100% RNN RWKV-LM trained on 12+ natural languages. Demo limited to ctxlen 4096. For best results, write short imperative prompts like commands and requests. Example: use "Tell me what my name is" instead of "What's my name?".
Clone this space for faster inference if you can run the app on GPU or better CPU. To use CUDA, replace strategy='cpu bf16' with strategy='cuda fp16' in app.py.