/magenta

Magenta: Music and Art Generation with Machine Intelligence

Primary LanguagePythonApache License 2.0Apache-2.0

Magenta is a research project exploring the role of machine learning in the process of creating art and music. Primarily this involves developing new deep learning and reinforcement learning algorithms for generating songs, images, drawings, and other materials. But it's also an exploration in building smart tools and interfaces that allow artists and musicians to extend (not replace!) their processes using these models. Magenta was started by some researchers and engineers from the Google Brain team but many others have contributed significantly to the project. We use TensorFlow and release our models and tools in open source on this GitHub. If you’d like to learn more about Magenta, check out our blog, where we post technical details. You can also join our discussion group.

Getting Started

Installation

Python Pip

Magenta maintains a pip package for easy installation. We recommend using Anaconda to install it, but it can work in any standard Python environment. We support both Python 2 (>= 2.7) and Python 3 (>= 3.5). These instructions will assume you are using Anaconda.

Note that if you want to enable GPU support, you should follow the GPU Installation instructions below.

Automated Install

If you are running Mac OS X or Ubuntu, you can try using our automated installation script. Just paste the following command into your terminal.

curl https://raw.githubusercontent.com/tensorflow/magenta/master/magenta/tools/magenta-install.sh > /tmp/magenta-install.sh
bash /tmp/magenta-install.sh

After the script completes, open a new terminal window so the environment variable changes take effect.

The Magenta libraries are now available for use within Python programs and Jupyter notebooks, and the Magenta scripts are installed in your path!

Note that you will need to run source activate magenta to use Magenta every time you open a new terminal window.

Manual Install

If the automated script fails for any reason, or you'd prefer to install by hand, do the following steps.

First, download the Python 2.7 Miniconda installer (you can skip this step if you already have any variant of conda installed).

Next, create and activate a Magenta conda environment using Python 2.7 with Jupyter notebook support:

conda create -n magenta python=2.7 jupyter
source activate magenta

Install the Magenta pip package:

pip install magenta

Note that in order to install the rtmidi package that we depend on, you may need to install headers for some sound libraries. On Linux, this command should install the necessary packages:

sudo apt-get install build-essential libasound2-dev libjack-dev

The Magenta libraries are now available for use within Python programs and Jupyter notebooks, and the Magenta scripts are installed in your path!

Note that you will need to run source activate magenta to use Magenta every time you open a new terminal window.

GPU Installation

If you have a GPU installed and you want Magenta to use it, you will need to follow the Manual Install instructions, but with a few modifications.

First, make sure your system meets the requirements to run tensorflow with GPU support.

Next, follow the Manual Install instructions, but install the magenta-gpu package instead of the magenta package:

pip install magenta-gpu

The only difference between the two packages is that magenta-gpu depends on tensorflow-gpu instead of tensorflow.

Magenta should now have access to your GPU.

Docker

Another way to try out Magenta is to use our Docker container. First, install Docker. Next, run this command:

docker run -it -p 6006:6006 -v /tmp/magenta:/magenta-data tensorflow/magenta

This will start a shell in a directory with all Magenta components compiled, installed, and ready to run. It will also map port 6006 of the host machine to the container so you can view TensorBoard servers that run within the container.

This also maps the directory /tmp/magenta on the host machine to /magenta-data within the Docker session. Windows users can change /tmp/magenta to a path such as C:/magenta, and Mac and Linux users can use a path relative to their home folder such as ~/magenta. WARNING: only data saved in /magenta-data will persist across Docker sessions.

The Docker image also includes several pre-trained models in /magenta/models. For example, to generate some MIDI files using the Lookback Melody RNN, run this command:

melody_rnn_generate \
  --config=lookback_rnn \
  --bundle_file=/magenta-models/lookback_rnn.mag \
  --output_dir=/magenta-data/lookback_rnn/generated \
  --num_outputs=10 \
  --num_steps=128 \
  --primer_melody="[60]"

NOTE: Verify that the --output_dir path matches the path you mapped as your shared folder when running the docker run command. This example command presupposes that you are using /magenta-data as your shared folder from the example docker run command above.

One downside to the Docker container is that it is isolated from the host. If you want to listen to a generated MIDI file, you'll need to copy it to the host machine. Similarly, because our MIDI instrument interface requires access to the host MIDI port, it will not work within the Docker container. You'll need to use the full Development Environment.

You may find at some point after installation that we have released a new version of Magenta and your Docker image is out of date. To update the image to the latest version, run:

docker pull tensorflow/magenta

Note: Our Docker image is also available at gcr.io/tensorflow/magenta.

Using Magenta

You can now train our various models and use them to generate music, audio, and images. You can find instructions for each of the models by exploring the models directory.

To get started, create your own melodies with TensorFlow using one of the various configurations of our Melody RNN model; a recurrent neural network for predicting melodies.

Playing a MIDI Instrument

After you've trained one of the models above, you can use our MIDI interface to play with it interactively.

We also have created several demos that provide a UI for this interface, making it easier to use (e.g., the browser-based AI Jam).

Development Environment

If you want to develop on Magenta, you'll need to set up the full Development Environment.

The installation has three components. You are going to need Bazel to build packages, TensorFlow to run models, and an up-to-date version of this repository.

First, clone this repository:

git clone https://github.com/tensorflow/magenta.git

Next, install Bazel. We require the latest version, currently 0.4.5.

You will also need to install some required python dependencies. We recommend using a conda environment and installing with pip:

pip install matplotlib scipy bokeh IPython pandas

Finally, install TensorFlow. To see what version of TensorFlow the code currently requires, check the dependency listed in setup.py.

Also, verify that your environment uses a supported Python version: >=2.7 for Python 2 or >=3.5 for Python 3.

After that's done, run the tests with this command:

bazel test //magenta/...

To build and install the pip package from source, follow the pip build instructions. You can also use our build script.

If you want to build and run commands with Bazel, you'll need to run the package that the build step generates. There are two ways to do this. The first option is to look at the output of the build command to find the path to the generated file. For example, if you want to build the melody_rnn_generate script:

$ bazel build //magenta/models/melody_rnn:melody_rnn_generate
INFO: Found 1 target...
Target //magenta/models/melody_rnn:melody_rnn_generate up-to-date:
  bazel-bin/magenta/models/melody_rnn/melody_rnn_generate

$ bazel-bin/magenta/models/melody_rnn/melody_rnn_generate --config=...

The other option is to use the bazel run command, which combines the two steps above. Note that if you use bazel run, you'll need to add an extra -- before the command line arguments to differentiate between Bazel arguments and arguments to the command.

$ bazel run //magenta/models/melody_rnn:melody_rnn_generate -- --config=...