BlinkDL / RWKV-LM
Author description: 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.
- Repository stars
- 14,771All-time #3,348
- Fork
- 1,026
- Last push
- updated 2 days ago
- Repository created
- since 2021
About
Summarised by AI from the README; the original prevails
Author description: Author description: 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.
Basic listing based on official public information; functionality has not been independently evaluated.
- Platform
- Not verified
- Pricing
- Pricing not verifiedCheck current official software, hosting, model and infrastructure pricing.
- Maintenance notes
- Not verified