Awesome

Frad

This repository contains the official implementation of the paper titled "Fractional Denoising for 3D Molecular Pre-training," accepted by ICML23.

Pre-training

Pre-traing for Frad:

python -u scripts/train.py --conf examples/ET-PCQM4MV2_dih_var0.04_var2_com_re.yaml --layernorm-on-vec whitened --job-id frad_pretraining --num-epochs 8 

The data used to pre-train the model is provided here: https://drive.google.com/drive/folders/1F9CyD4HkVL0XFNwtSOTHaXgqZfzLAqFf?usp=sharing

Fine-Tuning

Pre-trained Models

The pre-trained models can be accessed via the following links:

• For QM9: Download Link

• For MD17: Download Link

Finetune on QM9

Below is the script for fine-tuning the QM9 task. Ensure to replace pretrain_model_path with the actual model path. In this script, the subtask is set to 'homo', but it can be replaced with other subtasks as well.

python -u scripts/train.py --conf examples/ET-QM9-FT_dw_0.2_long.yaml --layernorm-on-vec whitened --job-id frad_homo --dataset-arg homo  --denoising-weight 0.1 --dataset-root $datapath --pretrained-model $pretrain_model_path

Finetune on MD17

Below is the script for fine-tuning the MD17 task. Replace pretrain_model_path with the actual model path. In this script, the subtask is set to 'aspirin', but it can be replaced with other subtasks such as {'benzene', 'ethanol', 'malonaldehyde', 'naphthalene', 'salicylic_acid', 'toluene', 'uracil'}.

python -u scripts/train.py --conf examples/ET-MD17_FT-angle_9500.yaml  --job-id frad_aspirin --dataset-arg aspirin --pretrained-model $pretrain_model_path --dihedral-angle-noise-scale 20 --position-noise-scale 0.005 --composition true --sep-noisy-node true --train-loss-type smooth_l1_loss

How to Cite

If you find this work helpful, please consider citing us:

@InProceedings{pmlr-v202-feng23c,
  title = 	 {Fractional Denoising for 3{D} Molecular Pre-training},
  author =       {Feng, Shikun and Ni, Yuyan and Lan, Yanyan and Ma, Zhi-Ming and Ma, Wei-Ying},
  booktitle = 	 {Proceedings of the 40th International Conference on Machine Learning},
  pages = 	 {9938--9961},
  year = 	 {2023},
  volume = 	 {202},
  series = 	 {Proceedings of Machine Learning Research},
  month = 	 {23--29 Jul},
  publisher =    {PMLR},
  pdf = 	 {https://proceedings.mlr.press/v202/feng23c/feng23c.pdf},
  url = 	 {https://proceedings.mlr.press/v202/feng23c.html},
}