Evaluating the MuseGAN jazz sample model

To evaluate the pre-trained MuseGAN jazz genre model

1. Open the AWS DeepComposer console.

2. In the navigation pane, choose Models.

3. On the Models page, under Sample model, choose the Jazz pre-trained model.

4. On the training results page for the model, choose Loss function to review the loss function graph.

   

   The generator loss and discriminator loss are superimposed onto the same graph, but on different scales. The scale of the generator loss is shown on the left side and the scale of the discriminator loss is shown on the right side.

   In this training job, the generator loss plateaus around the 50th epoch, when it stops significantly improving its ability to generate realistic music.

5. Choose Sample output to listen to accompaniment tracks that would be generated had inference been performed at that specific epoch.

   For every 50th epoch, you can listen to the accompaniment tracks that could have been generated.

   The discriminator loss shows similar behavior, but is less noisy after it plateaus.

6. Choose the Generated samples tab.

   You can choose one of the available metrics: Pitches used, Pitch classes used, Empty bar rate, Polyphonic range, In scale ratio, Drum in pattern.

   To view the definition of a generated sample metric, choose the corresponding tab. For each generated sample, a graph shows output samples compared with ground truth samples. Over time, the generated samples should converge to the target line. The plot in the tab shows the metric plotted for the generated samples at each epoch (scatter) vs. the ground truth samples (line), respectively. The ground truth samples allow image data to be related to real images as opposed to images provided by inference.