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

Method setup_async

tensorrt_llm/_torch/async_llm.py:40–46  ·  view source on GitHub ↗

Setup the LLM asynchronously.

(self)

Source from the content-addressed store, hash-verified

38 self._async_initialized = False
39
40 async def setup_async(self):
41 """Setup the LLM asynchronously."""
42 if not self._async_initialized:
43 await self._executor.init_workers_async()
44 await self._executor.setup_engine_remote_async()
45 self._async_initialized = True
46 return self
47
48 async def release(self, tags: list[str]):
49 """Release the GPU memory used by the LLM asynchronously.

Callers 3

__await__Method · 0.95
__aenter__Method · 0.95
init_llmMethod · 0.80

Calls 2

init_workers_asyncMethod · 0.80

Tested by 1

init_llmMethod · 0.64