How can I reuse the last state of RWKV-v6?

[ddddewang0425]

Hello. I want to reuse the last state of RWKV-v6 in order to input something again after first inference.
But, I think, Because wkv6_cuda.cu doesn't returns state information, I have to change many things of your code.
Is it right or there is something easy way to reuse the last state of RWKV-v6?
I mean how can I make RWKV returns last hidden state? for v6.

[BlinkDL]

check https://pypi.org/project/rwkv/ and https://github.com/BlinkDL/ChatRWKV/tree/main/rwkv_pip_package/src/rwkv/cuda

[ddddewang0425]

Very Very thank you~!!!
I really appreciate that.