- Download the data from here. Save the individual files to a directory, say,
/path/to/data(you now should have/path/to/data/{gammma, eplus, piplus}.hdf5). - Edit the configuration file
config.json(or make a copy) to point to this directory, and edit the config to point to a location where you want model metadata & logging to occur (say,/path/to/save/things). - From the directory, run
python trainer.py config.json, and profit!
Just run pip install -r requirements.txt (or, pip install -r requirements-gpu.txt if you've got a CUDA-enabled graphics card).
[update 10/13/20] This requires some old software, so please consider using a virtual environment. The specific versions of Keras and TensorFlow matter.
We recommend using Python 3. If you need to use Python 2, please downgrade the TensorFlow version to 1.15.0.
- Keras==2.0.8
- Keras-contrib (from our fork, on branch
densenet-mod) - Pandas
- Numpy
- Scikit learn
- h5py
- TensorFlow<=1.15.4 (make sure to install the GPU version if you can)