Seq2Seq Audio

This is a small repository on using pytorch to make many to one and many to many models to generate audio.

Navigation around the repo

There are a few folders; search and notebooks. Both contain unfinished (and untested) code. The search folder contains the code for using a metaheuristic search over the hyperparameters and is not ready to be used yet. The notebooks folder contains the notebooks that were initially hosted on google.colab.

Notebooks

Again these notebooks were made for google colab sessions, which required frequent reinstallation of libraries in bash. Certain code boxes may need to be edited/deleted.

1. Seq2Seq Audio
2. Seq2Seq Audio + GAN
3. Seq2Seq Audio + Wasserstein GAN
4. Seq2Seq Audio (Unoptimised (for loops))
5. LSTM RNN Audio (Many to One)