MCPcopy Create free account
hub / github.com/NVIDIA/TensorRT-LLM / synchronize

Method synchronize

tensorrt_llm/_torch/pyexecutor/sampler.py:105–108  ·  view source on GitHub ↗
(self)

Source from the content-addressed store, hash-verified

103 worker_futures: Optional[list[futures.Future[Any]]] = None
104
105 def synchronize(self):
106 if self.worker_futures:
107 futures.wait(self.worker_futures)
108 self.cuda_event.synchronize()
109
110
111@dataclass(kw_only=True)

Callers 15

run_testFunction · 0.80
syncMethod · 0.80
profile_allreduceFunction · 0.80
cuda_error_early_quitFunction · 0.80
process_gpu_memory_info_availableFunction · 0.80
test_logits_post_processorFunction · 0.80
_construct_executionMethod · 0.80
test_gpt_attentionMethod · 0.80
_construct_executionMethod · 0.80
torch_execMethod · 0.80
run_engineMethod · 0.80
hfMethod · 0.80

Calls 1

waitMethod · 0.80

Tested by 15

run_testFunction · 0.64
cuda_error_early_quitFunction · 0.64
process_gpu_memory_info_availableFunction · 0.64
test_logits_post_processorFunction · 0.64
_construct_executionMethod · 0.64
test_gpt_attentionMethod · 0.64
_construct_executionMethod · 0.64
torch_execMethod · 0.64
run_engineMethod · 0.64
hfMethod · 0.64
test_vgqa_model_runner_allcloseMethod · 0.64
test_phiMethod · 0.64