gryf/coach
1
0
Fork 0
mirror of https://github.com/gryf/coach.git synced 2026-08-27 22:21:52 +02:00
Code Issues Projects Releases Wiki Activity
511 Commits 4 Branches 2 Tags
59e08034c685c2b9c9d649ee0f00a90b8b6eed94
Commit Graph
84 Commits
This Branch
This Branch
All Branches
Author SHA1 Message Date
Gal Leibovich e288a552dd Update requirements.txt (#422) 2019-10-28 18:30:48 +02:00
Gal Leibovich 66fada7f78 Remove assertion from BatchRLGraphManager 2019-10-22 11:54:14 +03:00
Gal Leibovich 79a4161eca Workaround for dumping gifs through the Python API (#405) 2019-09-26 12:21:25 +03:00
Gal Leibovich c7949d7011 Fix Atari Schedule Heatup 2019-09-08 16:57:38 +03:00
Gal Leibovich 138ced23ba RL in Large Discrete Action Spaces - Wolpertinger Agent (#394) 
* Currently this is specific to the case of discretizing a continuous action space. Can easily be adapted to other case by feeding the kNN otherwise, and removing the usage of a discretizing output action filter
 2019-09-08 12:53:49 +03:00
Gal Leibovich c1d1fae342 Distiller's AMC induced changes (#359) 
* override episode rewards with the last transition reward

* EWMA normalization filter

* allowing control over when the pre_network filter runs
 2019-08-05 10:24:58 +03:00
Gal Leibovich 718597ce9a Fixes to Batch RL tutorial (#378) 2019-07-16 11:22:42 +03:00
Gal Leibovich 19ad2d60a7 Batch RL Tutorial (#372) 2019-07-14 18:43:48 +03:00
Gal Leibovich 587b74e04a Remove double call to reset_internal_state() on gym environments (#364) 2019-07-02 13:43:23 +03:00
Gal LeibovichandJames Casbon d6795bd524 batchnorm fixes + disabling batchnorm in DDPG (#353) 
Co-authored-by: James Casbon <casbon+gh@gmail.com>
 2019-06-23 11:28:22 +03:00
Gal Leibovich 7eb884c5b2 TD3 (#338) 2019-06-16 11:11:21 +03:00
Gal Leibovich a1bb8eef89 DDPG Critic Head Bug Fix (#344) 
* A bug fix for DDPG, where the update to the policy network was based on the sum of the critic's Q predictions on the batch instead of their mean
 2019-06-05 17:47:56 +03:00
Gal Leibovich 4c996e147e applying filters for a csv loaded dataset + some bug-fixes in data loading (#319) 2019-05-28 15:44:55 +03:00
Gal Leibovich 251dc9ccc0 Preset dependent number of csv read attempts in golden testing (#334) 2019-05-28 12:19:57 +03:00
Gal Leibovich 9e9c4fd332 Create a dataset using an agent (#306) 
Generate a dataset using an agent (allowing to select between this and a random dataset)
 2019-05-28 09:34:49 +03:00
Gal Leibovich 30c2b2fc45 moving to skimage.transform.resize (#321) 2019-05-23 13:38:01 +03:00
Gal Leibovich acceb03ac0 bug fixes for OPE (#311) 2019-05-21 16:39:11 +03:00
Gal Leibovich deb0251367 bug fix following PR #191 (#313) 2019-05-12 13:42:45 -07:00
Gal Leibovich 582921ffe3 OPE: Weighted Importance Sampling (#299) 2019-05-02 19:25:42 +03:00
Gal Leibovich 9f625c197b fix for fetch rendering (#297) 
* fix for fetch rendering - removing code which was once required with older gym versions. images are now rendered correctly by default with the latest gym.

* fixing mujoco camera id failure
 2019-04-21 17:37:14 +03:00
Gal Leibovich 4741b0b916 BCQ variant on top of DDQN (#276) 
* kNN based model for predicting which actions to drop
* fix for seeds with batch rl
 2019-04-16 17:06:23 +03:00
Gal Leibovich 310d31c227 integration test changes to reach the train part (#254) 
* integration test changes to override heatup to 1000 steps +  run each preset for 30 sec (to make sure we reach the train part)

* fixes to failing presets uncovered with this change + changes in the golden testing to properly test BatchRL

* fix for rainbow dqn

* fix to gym_environment (due to a change in Gym 0.12.1) + fix for rainbow DQN + some bug-fix in utils.squeeze_list

