Skip to content

scheduler orchestration optimization 1 #381

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
mayinghan merged 6 commits into from
Dec 19, 2025
Merged

scheduler orchestration optimization 1 #381

Show file tree
Hide file tree
Changes from 5 commits
Commits
Show all changes
6 commits
Select commit Hold shift + click to select a range
29c6d06
fix scheduler concurrency and orchestration
mayinghan Dec 17, 2025
c9344c3
remove comment
mayinghan Dec 17, 2025
1dd4169
add
mayinghan Dec 18, 2025
692f3ad
resolve comments
mayinghan Dec 18, 2025
c726a57
fix
mayinghan Dec 19, 2025
fb8debe
remove print
mayinghan Dec 19, 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
5 changes: 4 additions & 1 deletion eval_protocol/pytest/default_single_turn_rollout_process.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -39,7 +39,7 @@ def __call__(self, rows: List[EvaluationRow], config: RolloutProcessorConfig) ->
async def process_row(row: EvaluationRow) -> EvaluationRow:
"""Process a single row asynchronously."""
start_time = time.perf_counter()

if len(row.messages) == 0:
raise ValueError("Messages is empty. Please provide a non-empty dataset")

Expand Down Expand Up @@ -97,7 +97,10 @@ async def process_row(row: EvaluationRow) -> EvaluationRow:
chunks.append(chunk)
response = litellm.stream_chunk_builder(chunks, messages_payload)
else:
tc = time.perf_counter()
# print(f"run_id {row.execution_metadata.run_id} request_params: {json.dumps(request_params)}")
response = await acompletion(**request_params)
print(f"run_id {row.execution_metadata.run_id} time taken: {time.perf_counter() - tc} speculation_enabled: {request_params.get('extra_body', {}).get('prediction', None) is not None}")
Copy link
Copy Markdown

@cursor cursor Bot Dec 19, 2025

Choose a reason for hiding this comment

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

Bug: Debug print statement left in production code

A debug print statement was left in the production code path. This prints timing information and speculation status for every non-streaming LLM completion call, which will pollute logs and stdout in production. The timing variable tc and the uncommented print at line 103 appear to be debugging/profiling code that wasn't removed before committing.

Fix in Cursor Fix in Web


assert response is not None, "Response is None"
assert isinstance(response, ModelResponse), "Response should be ModelResponse"
Expand Down
Loading
Loading