RNNControl
Jupyter notebooks that run experiments investigating the effect of conditional parameters on audio synthesis.
The code here is closely integrated with audioDataloader and paramManager.
Clone those repositories and run the notebooks from this repo alongside them as follows:
.
+--model
+--README.md
+--paramManager
+--audioDataloader
+--explore_generation.ipynb
+--main.ipynb
+--data
|+--dataset
|+--dataparams
+--utils
|+--myUtils.py
|+--architectures.py
Files
- main.ipynb: for training and saving model (an RNN)
- explore_generation.ipynb: to do cool stuff i.e. audio synthesis with the trained model
- model: pre-trained models can be found here
- myUtils.py: contains some utility functions e.g for plotting
- architectures.py: the model architectures should be loaded from here
Dependencies
- pytorch 0.4.0
- audioDataloader
- paramManager
Authors
- Muhammad Huzaifah
NOTE: this project is incomplete, run at your own risk.
To do
- Speed up generation
- Explore conditional parameters