Commit Graph

22 Commits

Author SHA1 Message Date
Trevor Blackwell
cd963185b9 Revert "Revert "Revert "Revert "Optimize Atari envs""""
This reverts commit f719298cf0.
2017-03-28 10:51:47 -07:00
Trevor Blackwell
f719298cf0 Revert "Revert "Revert "Optimize Atari envs"""
This reverts commit 8f62c0ab91.
2017-03-28 10:50:48 -07:00
Trevor Blackwell
8f62c0ab91 Revert "Revert "Optimize Atari envs""
This reverts commit c8a84d3c1d.
2017-03-28 10:50:09 -07:00
Trevor Blackwell
c8a84d3c1d Revert "Optimize Atari envs"
This reverts commit 3bc4692d12.
2017-03-28 10:49:51 -07:00
Trevor Blackwell
3bc4692d12 Optimize Atari envs
Get screen pixels in a way that doesn’t require swapping the RGB
channels.
SeaquestNoFrameskip used to take 3.5 seconds to do 10k steps, now it
takes 2.6 seconds.
When also using opencv to resize to (84,84,3), it used to take 7.8
seconds and now takes 7.0.
2017-03-28 09:47:08 -07:00
Trevor Blackwell
bd36540a19 Clarify atari observation space comment 2017-03-15 11:55:07 -07:00
Tom Brown
d337f4e571 TimeLimit refactor with Monitor Simplification (#482)
* fix double reset, as suggested by @jietang

* better floors and ceilings

* add convenience methods to monitor

* add wrappers to gym namespace

* allow playing Atari games, with potentially more coming in the future

* simplify example in docs

* Move play out of the Env

* fix tests

* no more deprecation warnings

* remove env.monitor

* monitor simplification

* monitor simplifications

* monitor related fixes

* a few changes suggested by linter

* timestep_limit fixes

* keep track of gym env variables for future compatibility

* timestep_limit => max_episode_timesteps

* don't apply TimeLimit wrapper in make for VNC envs

* Respect old timestep_limit argument

* Pass max_episode_seconds through registration

* Don't include deprecation warnings yet
2017-02-01 13:10:59 -08:00
catherio
061ca333d7 Expose ale lives in info dict 2017-01-13 14:09:56 -08:00
catherio
f412225f9e Revert "atari envs where life loss triggers new episode"
We intended to use this functionality in Universe, but have
decided to keep it within the Universe repo, and expose
ALE lives in the info dict instead (see next commit)
2017-01-13 14:09:20 -08:00
damodei
d626493705 atari envs where life loss triggers new episode 2017-01-12 23:19:22 -08:00
Greg Brockman
66bcbcdab6 Move repeat_action_probability before loading the ROM (#362)
Closes #349
2016-09-29 02:49:48 -07:00
Greg Brockman
2761741ad3 Disable ALE's underlying repeat_action_probability (#354)
* Disable ALE's underlying repeat_action_probability

* Add assertion for repeat_action_probability

* Properly encode for py3
2016-09-21 00:36:56 -07:00
Greg Brockman
e3e735b1ad Add Atari environments with deterministic frameskip 2016-08-25 08:58:09 -07:00
Rafal Jozefowicz
c11cacd50d Remove unnecessary copies from atari environments (#317) 2016-08-24 00:15:27 -07:00
Nathan Daly
b46f30dd4b Fixed quotes in NotInstalled error message (#261)
s/.)'/'.)/ -- the quote accidentally went around too much.
2016-08-06 00:15:12 -07:00
Mohammad Babaeizadeh
f5f260129d exposing save, load, clone and restore states at enviroment level (#229)
* exposing save, load, clone and restore states at enviroment level for more visibility

* add save, load, restore, clone states as a comment
2016-06-28 17:59:53 -07:00
Greg Brockman
8a535ca6f2 Switch to a global PRNG for action/observation spaces (#144)
cf 58e6aa95e5 (commitcomment-17669277)
2016-05-30 18:07:59 -07:00
Greg Brockman
58e6aa95e5 [WIP] add support for seeding environments (#135)
* Make environments seedable

* Fix monitor bugs

- Set monitor_id before setting the infix. This was a bug that would yield incorrect results with multiple monitors.
- Remove extra pid from stats recorder filename. This should be purely cosmetic.

* Start uploading seeds in episode_batch

* Fix _bigint_from_bytes for python3

* Set seed explicitly in random_agent

* Pass through seed argument

* Also pass through random state to spaces

* Pass random state into the observation/action spaces

* Make all _seed methods return the list of used seeds

* Switch over to np.random where possible

* Start hashing seeds, and also seed doom engine

* Fixup seeding determinism in many cases

* Seed before loading the ROM

* Make seeding more Python3 friendly

* Make the MuJoCo skipping a bit more forgiving

* Remove debugging PDB calls

* Make setInt argument into raw bytes

* Validate and upload seeds

* Skip box2d

* Make seeds smaller, and change representation of seeds in upload

* Handle long seeds

* Fix RandomAgent example to be deterministic

* Handle integer types correctly in Python2 and Python3

* Try caching pip

* Try adding swap

* Add df and free calls

* Bump swap

* Bump swap size

* Try setting overcommit

* Try other sysctls

* Try fixing overcommit

* Try just setting overcommit_memory=1

* Add explanatory comment

* Add what's new section to readme

* BUG: Mark ElevatorAction-ram-v0 as non-deterministic for now

* Document seed

* Move nondetermistic check into spec
2016-05-29 09:07:09 -07:00
Greg Brockman
2ae0dbc846 Discard viewer object after render with close=True
Fixes #95
2016-05-15 17:22:38 -07:00
Josh Marlow
9529fa731b Replaced is string comparison with == (#77) 2016-05-09 17:51:04 -07:00
Jonas Schneider
5065950a09 Fix Python3 compat of import dependencies 2016-04-27 18:03:29 -07:00
Greg Brockman
e8f2980603 Initial release. Hello world :). 2016-04-27 08:00:58 -07:00