Commit Graph
1840 Commits
Author SHA1 Message Date
Wang QingandRobert Nishihara 3b141b26cd Fix global_state not disconnected after ray.shutdown (#4354) 2019-03-18 16:44:49 -07:00
Kristian HartikainenandRichard Liaw 2a046116ce [tune] Fix _SafeFallbackEncoder type checks (#4238)
* Fix numpy type checks for _SafeFallbackEncoder

* Format changes

* Fix usage of nan_str in _SafeFallbackEncoder
2019-03-18 16:34:56 -07:00
Eric LiangandGitHub 27cd6ea401 [rllib] Flip sign of A2C, IMPALA entropy coefficient; raise DeprecationWarning if negative (#4374) 2019-03-17 18:07:37 -07:00
Richard LiawandGitHub ea5a6f8455 [tune] Simplify API (#4234)
Uses `tune.run` to execute experiments as preferred API.

@noahgolmant

This does not break backwards compat, but will slowly internalize `Experiment`. 

In a separate PR, Tune schedulers should only support 1 running experiment at a time.
2019-03-17 13:03:32 -07:00
markgoodheadandRichard Liaw 20a155d03d [Tune] Support initial parameters for SkOpt search algorithm (#4341)
Similar to the recent change to HyperOpt (#https://github.com/ray-project/ray/pull/3944) this implements both:
1. The ability to pass in initial parameter suggestion(s) to be run through Tune first, before using the Optimiser's suggestions. This is for when you already know good parameters and want the Optimiser to be aware of these when it makes future parameter suggestions.
2. The same as 1. but if you already know the reward value for those parameters you can pass these in as well to avoid having to re-run the experiments. In the future it would be nice for Tune to potentially support this functionality directly by loading previously run Tune experiments and initialising the Optimiser with these (kind of like a top level checkpointing functionality) but this feature allows users to do this manually for now.
2019-03-16 23:11:30 -07:00
Eric LiangandGitHub b513c0f498 [autoscaler] Restore error message for setup 2019-03-16 18:00:37 -07:00
Richard LiawandGitHub 5e95abe63e [tune] Fix performance issue and fix reuse tests (#4379)
* fix tests

* better name

* reduce warnings

* better resource tracking

* oops

* revertmessage

* fix_executor
2019-03-16 13:52:02 -07:00
Eric LiangandGitHub a45019d98c [rllib] Add option to proceed even if some workers crashed (#4376) 2019-03-16 13:34:09 -07:00
Philipp MoritzandRobert Nishihara c5e2c9af4d Build wheels for macOS with Bazel (#4280) 2019-03-15 10:37:57 -07:00
Leon SieversandEric Liang 6b93ec3034 Fixed calculation of num_steps_trained for multi_gpu_optimizer (#4364) 2019-03-14 19:46:02 -07:00
Eric LiangandGitHub 2c1131e8b2 [tune] Add warnings if tune event loop gets clogged (#4353)
* add guards

* comemnts
2019-03-14 19:44:01 -07:00
Yuhong GuoandGitHub becffc6cef Fix checkpoint crash for actor creation task. (#4327)
* Fix checkpoint crash for actor creation task.

* Lint

* Move test to test_actor.py

* Revert unused code in test_failure.py

* Refine test according to Raul's suggestion.
2019-03-14 23:42:57 +08:00
Philipp MoritzandRobert Nishihara 2f37cd7e27 fix wheel building doc (#4360) 2019-03-13 23:11:30 -07:00
Philipp MoritzandRobert Nishihara b0c4e60ffb Build wheels for Linux with Bazel (#4281) 2019-03-13 15:57:33 -07:00
Ameer Haj AliandEric Liang 8a6403c26e [rllib] bug fix: merging --config params with params.pkl (#4336) 2019-03-13 11:26:55 -07:00
Andrew TanandRichard Liaw 87bfa1cf82 [tune] add output flag for Tune CLI (#4322) 2019-03-12 23:56:59 -07:00
Eric LiangandGitHub d5f4698305 [tune] Avoid scheduler blocking, add reuse_actors optimization (#4218) 2019-03-12 23:49:31 -07:00
Stefan PanticandEric Liang 2202a81773 Fix multi discrete (#4338)
* Revert "Revert "[wingman -> rllib] IMPALA MultiDiscrete changes (#3967)" (#4332)"

This reverts commit 3c41cb9b60.

* Fix a bug with log rhos for vtrace

* Reformat

* lint
2019-03-12 20:32:11 -07:00
Eric LiangandGitHub 3c41cb9b60 Revert "[wingman -> rllib] IMPALA MultiDiscrete changes (#3967)" (#4332)
This reverts commit 962b17f567.
2019-03-11 22:51:26 -07:00
Kai YangandStephanie Wang 7ff56ce826 Introduce set data structure in GCS (#4199)
* Introduce set data structure in GCS. Change object table to Set instance.

* Fix a logic bug. Update python code.

* lint

* lint again

* Remove CURRENT_VALUE mode

* Remove 'CURRENT_VALUE'

* Add more test cases

* rename has_been_created to subscribed.

