/Latent-Diffusion-EBM

[ICML 2022] Latent Diffusion Energy-Based Model for Interpretable Text Modeling

Primary LanguageOpenEdge ABL

Latent Diffusion Energy-Based Model for Interpretable Text Modeling

teaser

[Paper] [Code]

The official code repository for ICML 2022 paper "Latent Diffusion Energy-Based Model for Interpretable Text Modeling".

Installation

The implementation is based on python 3.6.13 and depends on the following commonly used packages, most of which can be directly installed via conda.

Package Version
PyTorch 1.11.0
faiss 1.7.0
numpy 1.19.5
nltk 3.6.5
sklearn 0.24.2

Please refer to this repo if you're having trouble installing faiss.

Datasets and Pretrained Models

Pretrained models are available at: https://drive.google.com/drive/folders/1XWu7olAoYbrKmh8Hnu_zROKhhd3TtmhS?usp=sharing

Training

# Run the corresponding task scripts
python <TASK_SCRIPT>.py --gpu <GPU_ID> --max_kl_weight <WEIGHT_OF_KLD> --mutual_weight <WEIGHT_OF_MI> --cls_weight <WEIGHT_OF_CLS_LOSS>

You may specify the value of arguments for training. Please find the available arguments in the corresponding task scripts in the workspace folder.

Testing

# Evaluate the trained model
python <TASK_SCRIPT>.py --gpu <GPU_ID> --forward_only True

Citation

@inproceedings{yu2022latent,
  author = {Yu, Peiyu and Xie, Sirui and Ma, Xiaojian and Jia, Baoxiong and Pang, Bo and Gao, Ruiqi and Zhu, Yixin and Zhu, Song-Chun and Wu, Ying Nian},
  title = {Latent Diffusion Energy-Based Model for Interpretable Text Modeling},
  booktitle = {Proceedings of International Conference on Machine Learning (ICML)},
  month = {July},
  year = {2022}
}