A PyTorch implementation of SampleRNN: An Unconditional End-to-End Neural Audio Generation Model.

It's based on the reference implementation in Theano: Unlike the Theano version, our code allows training models with arbitrary number of tiers, whereas the original implementation allows maximum 3 tiers. However it doesn't allow using LSTM units (only GRU). For more details and motivation behind rewriting this model to PyTorch, see our blog post:


This code requires Python 3.5+ and PyTorch 0.1.12+. Installation instructions for PyTorch are available on their website: You can install the rest of the dependencies by running pip install -r requirements.txt.


We provide a script for creating datasets from YouTube single-video mixes. It downloads a mix, converts it to wav and splits it into equal-length chunks. To run it you need youtube-dl (a recent version; the latest version from pip should be okay) and ffmpeg. To create an example dataset - 4 hours of piano music split into 8 second chunks, run:

cd datasets
./ "" 8 piano

You can also prepare a dataset yourself. It should be a directory in datasets/ filled with equal-length wav files. Or you can create your own dataset format by subclassing It's easy, take a look at dataset.FolderDataset in this repo for an example.


To train the model you need to run All model hyperparameters are settable in the command line. Most hyperparameters have sensible default values, so you don't need to provide all of them. Run python -h for details. To train on the piano dataset using the best hyperparameters we've found, run:

python --exp TEST --frame_sizes 16 4 --n_rnn 2 --dataset piano

The results - training log, loss plots, model checkpoints and generated samples will be saved in results/.

We also have an option to monitor the metrics using CometML. To use it, just pass your API key as --comet_key parameter to