Commit graph

1153 commits

Author SHA1 Message Date
Stephanie Wang
15486a14a0 Refactor actor task queues (#1118)
* Refactor add_task_to_actor_queue into queue_actor_task and insert_actor_task_queue

* Refactor actor task queue to share the waiting task queue

* Fix
2017-10-13 20:52:11 -07:00
Eric Liang
79ea205b3e [rllib] Initial work on integrating hyperparameter search tool (#1107)
* clean up train

* update

* update train script

* add tuned examples

* add agent catalog

* add tune lib

* update

* fix

* testS

* remove

* train docs

* comments

* todo

* fix resource parsing

* fix cr test

* add test

* try to fix travis test
2017-10-13 16:18:16 -07:00
Robert Nishihara
486cb64e3f Compile with -Werror and -Wall (#1116)
* Compile global scheduler with -Werror -Wall.

* Compile plasma manager with -Werror -Wall.

* Compile local scheduler with -Werror -Wall.

* Compile common code with -Werror -Wall.

* Signed/unsigned comparisons.

* More signed/unsigned fixes.

* More signed/unsigned fixes and added extern keyword.

* Fix linting.

* Don't check strict-aliasing because Python.h doesn't pass.
2017-10-12 21:00:23 -07:00
Stephanie Wang
3764f2f2e1 Actor checkpointing with object lineage reconstruction (#1004)
* Worker reports error in previous task, actor task counter is incremented after task is successful

* Refactor actor task execution

- Return new task counter in GetTaskRequest
- Update worker state for actor tasks inside of the actor method
  executor

* Manually invoked checkpoint method

* Scheduling for actor checkpoint methods

* Fix python bugs in checkpointing

* Return task success from worker to local scheduler instead of actor counter

* Kill local schedulers halfway through actor execution instead of waiting for all tasks to execute once

* Remove redundant actor tasks during dispatch, reconstruct missing dependencies for actor tasks

* Make executor for temporary actor methods

* doc

* Set default argument for whether the previous task was a success

* Refactor actor method call

* Simplify checkpoint task submission

* lint

* fix philipp's comments

* Add missing line

* Make actor reconstruction tests run faster

* Unimportant whitespace.

* Unimportant whitespace.

* Update checkpoint method signature

* Documentation and handle exceptions during checkpoint save/resume

* Rename get_task message field to actor_checkpoint_failed

* Fix bug.

* Remove debugging check, redirect test output
2017-10-12 09:53:32 -07:00
Robert Nishihara
b585001881 When a task is passed to the global scheduler, if it is not received,… (#1106)
* When a task is passed to the global scheduler, if it is not received, then try again.
* Call give_task_to_global_scheduler directly (same with local).
2017-10-12 00:04:38 -07:00
Philipp Moritz
6109cc0782 Add instructions to clean the source tree (#1114)
* add instructions to clean the source tree

* update
2017-10-11 23:13:07 -07:00
Robert Nishihara
7a954f4b5f Use monotonic clock for some python tests. (#1112) 2017-10-11 19:58:59 -07:00
Robert Nishihara
9f1e385335 Return errno from handle_sigpipe. (#1051) 2017-10-11 18:36:28 -07:00
Richard Liaw
379b0604b4 Fixing MeanSTDFilter (#1101)
* Fixing MeanSTDFilter

* documentation
2017-10-11 18:16:08 -07:00
Peter Schafhalter
46f6c163dc Converted ClientConnection to C++ standard library (#1099) 2017-10-11 11:12:15 -07:00
Stephanie Wang
1e0ab3d386 Switch to monotonic clock (#1100) 2017-10-10 22:35:21 -07:00
Philipp Moritz
0684258d2e Update arrow to include pandas serialization (#1102)
* update arrow to include pandas serialization

* update
2017-10-10 22:16:35 -07:00
Eric Liang
b1660c4edf [rllib] Refactor to support passing custom env_creator function (#1096)
* refactor to use env creator

* doc

* lint
2017-10-10 12:49:42 -07:00
Robert Nishihara
1837824881 Don't install boost when building on Linux. (#1095) 2017-10-09 23:39:03 -07:00
Robert Nishihara
8f1a73f041 Allow Ray to be built without UI by setting INCLUDE_UI=0. (#1094)
* Allow building Ray without UI by setting INCLUDE_UI=0.

* Fix bash.

* Fix linting.
2017-10-09 23:32:38 -07:00
Robert Nishihara
3944e9a450 Upload wheels to S3 from Travis. (#1097)
* Upload wheels to S3 from Travis.

