google-deepmind/xitari

This is the 0.4 release of the Arcade Learning Environment (ALE), a platform designed for AI research. ALE is based on Stella, an Atari 2600 VCS emulator.

Overview

Xitari is a fork of the Arcade Learning Environment v0.4.

Original Readme.txt from ALE 0.4, with tidy up by Marc G. Bellemare

This is the 0.4 release of the Arcade Learning Environment (ALE), a platform designed for AI research. ALE is based on Stella, an Atari 2600 VCS emulator. More information and ALE-related publications can be found at

http://www.arcadelearningenvironment.org

We encourage you to use the Arcade Learning Environment in your research. In return, we would appreciate if you cited ALE in publications that rely on it (BibTeX entry at the end of this document).

Feedback and suggestions are welcome and may be addressed to any active member of the ALE team.

Enjoy, The ALE team

List of command-line parameters

Execute ./ale -help for more details; alternatively, see documentation available at http://www.arcadelearningenvironment.org.

-random_seed [n] -- sets the random seed; defaults to the current time

-game_controller [fifo|fifo_named|internal] -- specifies how agents interact with ALE; see Java agent documentation for details

-config [file] -- specifies a configuration file, from which additional parameters are read

-output_file [file] -- if set, standard output is redirected to the given file. Do not use in conjunction with -game_controller fifo_named

-run_length_encoding [false|true] -- determine whether run-length encoding is used to send data over pipes; irrelevant when -game_controller internal is set

-max_num_frames_per_episode [n] -- sets the maximum number of frames per episode. Once this number is reached, a new episode will start. Currently implemented on a per-agent basis with internal agents, or for all agents when using pipes (fifo/fifo_named)

Sample agents command-line parameters

These parameters are only relevant when using one of the sample agents under src/agents.

-max_num_episodes [n] -- sets the maximum number of episodes

-max_num_frames [n] -- sets the maximum number of frames (independent of how many episodes are played)

Building

xitari relies on cmake and make.

To compile source code, run:

cmake .
make install

Citing The Arcade Learning Environment: An Evaluation Platform for General Agents

If you use ALE in your research, we ask that you please cite the following.

M. G. Bellemare, Y. Naddaf, J. Veness and M. Bowling. The Arcade Learning Environment: An Evaluation Platform for General Agents, Journal of Artificial Intelligence Research, Volume 47, pages 253-279, 2013.

In BibTeX format:

@ARTICLE{bellemare13arcade,
  author = {{Bellemare}, M.~G. and {Naddaf}, Y. and {Veness}, J. and {Bowling}, M.},
  title = {The Arcade Learning Environment: An Evaluation Platform for General Agents},
  journal = {Journal of Artificial Intelligence Research},
  year = "2013",
  month = "jun",
  volume = "47",
  pages = "253--279",
}

xitari's Issues

Cmake failed

When I run $PREFIX/bin/luarocks make under /tmp/xitari, I got the following error:

CMake Error: Generator: execution of make failed. Make command was: /tmp/xitari/make "cmTryCompileExec69850377/fast"
-- Check for working C compiler: /usr/bin/cc -- broken
CMake Error at /usr/share/cmake-2.8/Modules/CMakeTestCCompiler.cmake:61 (message):
The C compiler "/usr/bin/cc" is not able to compile a simple test program.

It fails with the following output:

Change Dir: /tmp/xitari/CMakeFiles/CMakeTmp

Run Build Command:/tmp/xitari/make "cmTryCompileExec69850377/fast"

Permission denied

Generator: execution of make failed. Make command was: /tmp/xitari/make
"cmTryCompileExec69850377/fast"

Can anyone help me with this?

Trained DQN models

Maybe you guys could publish trained models along with DQN code, it would save the world some energy.

xitari make error

when use luarocks make xitari-0-0.rockspec in win.

[ 50%] Built target ale
[100%] Built target xitari
Install the project...
-- Install configuration: "Release"
-- Up-to-date: C:/torch/luarocks/xitari/0-0/include/xitari/ale_interface.hpp
CMake Error at cmake_install.cmake:56 (file):
file INSTALL cannot find "C:/torch/xitari-master/libxitari.a": No error.

NMAKE : fatal error U1077: “D:\cmake\bin\cmake.exe”: 返回代码“0x1”
Stop.

Error: Build error: Failed installing.

help! please!!!

Problem with running ROMs with improper names

This bug is regarding the DQN demo provided here.

There is an undocumented and undisplayed error in the code that should be addressed. If an improper name of the ROM is used (for example "Breakout" instead of "breakout"), the program simply crashes with a SEGFAULT leaving the user clueless of what they are doing wrong.

The problem is caused by the buildRomRLWrapper function within the games/Roms.cpp file, where the loop is looking through a list of supported ROMs. If the name isn't matched exactly, the method silently returns NULL, causing a crash down the road.

Here are a couple of proposed solutions:

1. Make a note in the README files about the list of supported ROMs and their proper naming convention
2. If possible, make the program output a warning inside the buildRomRLWrapper function before the NULL is returned. Say something like "The ROM XXX isn't supported. Please make sure the name is written correctly!"
3. Fix this entirely, by asking the user to provide the codename (or some kind of ID) of the game, together with the ROM file, instead of relying on the file name itself - which is a poor practice.
4. Use md5sum in a hashmap to determine which game ROM it is that you are trying to load