MCPcopy Create free account
hub / github.com/PaddlePaddle/FastDeploy / sot_warmup

Method sot_warmup

fastdeploy/worker/gpu_model_runner.py:2198–2206  ·  view source on GitHub ↗
(self)

Source from the content-addressed store, hash-verified

2196
2197 @sot_warmup_guard(True)
2198 def sot_warmup(self) -> None:
2199 start_time = time.perf_counter()
2200 for batch_size in self.sot_warmup_sizes:
2201 self._dummy_run(
2202 num_tokens=self.fd_config.get_max_chunk_tokens(),
2203 batch_size=batch_size,
2204 )
2205 logger.info(f"SOT warmup the model with the batch size:{batch_size}")
2206 logger.info(f"SOT warmup took {time.perf_counter() - start_time} seconds")
2207
2208 def _get_p_done_idxs_gd(self, model_forward_batch: Optional[List[Request]], num_running_requests: int):
2209 """

Callers

nothing calls this directly

Calls 3

_dummy_runMethod · 0.95
get_max_chunk_tokensMethod · 0.80
infoMethod · 0.45

Tested by

no test coverage detected