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
gwern
c1bc07bcac
Generalized Kelly Coinflip game: persist the hyperparameters as well to avoid confusing users
...
before, hyperparameters would get reset to the defaults by _reset since any user-specified pareto/beta/normal hyperparameters weren't saved after the original init; tediously store them and pass them back into init on reset
2017-03-28 10:44:56 -07:00
gwern
a844c3b6de
Generalized Kelly Coinflip game: add a check to init in case anyone accidentally calls init with seed=False, per tlbtlbtlb
2017-03-28 10:44:56 -07:00
gwern
7c813e87c2
Generalized Kelly Coinflip game: right, observation space has to match the sufficient statistics...
2017-03-28 10:44:56 -07:00
gwern
1de6febf46
Generalized Kelly Coinflip game: preserve the random seed across resets
2017-03-28 10:44:56 -07:00
gwern
dc9e36b895
Kelly Coinflip game: add a generalized version with randomized parameters that turns it into a POMDP for deep RL
2017-03-28 10:44:56 -07:00
gwern
61bf52da6f
Kelly Coinflip Game: document
2017-03-28 10:44:56 -07:00
Trevor Blackwell
f19f90452c
bump atari-py
2017-03-28 10:41:20 -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
Szymon Sidor
255a261e5d
bump atari-py requirement
2017-03-22 16:45:10 -07:00
Szymon Sidor
33e96e7341
add hero
2017-03-19 16:14:17 -07:00
andrewschreiber
1b6cb28192
fix: remove duplicate PyOpenGL declaration
2017-03-16 10:51:58 -07:00
Trevor Blackwell
dfd2c475fd
Fencepost bug on kellycoinflip observation space
2017-03-16 10:50:14 -07:00
gwern
7a2adc688a
+toy text game: The Kelly Coinflip Game ( https://www.gwern.net/Coin-flip )
...
this is a gambling game introduced by 'Rational Decision-Making Under Uncertainty: Observed Betting Patterns on a Biased Coin', by Haghani & Dewey 2016; starting from $25, one makes double-or-nothing bets trying to reach the $250 ceiling in 300 rounds
2017-03-16 10:50:14 -07:00
Trevor Blackwell
85b2623a95
Fix FAQ link
2017-03-15 12:38:44 -07:00
Trevor Blackwell
72da2d5a16
Link to FAQ
2017-03-15 12:36:39 -07:00
Trevor Blackwell
c4321c2791
Bump version
2017-03-15 11:55:15 -07:00
Trevor Blackwell
bd36540a19
Clarify atari observation space comment
2017-03-15 11:55:07 -07:00
Trevor Blackwell
cb8ca5a4aa
Address https://github.com/openai/gym/issues/523
2017-03-15 11:54:44 -07:00
Trevor Blackwell
2a0a2a3a2b
Forward spec down the tree, rather than copy at init time, since it can be set after wrapping
2017-03-09 18:16:26 -08:00
Jie Tang
7c7d2f820e
Add retries when creating benchmark runs
2017-03-08 11:45:56 -08:00
Jie Tang
28913dad94
Add retries to benchmark upload
2017-03-06 13:46:37 -08:00
Jie Tang
07d9885471
Tighten up this docstring
2017-03-06 13:46:37 -08:00
Jie Tang
f869ed01fe
Add skip_videos flag for skipping video uploads on large benchmarks
2017-03-06 13:46:37 -08:00
Jie Tang
ffc30879de
Refactor benchmark / evaluation uploads
2017-03-06 13:46:37 -08:00
Jie Tang
b76ea32a46
Revert unused scorer descriptions
2017-03-06 13:46:37 -08:00
Oleg Klimov
d6714824ec
README: link to What’s New, fix two other glitches
2017-03-07 00:13:51 +03:00
John Schulman
ed35158524
simplify renderer closing
2017-03-05 20:11:24 -08:00
Greg Brockman
c16f6c8a23
Bump version
2017-03-05 16:30:17 -08:00
Greg Brockman
3174039cbf
Revert render change for now
2017-03-05 16:30:04 -08:00
Tom Brown
842de12611
Clarify instructions for resolving env.configure related issues
2017-03-05 14:11:43 -08:00
Tom Brown
90a776b4b3
Bump version, note backwards incompatibility
2017-03-05 13:44:39 -08:00
John Schulman
0ee49635f1
add *args, **kwargs to (non-functional) configure
2017-03-05 13:17:32 -08:00
John Schulman
1fb81d4e3f
remove MultiDiscrete adaptors
2017-03-05 13:13:14 -08:00
John Schulman
387f14d381
remove configure, simplify core.py
2017-03-05 12:00:51 -08:00
Jie Tang
6f277090ed
* Add descriptions of scorers
...
* Make null_score a property
* Scorers are proper classes
2017-02-27 19:53:02 -08:00
Falcon Dai
ceda767e34
fix pyglet.gl imports
2017-02-27 18:48:19 -08:00
Falcon Dai
1e75f670ed
move rendering import. fix issue-518
2017-02-27 18:48:19 -08:00
Trevor Blackwell
874b6fe2af
Better error when called on non-discrete action spaces
2017-02-27 10:00:58 -08:00
Trevor Blackwell
8c6468b34d
Don't crash if render called before reset
2017-02-27 10:00:48 -08:00
Toshiki Kataoka
93d554bdbb
fix typo
2017-02-25 18:51:37 -08:00
Alex Ray
f724a72006
Lint fixes to mujoco environment ( #377 )
2017-02-22 17:24:27 -08:00
Tom Brown
6d7e50e67b
Update environments.md
2017-02-22 17:23:07 -08:00
Jie Tang
77568accd7
Thread episode lengths through when scoring, add tests
2017-02-13 12:29:11 -08:00
Jie Tang
f1f13a2de1
* Print local benchmark score
...
* Turn off video recording for benchmark_runner
* Add force argument to benchmark runner
2017-02-13 12:28:48 -08:00
John Schulman
6f4f5653de
switch to pytest ( #495 )
...
* switch to pytest
* remove observation space sampling
* fix test
2017-02-11 22:17:02 -08:00
John Schulman
33a6112e07
don't unit test atari except pong -- it takes too long
2017-02-11 17:52:18 -08:00
John Schulman
1ef3996010
fix to previous commit
2017-02-11 17:43:11 -08:00
