AllReduce
communication
Ring or tree all-reduce across TP group; bandwidth-dominated
公式
FLOPs
0Bytes
2 * (N-1)/N * payload_bytes使用此算子的模型 (0)
尚无模型在算子拆解中引用此算子。
Ring or tree all-reduce across TP group; bandwidth-dominated
02 * (N-1)/N * payload_bytes尚无模型在算子拆解中引用此算子。