This is a weight transfer from RWKV-4 to FWKV. Why? They have very similar architectures, but FWKV is much faster and simpler than RWKV, and it computes WKV in a very different way. The weights are going to be released when I think that the model is stable or works properly.
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support