* fix for NEC agent
 2019-03-27 21:14:19 +02:00
Gal Leibovich 6e08c55ad5 Enabling-more-agents-for-Batch-RL-and-cleanup (#258) 
allowing for the last training batch drawn to be smaller than batch_size + adding support for more agents in BatchRL by adding softmax with temperature to the corresponding heads + adding a CartPole_QR_DQN preset with a golden test + cleanups
 2019-03-21 16:10:29 +02:00
Gal Leibovich abec59f367 fixes to rainbow dqn + a cartpole based golden test (#253) 2019-03-21 12:57:56 +02:00
Gal Leibovich e3c7e526c7 Batch RL (#238) 2019-03-19 18:07:09 +02:00
Gal Leibovich d6158a5cfc restoring from a checkpoint file (#247) 2019-03-17 16:28:09 +02:00
Gal Leibovich 8be9ea5dc9 Update setup.py (#245) 2019-03-12 11:08:10 +02:00
Gal Leibovich c02333b1ba fix dashboard to allow connections from a remote machine. (#231) 2019-03-10 13:15:14 +02:00
Gal Leibovich 9a895a1ac7 bug-fix for l2_regularization not in use (#230) 
* bug-fix for l2_regularization not in use
* removing not in use TF REGULARIZATION_LOSSES collection
 2019-03-03 15:11:06 +02:00
Gal Leibovich 4c914c057c fix for finding the right filter checkpoint to restore + do not update internal filter state when evaluating + fix SharedRunningStats checkpoint filenames (#147) 2018-12-17 21:36:27 +02:00
Gal Leibovich f9ee526536 Fix for issue #128 - circular DQN import (#130) 2018-12-16 16:06:44 +02:00
Gal Leibovich f12857a8c7 Docs changes - fixing blogpost links, removing importing all exploration policies (#139) 
* updated docs

* removing imports for all exploration policies in __init__ + setting the right blog-post link

* small cleanups
 2018-12-05 16:16:16 -05:00
Gal Leibovich 3e281b467b Update docs_raw README.md (#138) 
* Update README.md
 2018-12-03 05:39:17 -08:00
Gal Leibovich 5674749ed5 workaround for resolving the issue of restoring a multi-node training checkpoint to single worker (#156) 2018-11-26 00:08:43 +02:00
Gal Leibovich ab10852ad9 hacky way to resolve the checkpointing issue (#154) 2018-11-25 16:14:15 +02:00
Gal Leibovich 11170d5ba3 fix dist. tf (#153) 2018-11-25 14:02:24 +02:00
Gal Leibovich 2b4c9c6774 Removing grarph_manager param (#141) 2018-11-23 11:42:54 -08:00
Gal Leibovich a1c56edd98 Fixes for having NumpySharedRunningStats syncing on multi-node (#139) 
1. Having the standard checkpoint prefix in order for the data store to grab it, and sync it to S3.
2. Removing the reference to Redis so that it won't try to pickle that in.
3. Enable restoring a checkpoint into a single-worker run, which was saved by a single-node-multiple-worker run.
 2018-11-23 16:11:47 +02:00
Gal Leibovich a112ee69f6 Save filters' internal state (#127) 
* save filters internal state

