RWKV is an RNN with transformer-level LLM performance. It can be directly trained like a GPT (parallelizable). So it's combining the best of RNN and transformer - great performance, fast inference, saves VRAM, fast training, "infinite" ctx_len, and free sentence embedding.
Nice work!
And I wonder what is the difference between the rwkv-v1/v2-RNN/v3/v4/v4neo version,should we use v4 or v4neo now?