flashinfer.comm.moe_a2a_initialize¶
- flashinfer.comm.moe_a2a_initialize(workspace: Tensor, ep_rank: int, ep_size: int, max_num_tokens: int)¶
Back to top