A high-throughput and memory-efficient inference and serving engine for LLMs
Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git
Synced Nov 6, 21:45 UTC
[Model Runner V2] Remove unused warmup_for_prefill method (#36762)
Signed-off-by: Woosuk Kwon <woosuk@inferact.ai>
Woosuk Kwon committed
Mar 11, 2026 at 08:02 UTC
8ccbcda5c0d460b0189f274bfbfe4947b45bd5cb
1 file changed
-7
vllm/v1/worker/gpu/model_runner.py
-7
@@ -532,13 +532,6 @@ class GPUModelRunner(LoRAModelRunnerMixin):
532
)
533
return cuda_graph_size
534
535
- def warmup_for_prefill(self) -> None:
536
- # For FlashInfer, we would like to execute a dummy prefill run
537
- # to trigger JIT compilation.
538
- if all("FLASHINFER" in b.get_name() for b in self.attn_backends.values()):
539
- self._dummy_run(self.max_num_tokens, skip_attn=False)
540
- torch.accelerator.synchronize()
541
-
535
def finish_requests(self, scheduler_output: SchedulerOutput) -> None:
536
finished_req_ids = scheduler_output.finished_req_ids
537
preempted_req_ids = scheduler_output.preempted_req_ids