xinyun/vllm
1
0
Fork 0
mirror of https://git.datalinker.icu/vllm-project/vllm.git synced 2026-05-31 05:47:04 +08:00
Code Issues Packages Projects Releases Wiki Activity

remove debug logging

Browse Source 
Signed-off-by: Sage Moore <sage@neuralmagic.com>
This commit is contained in:
Sage Moore 2025-08-08 19:07:49 +00:00
parent 5bbfd95bdb
commit 2cf200c5b8
1 changed files with 8 additions and 8 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

16
vllm/v1/worker/gpu_model_runner.py
View File

@ -1903,27 +1903,27 @@ class GPUModelRunner(LoRAModelRunnerMixin, KVConnectorModelRunnerMixin):
is_dummy_run=is_dummy_run) is_dummy_run=is_dummy_run)
if num_scheduled_tokens not in self.cudagraphs \ if num_scheduled_tokens not in self.cudagraphs \
and not skip_cuda_graphs and build_cuda_graph: and not skip_cuda_graphs and build_cuda_graph:
if is_global_first_rank(): # if is_global_first_rank():
logger.info(f"CAPTURING {num_scheduled_tokens}") # logger.info(f"CAPTURING {num_scheduled_tokens}")
return self._capture_ubatches(ubatch_metadata, self.model) return self._capture_ubatches(ubatch_metadata, self.model)
elif num_scheduled_tokens in self.cudagraphs and not skip_cuda_graphs: elif num_scheduled_tokens in self.cudagraphs and not skip_cuda_graphs:
# assert False # assert False
cudagraph_metadata = self.cudagraphs[num_scheduled_tokens] cudagraph_metadata = self.cudagraphs[num_scheduled_tokens]
if is_global_first_rank(): # if is_global_first_rank():
logger.info(f"UBATCH REPLAY {num_scheduled_tokens}") # logger.info(f"UBATCH REPLAY {num_scheduled_tokens}")
cudagraph_metadata.cudagraph.replay() cudagraph_metadata.cudagraph.replay()
return cudagraph_metadata.outputs return cudagraph_metadata.outputs
else: else:
if is_global_first_rank(): # if is_global_first_rank():
logger.info(f"RUNNING NORMALLY {num_scheduled_tokens}") # logger.info(f"RUNNING NORMALLY {num_scheduled_tokens}")
return self._run_ubatches(ubatch_metadata, self.model) return self._run_ubatches(ubatch_metadata, self.model)
# run normal batch # run normal batch
else: else:
input_ids, positions, inputs_embeds, intermediate_tensors = \ input_ids, positions, inputs_embeds, intermediate_tensors = \
self.model_inputs(slice(0, num_scheduled_tokens), self.model_inputs(slice(0, num_scheduled_tokens),
scheduler_output, is_dummy_run) scheduler_output, is_dummy_run)
if is_global_first_rank(): # if is_global_first_rank():
logger.info(f"RUNNING FULL BATCH {num_scheduled_tokens}") # logger.info(f"RUNNING FULL BATCH {num_scheduled_tokens}")
skip_cuda_graphs = self.parallel_config.enable_microbatching skip_cuda_graphs = self.parallel_config.enable_microbatching
with set_forward_context(attn_metadata, with set_forward_context(attn_metadata,
vllm_config=self.vllm_config, vllm_config=self.vllm_config,