This package is a basic Reinforcement Learning package written in LUA for Torch. It implements some simple environments and learning policies (Policy Gradient and Deep Q Learning). It also can be easily used with the OpenAI Gym package by using lutorpy (example given in the opeanaigym directory).
Tutorials are provided in the tutorials directory
Lua:
- Torch7
- nn, dpnn
- logroll, json, alewrap
For using openAI Gym:
- openai gym
- lutorpy
- In the torch directory: luarocks make
- Install lutorpy and open AI
- lauch the python script (example.py)
WARNING : If you use an openAI Gym ATARI environment, a new sensor must be developped: it will be avaiable in the next few days (since openAI and alewrap do not store the ATARI images in the same format)
Author: Ludovic DENOYER