You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
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.
Implementation of a custom architecture on nanoRWKV: A nanoGPT-style adaptation of the RWKV Language Model, which combines the simplicity of RNNs with GPT-level performance for large language models (LLMs).
Support for running RWKV 4/5 World models locally from the command line for conversation and small talk. Support is provided through the prompt.txt defines the prompt for the model. Support Mirostat algorithm and traditional Top-k, Top-p for decoding.