/moolib

A library for distributed ML training with PyTorch

Primary LanguageC++MIT LicenseMIT

🐮
moolib

moolib - a communications library for distributed ML training

moolib offers general purpose RPC with automatic transport selection (shared memory, TCP/IP, Infiniband) allowing models to data-parallelise their training and synchronize gradients and model weights across many nodes.

moolib is an RPC library to help you perform distributed machine learning research, particularly reinforcement learning. It is designed to be highly flexible and highly performant.

It is flexible because it allows researchers to define their own training loops and data-collection policies with minimal interference or abstractions - moolib gets out of the way of research code.

It is performant because it gives researchers the power of efficient data-parallelization across GPUs with minimal overhead, in a manner that is highly scalable.

moolib aims to provide researchers with the freedom to implement whatever experiment loop they desire, and the freedom to scale it up from single GPUs to hundreds at will (with no additional code). It ships with a reference implementations IMPALA on Atari that can easily be adapted to other environments or algorithms.

Installing

To compile moolib without CUDA support

EXPORT USE_CUDA=0

To install from GitHub:

pip install git+https://github.com/facebookresearch/moolib

To build from source:

git clone --recursive git@github.com:facebookresearch/moolib
cd moolib
pip install .

Run an Example

To run the example agent on a given Atari level:

First, start the broker:

python -m moolib.broker

It will output something like Broker listening at 0.0.0.0:4431.

Note that a single broker is enough for all your experiments.

Now take the IP address of your computer. If you ssh'd into your machine, this should work (in a new shell):

export BROKER_IP=$(echo $SSH_CONNECTION | cut -d' ' -f3)  # Should give your machine's IP.
export BROKER_PORT=4431

To start an experiment with a single peer:

python -m examples.vtrace.experiment connect=BROKER_IP:BROKER_PORT \
    savedir=/tmp/moolib-atari/savedir \
    project=moolib-atari \
    group=Zaxxon-Breakout \
    env.name=ALE/Breakout-v5

To add more peers to this experiment, start more processes with the same project and group settings, using a different setting for device (default: 'cuda:0').

Documentation

Benchmarks

Show results on Atari

atari_1 atari_2

Citation

@article{moolib2022,
  title  = {{moolib:  A Platform for Distributed RL}},
  author = {Vegard Mella and Eric Hambro and Danielle Rothermel and Heinrich K{\"{u}}ttler},
  year   = {2022},
  url    = {https://github.com/facebookresearch/moolib},
}

License

moolib is licensed under the MIT License. See LICENSE for details.