* Make wheels publicly readable and also only deploy from wheel builds.
2017-10-09 23:10:58 -07:00
Eric Liang
90013eda2d [rllib] Fix docs to reference new code locations (#1092)
* fix rllib docs

* Update example-a3c.rst
2017-10-09 22:58:58 -07:00
Robert Nishihara
a52a1e893f Automatically set CUDA_VISIBLE_DEVICES when worker gets task. (#1044)
* Automatically set CUDA_VISIBLE_DEVICES when worker gets task.

* Add test.
2017-10-06 18:38:08 -07:00
Robert Nishihara
4669c59fa8 Release GPU resources as soon as an actor exits. (#1088)
* Release GPU resources as soon as an actor exits.

* Add a test.

* Store local_scheduler_id and driver_id in the worker object instead of the actor object.
2017-10-06 17:58:19 -07:00
Stephanie Wang
aebe9f9374 Fix actor garbage collection by breaking cyclic references (#1064)
* Fix bug in wait_for_pid_to_exit, add test for actor deletion.

* Fix actor garbage collection by breaking cyclic references

* Add test for calling actor method immediately after actor creation.

* Fix bug, must dispatch tasks when workers are killed.

* Fix python test

* Fix cyclic reference problem by creating ActorMethod objects on the fly.

* Try simply increasing the time allowed for many_drivers_test.py.
2017-10-05 00:55:33 -07:00
Robert Nishihara
971becc905 [rllib] Use NoFilter instead of MeanStdFilter for PPO. (#1082)
* Make NoFilter the default observation filter for PPO.

* Make reward filter NoFilter for PPO.
2017-10-04 21:31:17 -07:00
shane
a72084c568 adding -x flag for better debugging during builds (#1079) 2017-10-04 13:56:14 -07:00
Eric Liang
6ecc899cf2 [rllib] Fix DQN checkpoint/restore and enable test in jenkins (#1063)
* fix dqn restore and add test

* Update .gitignore

* Update test_checkpoint_restore.py

* add checkpoint restore
2017-10-03 23:17:54 -07:00
Mitar
a0d3fb1de1 Fix Arrow's repository URL. (#1072)
Thanks!
2017-10-03 21:40:21 -07:00
Richard Liaw
cb6dea94bc [rllib] Fix Preprocessor for ATARI (#1066)
* Removing squeeze, fix atari preprocessing

* nit comment

* comments

* jenkins

* Lint
2017-10-03 18:45:02 -07:00
Robert Nishihara
0dcf36c91e Switch Arrow commit. (#1068) 2017-10-03 13:56:53 -07:00
Robert Nishihara
c2f5b4c2ed Add a 'Getting Involved' section to Readme. (#1067) 2017-10-03 10:23:47 -07:00
Philipp Moritz
57bd1d6ff5 Specialize Serialization for OrderedDict (#1035)
Specialize Serialization for OrderedDict and defaultdict
2017-10-02 17:33:10 -07:00
Philipp Moritz
b94d85fb5d [rllib] Fix logging to Athena (#1058)
* Fix logging to Athena

* fixes
2017-10-02 17:16:52 -07:00
Robert Nishihara
1488975d1b Add timing statement to loop that calls redis_get_cached_db_client be… (#1045)
* Add timing statement to loop that calls redis_get_cached_db_client because it has been slow in the past.

* Fix linting.

* Refactoring to make manager vectors into std::vector.

* Fix linting.

* Fixes.
2017-10-02 10:46:21 -07:00
Robert Nishihara
a31d138f21 Don't log when a worker can't be started. (#1056) 2017-10-02 10:32:46 -07:00
Richard Liaw
54aaeb25ab fixing (#1048) 2017-10-01 22:34:55 -07:00
Robert Nishihara
2892955119 Pass string into json.loads, not bytes object. (#991) 2017-10-01 14:56:39 -07:00
Robert Nishihara
6c31adf781 Add blog post for 0.2 release. (#1029)
* Draft of 0.2 release blog post.

* Add figure.

* Change date in filename.

* Shrink PNG.

* Add link to plasma store documentation.

* Small updates to blog post.
2017-10-01 14:40:58 -07:00
Robert Nishihara
6d393828db Bump version number to 0.2.1. (#1026) 2017-10-01 12:33:13 -07:00
Robert Nishihara
ad61af7333 Workaround for passing empty list to ray.wait. (#1043)
* Workaround for passing empty list to ray.wait.

