Implementation of Model-Agnostic Meta-Learning (MAML) applied on Reinforcement Learning problems in Pytorch. This repository includes environments introduced in (Duan et al., 2016, Finn et al., 2017): multi-armed bandits, tabular MDPs, continuous control with MuJoCo, and 2D navigation task.
To avoid any conflict with your existing Python setup, and to keep this project self-contained, it is suggested to work in a virtual environment with virtualenv. To install virtualenv:
pip install --upgrade virtualenv
Create a virtual environment, activate it and install the requirements in requirements.txt.
virtualenv venv
source venv/bin/activate
pip install -r requirements.txt
You can use the main.py script in order to run reinforcement learning experiments with MAML. This script was tested with Python 3.5. Note that some environments may also work with Python 2.7 (all experiments besides MuJoCo-based environments).
python main.py --env-name HalfCheetahDir-v1 --num-workers 8 --fast-lr 0.1 --max-kl 0.01 --fast-batch-size 20 --meta-batch-size 40 --num-layers 2 --hidden-size 100 --num-batches 1000 --gamma 0.99 --tau 1.0 --cg-damping 1e-5 --ls-max-steps 15 --output-folder maml-halfcheetah-dir --device cuda
This project is, for the most part, a reproduction of the original implementation cbfinn/maml_rl in Pytorch. These experiments are based on the paper
Chelsea Finn, Pieter Abbeel, and Sergey Levine. Model-agnostic meta-learning for fast adaptation of deep networks. International Conference on Machine Learning (ICML), 2017 [ArXiv]
If you want to cite this paper
@article{DBLP:journals/corr/FinnAL17,
author = {Chelsea Finn and Pieter Abbeel and Sergey Levine},
title = {Model-{A}gnostic {M}eta-{L}earning for {F}ast {A}daptation of {D}eep {N}etworks},
journal = {International Conference on Machine Learning (ICML)},
year = {2017},
url = {http://arxiv.org/abs/1703.03400}
}