RWKV-6-World-3B-v2.1-GGUF

This repo contains the RWKV-6-World-3B-v2.1-GGUF quantized with the latest llama.cpp(b3651).

How to run the model

• Get the latest llama.cpp:

git clone https://github.com/ggerganov/llama.cpp

• Download the GGUF file to a new model folder in llama.cpp(choose your quant):

cd llama.cpp
mkdir model
git clone https://huggingface.co/Lyte/RWKV-6-World-3B-v2.1-GGUF
mv RWKV-6-World-3B-v2.1-GGUF/RWKV-6-World-3B-v2.1-GGUF-Q4_K_M.gguf model/
rm -r RWKV-6-World-3B-v2.1-GGUF

• For Windows other than git cloning the repo, you just create the "model" folder inside llama.cpp folder and in here click "Files and versions" and download the model quant you want there.

• Now to run the model, you can use the following command:

./llama-cli -m ./model/RWKV-6-World-3B-v2.1-GGUF-Q4_K_M.gguf --in-suffix "Assistant:" --interactive-first -c 1024 -t 0.7 --top-k 50 --top-p 0.95 -n 128 -p "Assistant: Hello, what can i help you with today?\nUser:" -r "User:"