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.
增加（基本的）PyTorch Lightning 2.0支持。 Hopper和Ada Lovelace架构的GPU使用这个可以直接训练。