gitpulse

BlinkDL/RWKV-LM

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.

14,778 stars1,027 forksPythonApache-2.0pushed 2026-11-04

View on GitHub →

Star momentum

Today+5 (+0.0%)
This week+6 (+0.0%)
This month+28 (+0.2%)

Topics

attention-mechanismdeep-learninggptgpt-2gpt-3language-modellinear-attentionlstmpytorchrnntransformertransformersrwkvchatgpt