Skip to content

RWKV/RWKV-block

Repository files navigation

[WIP] RWKV: Receptive Weighted Key Value

WIP: This is meant to be reference block implmentaiton for various RWKV modules. It is not considered complete

Pytorch design decisions

  • RWKV states are passed around with native tuples, and list intentionally, benchmarking show this has a measurable speed bump compared to data state classes when needed.
  • init state tune weights should be made avaliable via init_state.x.wkv

Conda specific setup

conda create -n py-3-12 python=3.12 pip nvidia
conda activate py-3-12

# Install cuda in conda env
conda install cuda
conda install pytorch torchvision torchaudio pytorch-cuda=12.1 -c pytorch -c nvidia

# Install FLA and other required packages
pip3 install -r requirements.txt

# Optional test requirements
pip3 install -r test/requirements.txt

About

PyTorch implementation of RWKV blocks

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Contributors 3

  •  
  •  
  •