This is a super simple c++/cuda implementation of rwkv with no pytorch/libtorch dependencies.

included is a simple example of how to use in both c++ and python.

• Direct Disk -> Gpu loading ( practically no ram needed )
• Uint8 by default
• Incredibly fast
• No dependencies
• Simple to use
• Simple to build
• Optional Python binding using pytorch tensors as wrappers
• Native tokenizer!
• Windows Support!
• Distributable programs! (check actions for the prebuilt example apps)
• Godot module

• Optimize .pth converter (currently uses a lot of ram)
• Better uint8 support ( currently only uses Q8_0 algorythm)
• Fully fleshed out demos

1. go to the actions tab
2. find a green checkmark for your platform
3. download the executable
4. download or convert a model (downloads here)
5. place the model.bin file in the same place as the executable
6. run the executable

$./build.sh

> mkdir build
> cd build
> cmake ..
> cmake --build . --config release

You can find executable at build/release/rwkv.exe

Make sure you already installed CUDA Toolkit and Visual Studio 2022.

You can download the weights of the model here: https://huggingface.co/BlinkDL/rwkv-4-raven/tree/main

For conversion to a .bin model you can choose between 2 options:

Make sure you have python + torch, tkinter, tqdm and Ninja packages installed.

> cd converter
> python3 convert_model.py

Make sure you have python + torch, tqdm and Ninja packages installed.

> cd converter
> python3 convert_model.py your_downloaded_model.pth

• On Windows, please run the above commands in "x64 Native Tools Command Prompt for VS 2022" terminal.

C++ tokenizer came from this project: https://github.com/gf712/gpt2-cpp/