Beschrijving
RWKV (pronounced RwaKuv) is an RNN with great LLM performance, which can also be directly trained like a GPT transformer (parallelizable). We are at RWKV-7 "Goose". So it's combining the best of RNN and transformer - great performance, linear time, constant space (no kv-cache), fast training, infinite ctx_len, and free sentence embedding.
Atlas-score
- Apache-2.0 (permissive)
- Last commit 4 days ago
- Top 22% by stars in AI & ML::Machine Learning Frameworks