Skip to content

enh: use agent ref from data in rollouts #568

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
bxyu-nvidia merged 4 commits into from
Jan 13, 2026
Merged

enh: use agent ref from data in rollouts #568

Changes from 2 commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
61520c4
use agent ref from data in rollouts
gwarmstrong Jan 9, 2026
ab4a952
Merge branch 'main' into georgea/ng_collect_rollouts_use_agent_ref
gwarmstrong Jan 9, 2026
6ad03a0
ENH do error before rollouts collection
gwarmstrong Jan 13, 2026
12ca75c
Merge branch 'main' into georgea/ng_collect_rollouts_use_agent_ref
gwarmstrong Jan 13, 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
13 changes: 11 additions & 2 deletions nemo_gym/rollout_collection.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -51,7 +51,10 @@ class RolloutCollectionConfig(BaseNeMoGymCLIConfig):
```
"""

agent_name: str = Field(description="The agent to collect rollouts from.")
agent_name: Optional[str] = Field(
default=None,
description="The agent to collect rollouts from. If not specified, uses agent_ref from each data row.",
)
input_jsonl_fpath: str = Field(
description="The input data source to use to collect rollouts, in the form of a file path to a jsonl file."
)
Expand Down Expand Up @@ -108,8 +111,14 @@ async def run_from_config(self, config: RolloutCollectionConfig):

async def _post_coroutine(row: dict) -> None:
row["responses_create_params"] = row["responses_create_params"] | config.responses_create_params
# Use config.agent_name if specified, otherwise use agent_ref from the row
agent_name = config.agent_name or row.get("agent_ref", {}).get("name")
if not agent_name:
raise ValueError(
f"No agent specified for row. Either provide +agent_name config or include agent_ref in data. Row: {row.get('id', 'unknown')}"

Copilot AI Jan 9, 2026

Copy link

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

The error message could be more helpful by showing what was actually found in the row. Consider including the actual content of row.get('agent_ref') to help users debug whether the field exists but is malformed (e.g., missing 'name' key) versus completely absent.

Suggested change
raise ValueError(
f"No agent specified for row. Either provide +agent_name config or include agent_ref in data. Row: {row.get('id', 'unknown')}"
agent_ref_value = row.get("agent_ref")
raise ValueError(
f"No agent specified for row. Either provide +agent_name config or include agent_ref in data. "
f"Row id: {row.get('id', 'unknown')}, agent_ref: {agent_ref_value!r}"

Copilot uses AI. Check for mistakes.
Comment thread
bxyu-nvidia marked this conversation as resolved.
Outdated
)
async with semaphore:
response = await server_client.post(server_name=config.agent_name, url_path="/run", json=row)
response = await server_client.post(server_name=agent_name, url_path="/run", json=row)
await raise_for_status(response)
result = await response.json()
f.write(json.dumps(result) + "\n")
Expand Down