1. git clone https://github.com/raphaelspiekermann/TransformerForIMUbasedHAR.git
2. pip install -r requirements.txt
3. run main.py to generate config.json & meta_config.json
4. Setup config.json:
   • Add a Path to a directory where the data and experiment results will be saved:
     • Example: /home/user/transformer_experiments (folder "transformer_experiments" will be generated if not already existing)
     • 3 Subfolders "data" for storing dataset, "runs" for single experiments and "experiments" for experiments containing multiple subexperiments will be created
   • model: see models/model_configs.json
   • device_id = [cpu | cuda | cuda:idx]
   • split_type = [person | person_random]
5. run main.py
   • Args:
     • -v for verbose mode (logs printed in terminal)
     • -m for running multiple experiments from meta_config.json
     • --experiment name_of_experiment: experiments will be stored under path_from_config/experiments/name_of_experiment
   • Examples for basic usage:
     • python3 main.py -v: single experiment specified in config.json
     • python3 main.py -m --experiment exp1234: multiple experiment, settings from meta_config overwrite config for each run, usefull if for example you want to run the same experiment with different seeds