I choose to use Lightning to make implementation easier and more readable. All of this is logged to tensorboard. Best models are saved to disk, measured by the validation loss (from normal validation set).
Make sure you have packages in requirements.txt installed. You can install them by running:
python3 -m pip install -r requirements.txt
To train the model, run:
python3 main.py
Edit main.py to fit your use.
Base model is found in base_cnn.py. This model could be inherited from for constructing other designs.
To visualize the training process, run:
tensorboard --logdir=tb_logs/my_model/version_<version_number>
Some predictions about the design can be found by running design_finder.py and calcs.ipynb.
Evaluation of trained models can be done by running eval_runs.py.
Best, Morten Svendgård