To install the repo there is a conda environment that will install the necessary packages. Make sure you are in the TiDHy Github directory.
Use command:
conda env create -f environment.yaml
After installing activate the conda environment:
conda activate TiDHy
After pytorch is correctly installed run this command to install pip requirements:
pip install -r requirements.txt
If the requirements.txt file does not install pytorch with cuda, go to this site to install the appropriate pytorch version:
https://pytorch.org/get-started/locally/
To install TiDHy, in the repo folder use:
pip install -e .
For SLDS comparison install ssm package:
https://github.com/lindermanlab/ssm
To Train TiDHy you can run the Run_TiDHy.py script from the terminal with hydra overrides:
python Run_TiDHy.py Run_TiDHy.py dataset=SLDS dataset.train.gpu=0 version=Example
To add a custom dataset you can load data in any way you want. The final formatting should follow the convention of:
- train_data:
(time x features) - val_data:
(time x features) - test_data:
(time x features)
The data can then be stacked with overlapping windows using the stack_data function:
train_inputs = stack_data(train_inputs,cfg.train.sequence_length,overlap=cfg.train.sequence_length//cfg.train.overlap_factor)