Wednesday657

Hi, can i ask a simple question about RWKV kernel? The rwkv model without customized kernel uses a `for loop` here: https://github.com/huggingface/transformers/blob/3658488ff77ff8d45101293e749263acf437f4d5/src/transformers/models/rwkv/modeling_rwkv.py#L223-L241 I am not familiar with cuda kernel. So...