KO
|
EN
gitlite — search
Search
#python
#javascript
#react
#chatgpt
#reactjs
#nextjs
#openai
#gpt-3
#android
#machine-learning
#css
#html
RWKV-block
★ 31
Open GitHub ↗
PyTorch implementation of RWKV blocks
Download README (.md)
Explore Similar Repositories
ebpf_exporter
:
a pure-Go Prometheus exporter for the eBPF Linux subsystem
silisocs
:
Silicon Society Sandbox (SiliSocS) is an versatile and extensible experimentation system for EASE-configured generative multi-agent simulations
QT_UI_Simplify
:
使用QT_CPP实现一些极简主义的控件
checkers
:
No description available.
luckynotes
:
The faster and simpler way to take notes and start learn&build in public
// repository documentation
Was this content helpful?
★ 0
(0 ratings)
Select Rating:
★
★
★
★
★
Submit Feedback
Recent Feedback
×
Download README
Do you want to download the
README.md
file for
RWKV-block
?
Download (.md)
# [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 ```bash 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 ```
