Skip to content

GitLab

Commit 8d971060 authored by 王敏's avatar 王敏
Browse files

暂时去掉profilling标志位，避免影响其他模型

parent 7a97637e
Show whitespace changes
Inline Side-by-side
Showing with 3 additions and 3 deletions
vllm/v1/worker/gpu_model_runner.py
View file @ 8d971060
...@@ -2224,8 +2224,8 @@ class GPUModelRunner(LoRAModelRunnerMixin): ...@@ -2224,8 +2224,8 @@ class GPUModelRunner(LoRAModelRunnerMixin):
def profile_run(self) -> None: def profile_run(self) -> None:
# set profiling flag to avoid torch compile # set profiling flag to avoid torch compile
set_profilling(True) #set_profilling(True)
self._sync_device() #self._sync_device()
# Profile with multimodal encoder & encoder cache. # Profile with multimodal encoder & encoder cache.
# TODO: handle encoder-decoder models once we support them. # TODO: handle encoder-decoder models once we support them.
...@@ -2310,7 +2310,7 @@ class GPUModelRunner(LoRAModelRunnerMixin): ...@@ -2310,7 +2310,7 @@ class GPUModelRunner(LoRAModelRunnerMixin):
del hidden_states, output del hidden_states, output
self.encoder_cache.clear() self.encoder_cache.clear()
gc.collect() gc.collect()
set_profilling(False) #set_profilling(False)
def capture_model(self) -> None: def capture_model(self) -> None:
if not self.use_cuda_graph: if not self.use_cuda_graph:
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment