.global_steps = 0
self._load_checkpoint()
self.checkpoint_manager.update_weights(self.global_steps)

start_time = time.time()
if self.config.trainer.get("val_before_train", True):
Expand Down Expand Up @@ -213,6 +214,7 @@ def fit_agent(self):
with marked_timer("step", timing_raw):
# generate trajectories
final_gen_batch_output = self.generate_trajectories(batch=new_batch, timing_raw=timing_raw)
self.checkpoint_manager.sleep_replicas()

# need to repeat to make shape match
repeat_counts = final_gen_batch_output.meta_info["repeat_counts"]
Expand Down Expand Up @@ -473,6 +475,9 @@ def fit_agent(self):
with marked_timer("save_checkpoint", timing_raw, color="green"):
self._save_checkpoint()

# update weights from trainer to rollout
with marked_timer("update_weights", timing_raw, color="red"):
self.checkpoint_manager.update_weights(self.global_steps)
# Visualize some sample trajectories
if batch is not None and len(batch) > 0:
# Randomly select a few samples to visualize
Expand Down