* Make `changed` parameter type of `bool *`

* Rename mode to notification_mode

* fix build

* RAY.SET_REMOVE return error if entry doesn't exist

* lint

* Address comments

* lint and fix build
2019-03-11 14:42:58 -07:00
Andrew TanandRichard Liaw c435013b27 [tune] add-note command for Tune CLI (#4321)
Co-Authored-By: andrewztan <andrewztan12@gmail.com>
2019-03-11 14:16:44 -07:00
Stefan PanticandEric Liang 36cbde651a Add action space to model (#4210) 2019-03-09 19:23:12 -08:00
justinwyangandRobert Nishihara 5adb4a6941 Set _remote() function args and kwargs as optional (#4305) 2019-03-09 16:40:14 -08:00
Richard LiawandGitHub 6630a35353 [tune] Initial Commit for Tune CLI (#3983)
This introduces a light CLI for Tune.
2019-03-08 16:46:05 -08:00
Simon MoandRobert Nishihara 3064fad96b Add ray.experimental.serve Module (#4095) 2019-03-08 16:22:05 -08:00
Eric LiangandGitHub c7f74dbdc7 [rllib] Add async remote workers (#4253) 2019-03-08 15:39:48 -08:00
Robert NishiharaandPhilipp Moritz fd2d8c2c06 Remove Jenkins backend tests and add new long running stress test. (#4288) 2019-03-08 15:29:39 -08:00
Richard LiawandGitHub c3a3360a4a [tune] Add custom field for serializations (#4237) 2019-03-08 11:00:25 -08:00
Kristian HartikainenandRichard Liaw 7e4b4822cf [tune] Fix worker recovery by setting force=False when calling logger sync_now (#4302)
## What do these changes do?
Fixes a tune autoscaling problem where worker recovery causes things to stall.
2019-03-08 10:59:31 -08:00
Philipp MoritzandRobert Nishihara 95254b3d71 Remove the old web UI (#4301) 2019-03-07 23:15:11 -08:00
Yuhong GuoandGitHub b9ea821d16 Use strongly typed IDs in C++. (#4185)
*  Use strongly typed IDs for C++.

* Avoid heap allocation in cython.

* Fix JNI part

* Fix rebase conflict

* Refine

* Remove type check from __init__

* Remove unused constructor declarations.
2019-03-07 21:43:01 +08:00
Eric LiangandGitHub b0332551dd [rllib] Fix APPO + continuous spaces, feed prev_rew/act to A3C properly (#4286) 2019-03-06 21:36:26 -08:00
Hao ChenandGitHub f0465bc68c [Java] Refine tests and fix single-process mode (#4265) 2019-03-07 09:59:13 +08:00
Philipp MoritzandRobert Nishihara 39eed24d47 update version from 0.7.0.dev0 to 0.7.0.dev1 (#4282) 2019-03-06 14:43:09 -08:00
Eric LiangandGitHub 2781d74680 [rllib] Reserve CPUs for replay actors in apex (#4217) 2019-03-06 10:22:12 -08:00
Eric LiangandGitHub 6d705036f3 [rllib] Add callback accessor for raw observation, fix prev actions (#4212) 2019-03-06 10:21:05 -08:00
Eric LiangandGitHub 0e77a8f8c0 [rllib] Add end-to-end tests for RNN sequencing (#4258) 2019-03-06 09:55:07 -08:00
Philipp MoritzandRobert Nishihara ff5e3384ce Update version to 0.7.0.dev1 and update docs 0.6.3 -> 0.6.4 (#4276) 2019-03-05 22:22:29 -08:00
Stephanie WangandPhilipp Moritz b7ebf17650 Fix test (#4264) 2019-03-05 18:37:00 -08:00
Eric LiangandRobert Nishihara 78ad9c4cbb Add "ray timeline" command to auto-dump Chrome trace for the current Ray instance (#4239) 2019-03-05 16:28:00 -08:00
Richard LiawandGitHub a5441a3381 [tune] Fix testTrialNoSave (#4262)
Left a `last_result == None` after changing last_result to always be a
dict.



Fixes https://github.com/ray-project/ray/issues/4259.
2019-03-05 09:28:33 -08:00
Robert NishiharaandPhilipp Moritz fa8c07dd19 Sleep for half a second at exit in order to avoid losing log messages… (#4254) 2019-03-04 20:39:09 -08:00
Eric LiangandGitHub 30bf8e46c7 [rllib] Use nested scope in custom loss example 2019-03-04 18:29:22 -08:00
Kristian HartikainenandRichard Liaw df9beb7123 [tune] Fix trial result fetching (#4219)
* Fix trial results wait in RayTrialExecutor.get_next_available_trial

* Add comment for the results shuffling

* Remove timeout from the wait

* Change random.sample to random.shuffle
2019-03-04 14:26:10 -08:00
Eric LiangandGitHub 6e3384a719 [rllib] Add three new long-running stress tests {APEX, IMPALA, PBT} (#4215) 2019-03-04 14:05:42 -08:00
Stephanie WangandGitHub 8b871af555 Fix ray.wait bug for tasks on remote nodes and timeout=0 (#4242)
* Regression test

* Fix

* cleaner code
2019-03-04 11:46:06 -08:00
Hao ChenandGitHub a22d6ef955 Fix RemoteFunction._last_export_session (#4243) 2019-03-04 19:57:42 +08:00
Yuhong GuoandGitHub 5866fd7005 Add type check in free and change Exception to TypeError (#4221) 2019-03-04 16:40:04 +08:00
Philipp MoritzandRobert Nishihara e96e06e031 bump version to 0.6.4 (#4226) 2019-03-03 14:39:05 -08:00
Adi ZimmermanandRichard Liaw 9551f2a92e [tune] Properly handle closing files in Trainable (#4232)
Fixes #3965.

Using the with keyword/block will close to file immediately after the block ends
2019-03-03 14:23:05 -08:00