Supporting Codes of U-AnDi for data generation and model training
Xiang Qu, Yi Hu, Wenjie Cai, Yang Xu, Hu Ke, Guolong Zhu, and Zihan Huang
School of Physics and Electronics, Hunan University, Changsha 410082, China
E-mail: huangzih@hnu.edu.cn
Note: For code files in this repository:
- The suffix
reg
,cls
, andexp
correspond to subtask1, subtask2, and experimental data, respectively.
# Generation
python generate_data_reg.py --l 500 --N 1000000
# Pre-Processing for training
python pre_process_data_reg.py --l 500
l is the length, N is the number of trajectories.
The version of andi-datasets is 2.0.0 in subtask 1, and 0.0.5 in subtask 2 and experimental data.
python train-U-AnDi-reg.py --l 500
- OS: Ubuntu 16.04
- GPU: NVIDIA A100 40G
- Python==3.7.4
- PyTorch==1.12.1
- cuda==11.6