Audio Synthesis with Matrix Product States
- TensorFlow
- Python 3.6.6
- model.py
It defines our RNN model and contains other methods, to sample from the trained model for example.
- training_estimators.py
It loads the training data, and runs the training. A custom estimator is used.
- reader.py
It picks out a subset of the data, that contains a desired feature: pitch, instrument ...
- make-small-dataset.py
It imports reader.py to make the dataset.
- Sampling
- What about initialization? Glorot variance depends on size.
- Monitor change of state to see if it is small
- Exploring latent space