* moving the restore to be made from within NumpyRunningStats
 2018-11-20 17:21:48 +02:00
Gal Leibovich d4d06aaea6 remove kubernetes dependency (#117) 2018-11-18 18:10:22 +02:00
Gal Leibovich 430e286c56 muting pygame's hello message (#116) 2018-11-18 18:02:55 +02:00
Gal Leibovich ce85c8e8c3 Removing Egreedy from CartPole_ClippedPPO. ClippedPPO's default exploration policy is to be used instead. (#115) 2018-11-18 16:36:34 +02:00
Gal Leibovich 6caf721d1c Numpy shared running stats (#97) 2018-11-18 14:46:40 +02:00
Gal Leibovich 9fd4d55623 Making stop condition optional by using a flag (#113) 
* apply stop condition flag (default: ignore the stop condition)
 2018-11-18 13:37:39 +02:00
Gal Leibovich 449bcfb4e1 summing head losses instead of taking the mean (#98) 2018-11-18 12:20:00 +02:00
Gal Leibovich 49dea39d34 N-step returns for rainbow (#67) 
* n_step returns for rainbow
* Rename CartPole_PPO -> CartPole_ClippedPPO
 2018-11-07 18:33:08 +02:00
Gal Leibovich 5a8da90d32 bug-fix for dumping movies (+ small refactoring and rename 'VideoDumpMethod -> 'VideoDumpFilter') 2018-10-21 17:29:10 +03:00
Gal Leibovich 72ea933384 bug-fix for clipped_ppo not logging several signals + small cleanup 2018-10-02 14:22:37 +03:00
Gal Leibovich 08a557bfd1 updated the benchmarks for space invaders with dueling ddqn variants 2018-09-06 12:13:49 +03:00
Gal Leibovich 5aca3a5ed1 Update README.md 2018-08-30 23:33:44 +03:00
Gal Leibovich d862a3be83 rainbow dqn hyper-parameter updates 2018-08-30 20:41:38 +03:00
Gal Leibovich ebe574e463 add missing hidden layer in rainbow_q_head 2018-08-30 19:34:27 +03:00
Gal Leibovich ea294de7fd adding dueling support for rainbow dqn (now only missing n-step) 2018-08-30 18:15:59 +03:00
Gal Leibovich d2623c0eee bug-fix in dueling dqn 2018-08-30 18:14:53 +03:00
Gal Leibovich bbe7ac3338 Rainbow DQN agent (WIP - still missing dueling and n-step) + adding support for Prioritized ER for C51 2018-08-30 18:14:53 +03:00
Gal Leibovich 1aa2ab0590 parameter noise exploration - using Noisy Nets 2018-08-27 18:19:01 +03:00
Gal Leibovich d826382b11 removing test from Doom_Health_Supreme_DFP + relaxing time limit on reward tests 2018-08-26 18:42:41 +03:00
Gal Leibovich 2021490caa small adjustment to golden tests + fixes for Doom_Health_DFP and Doom_Health_Supreme_DFP 2018-08-26 18:42:41 +03:00
Gal Leibovich 9bb7bd2e9c bug-fix in local_batch_run_coach and rename to run_multiple_seeds 2018-08-23 14:39:11 +03:00
Gal Leibovich a4471389a4 brightened starcraft.gif 2018-08-20 13:50:09 +03:00
Gal Leibovich 904570000a Update README.md 2018-08-20 12:04:29 +03:00
Gal Leibovich 5e275e9795 update starcraft gif 2018-08-20 11:49:19 +03:00
Gal Leibovich c1f428666e bug-fix for checkpointing for single-worker algorithms 2018-08-19 20:17:15 +03:00
Gal Leibovich 23d2945bf8 Update README.md 2018-08-19 11:02:45 +03:00
Gal Leibovich 8f99409387 updating algorithms.png for README 2018-08-16 16:46:26 +03:00
Gal Leibovich ab5a81c7ee fix for dumping movies, without rendering, for pendulum_with_goals 2018-08-14 18:13:44 +03:00
Gal Leibovich e783157b15 Update README.md 2018-08-14 16:16:41 +03:00
Gal Leibovich 7a76d63da4 Update README.md 2018-08-13 17:19:47 +03:00
Gal Leibovich 2807c29f27 fix for measurements in the initial state (fix for DFP) 2018-05-29 16:47:38 +03:00
Gal Leibovich 16c5032735 fix for tensorboard visualization slowing execution even when it is off 
apparently tensorflow still collect summary data even when no summary FileWriter is defined.
 2018-02-18 16:35:24 +02:00
galleibo-intel 4025496783 Setting tensorflow-gpu version to 1.4.1 (1.5.0 is not tested yet) 2018-02-05 15:48:00 +02:00
Gal Leibovich 7c8962c991 adding support in tensorboard (#52) 
* bug-fix in architecture.py where additional fetches would acquire more entries than it should
* change in run_test to allow ignoring some test(s)
 2018-02-05 15:21:49 +02:00
galleibo-intel 3c330768f0 Fix for NEC not saving the DND when saving a model 2017-11-09 19:13:23 +02:00
galleibo-intel f47b8092af fix for intel optimized tensorflow on distributed runs + adding coach_env to .gitignore 2017-11-06 19:41:32 +02:00
Gal Leibovich eb0b57d7fa Updating PPO references per issue #11 2017-10-24 16:57:44 +03:00
Gal Leibovich 1a09b7cec3 changing python to python3 everywhere to make the supported version of python explicit 2017-10-23 13:07:54 +03:00
Gal Leibovich 16604b11df Update install.sh for missing AEON dependencies 
AEON (nervana data loader) is being installed as part of the neon installation
 2017-10-23 11:50:41 +03:00
Gal Leibovich 7c5d0ed01b Update install.sh to install virtualenv when installing neon 
this is required for a case the user asks the user *not* to install Coach in a virtualenv (and then the installer does not install virtualenv, which is required for neon)
 2017-10-23 10:51:48 +03:00
Gal Leibovich 2a3a6f4a68 small fix in neon installation 2017-10-22 16:43:02 +03:00
Gal Leibovich cc9580a949 updated docs with links to github + a few more words on Dashboard functionality 2017-10-22 16:33:49 +03:00
Gal Leibovich c3501653f7 Update README.md 2017-10-22 09:09:37 +03:00
Gal Leibovich 79bb44d5be Update README.md with a link to Coach documentation. 2017-10-20 14:26:07 +03:00
Gal Leibovich 8c708820a9 docs update + removing unused code from parallel_actor 2017-10-19 17:07:30 +03:00
Gal Leibovich 1d4c3455e7 coach v0.8.0 2017-10-19 13:10:15 +03:00