Description.md

1. Baseline vocoder

2. Pruning gru1. Target density 10%

3. Single GRU layer.

4. Resnet depth of 3.

5. Training length 5->7.

6. Trained on northandsouth.

7. Rescaling True.

8. Fixed aux net count.

9. FC layers 3->2

10. Prune FC layers.