flashinfer.comm.vllm_init_custom_ar

flashinfer.comm.vllm_init_custom_ar(ipc_tensors: List[int], rank_data: Tensor, rank: int, full_nvlink: bool) → int