Skip to content

Latest commit

 

History

History
38 lines (28 loc) · 1.17 KB

README.md

File metadata and controls

38 lines (28 loc) · 1.17 KB
Raw

Attention as a Hypernetwork

Official code to reproduce experiments in Attention as a Hypernetwork.

Installation

Install jax according to the instructions for your platform after which you can install the remaining dependencies with:

pip install -r requirements.txt

Structure

All experiments have a corresponding sweep file in sweeps/ and can be run using

`wandb sweep /sweeps/[name].yaml`

Default hyperparameters for all methods and experiments can be found in configs/. If you'd like to directly run a specific experiment for a single seed you can use:

python run.py --config 'configs/[experiment].py:[method]'

where experiment $\in$ [logic, raven, wiki] and method $\in$ [softmax_attention, linear_attention, linear_hypatt].

Citation

If you use this code in your research, please cite the paper:

@article{2024hyperatt,
  title={Attention as a Hypernetwork}, 
  author={Simon Schug and Seijin Kobayashi and Yassir Akram and João Sacramento and Razvan Pascanu},
  year={2024},
  url = {https://arxiv.org/abs/2406.05816},
}