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.
Are there any plans to train RWKV on this? https://huggingface.co/datasets/togethercomputer/RedPajama-Data-1T https://huggingface.co/datasets/HuggingFaceH4/databricks_dolly_15k https://huggingface.co/datasets/OpenAssistant/oasst1