This repository is the code base for evaluation of time series generation
The code has been tested successfully using Python 3.8 and pytorch 1.11.0. A typical process for installing the package dependencies involves creating a new Python virtual environment.
To install the required packages, run the following:
conda install pytorch==1.11.0 torchvision==0.12.0 torchaudio==0.11.0 cudatoolkit=10.2 -c pytorch
pip install signatory==1.2.6.1.9.0 --no-cache-dir --force-reinstall
pip install cupy-cuda102
pip install git+https://github.com/tgcsaba/ksig.git --no-deps
pip install -r requirements.txtNote that this pipeline relies on wandb for logging and tracking experiemnts, user needs to create a personal wandb account and specify the personal wandb api key in configs/train_gan.yaml.
We aim to create an pipeline with easy configuration interface for controlling on various GAN baseline models, datasets and evaluation metrics. The pipeline is still under development and may change based on the objective of the paper, it currently support 3 generative models on three different datasets. One should able to run experiments with
python run.py --algo ALGO_NAME --dataset DATAwhere ALGO_NAME is a choice of TimeGAN, RCGAN.TimeVAE. DATA is a choice AR1, ROUGH,GBM.
The run.py will retrieve configuration from configs/train_gan.yaml and complete the model training on the specified dataset and evaluation. One can modify configs/train_gan.yaml to further specify model parameters.
After the model training and evaluation, the related model and plots will be store in the numerical_results folder, these files will also uploaded the wandb online folder for each individual run.