* Add test for passing empty list to wait.
2017-10-01 11:45:02 -07:00
Philipp Moritz
79e013e876 upgrade to latest arrow to fix XCode 9 problem (#1042) 2017-09-30 16:24:59 -07:00
Richard Liaw
d482aa5aed Making our landing pages look better (#1022)
* Teaser in the front page of docs

* TEASER

* try changing README

* spacing

* readme

* power words

* readme obj_id change

* obj_ids change for docs

* Updates to readme.

* Update landing page in documentation.

* Fix linting.
2017-09-30 15:37:28 -07:00
Augusto Yao
0d90a17426 Pass cleanup argument to start_monitor. (#1040) 2017-09-30 15:35:25 -07:00
Wapaul1
97b3355adc Register Class Only Creates Entry in Redis Once (#1038)
Don't export the same custom class definition multiple times.
2017-09-30 15:30:27 -07:00
Richard Liaw
16e82b43d1 [rllib] Changes for preprocessors (#1033)
* Changes for preprocessors

* removed comments

* Changes + push for lint

* linted

* adding dependency for travis

* linting won't pass

* reordering

* needed for testing

* added comments

* pip it

* pip dependencies
2017-09-30 13:11:20 -07:00
Alexey Tumanov
2d0f439b7b hugepage + plasma directory support plumbing + documentation (#1030)
* hugepage + plasma directory support plumbing + documentation

* Indentation fix.

* huge_pages_enabled --> huge_pages

* One more change
2017-09-30 09:56:52 -07:00
Robert Nishihara
ce278aa06a Fix valgrind tests. (#1037)
* Comment out local scheduler valgrind test.

* Fix free/delete error.

* More free -> delete errors

* One more free -> delete and also clean up callback state in plasma manager.

* Add set -x to run_valgrind scripts.

* Fix valgrind error in CreateLocalSchedulerInfoMessage.
2017-09-30 00:11:09 -07:00
Eric Liang
ba153adc4c Downgrade severity of most common messages (#1039)
* downgrade severity of most common messages

* update
2017-09-30 00:01:49 -07:00
Eric Liang
b118cef49e [webui] Allow timeline scroll-to-zoom without holding ALT (#993)
* Allow timeline scroll-to-zoom without holding ALT

* Update build_ui.sh

* Update build_ui.sh

* Update build_ui.sh

* Update build_ui.sh

* Retry when getting catapult.
2017-09-29 21:35:12 -07:00
Robert Nishihara
b991dc8900 Add flag for ignoring the UI, don't start UI in jenkins tests. (#1021) 2017-09-29 15:22:51 -07:00
Eric Liang
9f3a4fce50 [rllib] Parallelize sample collection and gradient computation in DQN (#746)
* wip

* works with cartpole

* lint

* fix pg

* comment

* action dist rename

* preprocessor

* fix test

* typo

* fix the action[0] nonsense

* revert

* satisfy the lint

* wip

* wip

* works with cartpole

* lint

* fix pg

* comment

* action dist rename

* preprocessor

* fix test

* typo

* fix the action[0] nonsense

* revert

* satisfy the lint

* Minor indentation changes.

* fix merge

* add humanoid

* initial dqn refactor

* remove tfutil

* fix calls

* fix tf errors 1

* closer

* runs now

* lint

* tensorboard graph

* fix linting

* more 4 space

* fix

* fix linT

* more lint

* oops

* es parity

* remove example.py

* fix training bug

* add cartpole demo

* try fixing cartpole

* allow model options, configure cartpole

* debug

* simplify

* no dueling

* avoid out of file handles

* Test dqn in jenkins.

* Minor formatting.

* lint

* fix py3

* fix issue

* remove chekcpoint

* revert

* Fixit

* sanity check configs

* update cuda

* fix

* parallel gradient computation

* update

* upd

* bug

* upd

* always record training stats

* fix

* comments

* revert assert

* add gpu mask

* fofset

* a tie

* Merge

* fix

* fix

* fix examples

* A3C -> DQN

* fix dqn test

* remove submodule

* fix linting
2017-09-29 00:06:51 -07:00
Peter Schafhalter
10027974b1 Replaced ObjectWaitRequests with unordered map (#990)
* Replaced ObjectWaitRequests with unordered map

* Pass C++ STL object by reference

* Formatting changes and typos.
2017-09-28 15:29:26 -07:00
Eric Liang
19562f6ce5 [rllib] Fix issues with PPO model restoration (#1018)
* fix filter

* add test

* lint

* fix

* commit

* Update a3c.py
2017-09-28 13:12:06 -07:00
Zongheng Yang
427dee511b Fill out specs of the task table in ray_redis_module.cc. (#1024)
* Fill out specs of the task table in ray_redis_module.cc.

* local scheduler field in task table

* linting
2017-09-27 23:45:58 -07:00