Skip to content

graceful shutdown of vllm async #1064

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

Open
wants to merge 1 commit into
base: main
Choose a base branch
from
Open

graceful shutdown of vllm async #1064

Show file tree
Hide file tree
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
3 changes: 1 addition & 2 deletions src/lighteval/models/vllm/vllm_model.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -544,6 +544,7 @@ class AsyncVLLMModel(VLLMModel):
is_async = True

def cleanup(self):
self.model.shutdown()
gc.collect()
destroy_distributed_environment()
torch.cuda.empty_cache()
Expand Down Expand Up @@ -627,7 +628,6 @@ async def _async_batch(self, docs: list[Doc], generative: bool) -> list:
results = await asyncio.gather(*processed_requests)
return results

@cached(SamplingMethod.GENERATIVE)
async def greedy_until(
self,
docs: list[Doc],
Expand Down Expand Up @@ -661,7 +661,6 @@ async def greedy_until(

return results

@cached(SamplingMethod.LOGPROBS)
async def loglikelihood(
self,
docs: list[Doc],
Expand Down
7 changes: 4 additions & 3 deletions src/lighteval/pipeline.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -308,6 +308,8 @@ async def _run_model_async(self):
model_outputs = await self.model.loglikelihood(docs)
outputs[sampling_method] = model_outputs

self.model.cleanup()

return outputs

def _run_model_sync(self):
Expand All @@ -327,6 +329,8 @@ def _run_model_sync(self):
model_outputs = self.model.loglikelihood_rolling(docs)
outputs[sampling_method] = model_outputs

self.model.cleanup()

return outputs

def _run_model(self):
Expand All @@ -339,9 +343,6 @@ def _run_model(self):
else:
outputs = self._run_model_sync()

# Cleaning up the model before running metrics
self.model.cleanup()

return outputs

def _post_process_outputs(self, sampling_method_responses: dict[str, list[ModelResponse]]):
Expand Down