Running training.py will train the four models mentioned in the report.
Running evaluation.py will display the spctrogram model output of all of the testing set (all 270) for OpenUnmix and Unmix-AO, then display the BSS statistics of all four model. If you don't want to see the spectrogram output, simply comment out line 511 in evaluation.py