Extract data.zip to the same folder with scripts.
To run training with cuda:
python train.py cuda
To run training with cpu:
python train.py
During the run program outputs some information about current hyperparameters and losses, accuracy. \
Program will save the current best performing network and overwrite the previous one during all the run.
As a result, after program finishes we have the best network saved in 'model_state_dict' file.
To run testing with cuda:
python test.py cuda
To run testing with cpu:
python test.py
Testing should be done only after training program completed! Because it needs to first load saved best network.
- Sanity check:
python sanity_check.py
- Plot training and validation losses of best network (should run training first):
python plot_best.py