Commit graph

1510 commits

Author SHA1 Message Date
Eric Liang
7d5ef6d99c
[rllib] Support continuous action distributions in IMPALA/APPO (#4771) 2019-05-16 22:05:07 -07:00
Richard Liaw
9f2645d6ea [tune] Fix CLI test (#4801) 2019-05-16 13:50:03 -07:00
Devin Petersohn
1490a98a71 Bump version to 0.7.0 (#4791) 2019-05-15 22:55:21 -07:00
Richard Liaw
3bbafc7105
[autoscaler] Fix submit (#4782) 2019-05-14 19:52:28 -07:00
Jones Wong
c5161a2c4d [rllib] fix clip by value issue as TF upgraded (#4697)
*  fix clip_by_value issue

*  fix typo
2019-05-13 15:39:25 -07:00
Qing Wang
62c949bbd5 Fix ray stop by killing raylet before plasma (#4778) 2019-05-13 14:53:10 +08:00
Eric Liang
69352e3302
[rllib] Implement learn_on_batch() in torch policy graph 2019-05-12 21:29:58 -07:00
Romil Bhardwaj
004440f526 Dynamic Custom Resources - create and delete resources (#3742) 2019-05-11 20:06:04 +08:00
Eric Liang
351753aae5
[rllib] Remove dependency on TensorFlow (#4764)
* remove hard tf dep

* add test

* comment fix

* fix test
2019-05-10 20:36:18 -07:00
cgraywang
584adb45b8 [tune] Add MXNet Gluon example on CIFAR-10 (#4683) 2019-05-08 21:39:07 -07:00
Adi Zimmerman
28d381373d [tune] Add Ax to Tune (#4731) 2019-05-08 15:54:29 -07:00
Romil Bhardwaj
0421cba4e8 Autoscaler hotfix for #4555. (#4653) 2019-05-08 14:50:52 -07:00
Jacob Beck
28496c8b50 [rllib] Qmix padding patch (#4735)
* Qmix padding patch

* Update qmix_policy_graph.py

* lint errors

* more linting

* Update qmix_policy_graph.py
2019-05-08 14:07:29 -07:00
Devin Petersohn
edb8465910 [ray-core] Initial addition of performance integration testing files (#4325) 2019-05-08 13:40:54 -07:00
Richard Liaw
7f50c96adb
[tune] Reduce sampling API clutter (#4739)
Adds some sugar for tune sampling API (for commonplace sampling idioms).
2019-05-06 17:42:39 -07:00
Eric Liang
71b2dec3b4
[rllib] Fix bounds of space returned by preprocessor.observation_space (#4736) 2019-05-05 18:25:38 -07:00
Si-Yuan
bd00735fe8 Fix tempfile issues (#4605) 2019-05-05 16:06:15 -07:00
Daniel Ho
dca1c25d88 [tune] Fix setup-dev relative path (#4747) 2019-05-05 00:39:07 -07:00
Richard Liaw
f2faf5ce75 [tune] Contributor Guide and Design Page (#4716)
* Move setup script out

* some changes

* Finished Contributor guide

* some comments to the design

* move

* Apply suggestions from code review

Co-Authored-By: richardliaw <rliaw@berkeley.edu>

* sourcecode

* comments
2019-05-05 00:04:13 -07:00
Robert Nishihara
d81e71e297 Enable actor methods to be decorated on the caller side also and get postprocessors. (#4732)
* Allow decorating ray actor methods.

* Add test.

* Add get postprocessors.

* Improve documentation.

* Make it work for remote functions.

* Temporary fix.
2019-05-04 11:53:47 -07:00
Peng Zhenghao
897b35ce36 [tune] fix restore error at tune.run() (#4733) 2019-05-04 02:56:15 -04:00
Adi Zimmerman
36b71d1446 [Tune] Post-Experiment Tools (#4351) 2019-05-04 02:51:26 -04:00
Federico Fontana
78bb26286e Replaced discontinued rnn_cell.BasicLSTMCell with rnn_cell.LSTMCell (#4703)
* Fixed bug in Dirichlet (#4440)

* Replaced deprecated rnn_cell.BasicLSTMCell with rnn_cell.LSTMCell
2019-05-02 13:19:27 -07:00
Andrew Tan
f87235f232 [tune] Example for Tune blog post (#4673) 2019-05-02 13:16:48 -04:00
Andrew Tan
23ae73135e [tune] Tune CLI Fixes (#4659)
What do these changes do?
  Add --limit flag for ls
  Add ordering functionality to --sort flag
  Remove last_result from the names of columns for ls
  Fix weird double quote error messages (\")
2019-04-30 18:21:33 -07:00
Yuhong Guo
448a7bd08d Add lock in fetch_and_execute_function_to_run of import_thread.py (#4718) 2019-04-30 10:47:16 -07:00
Yuhong Guo
4eade036a0
Separate thread locks for worker and function manager. (#4499)
* Separate lock for function manager and worker

* Lint

* Add test case

* Remove print in remote function.

* Remove test and add ray.exit_actor

* Update python/ray/worker.py

Co-Authored-By: guoyuhong <guoyuhong1985@outlook.com>

* Move exit_actor from worker.py to actor.py

* Update actor.py

* Update actor.py
2019-04-29 14:55:37 +08:00
Kristian Hartikainen
69da6d0fc8 [autoscaler] Remove unnecessary apt installations in docker commands (#4577)
* Remove unnecessary apt installations in docker commands

* Add example for different head/worker image in gcp gpu example

* Update gcp gpu example docker image to tf 1.13

* Change the VM sourceImage for gcp/example-full.yaml

* Change the gcp gpu docker VM images for consistency

* Change gcp default project id to be consistent with other examples
2019-04-28 14:58:51 -07:00
Robert Nishihara
e9b351e749 Reduce memory usage of test_simple in test_stress.py. (#4709) 2019-04-28 07:50:23 -07:00
Eric Liang
b1c9ea7ffc
Update test_trial_scheduler.py (#4710) 2019-04-27 23:11:05 -07:00
Daniel Ho
d7d2694b57 [tune] Add config logging functionality to PBT scheduler (#4680) 2019-04-27 19:32:19 -07:00
Romil Bhardwaj
686d4caefe Updates to scheduling objects to support dynamic custom resources (#4465) 2019-04-27 18:45:23 -07:00
Si-Yuan
9ce3039390
Fix webui api (#4686)
* fix webui

* Apply suggestions from code review

lint

Co-Authored-By: suquark <suquark@gmail.com>

* add dependencies for this unittest

* move dependencies to the script file
2019-04-27 15:23:56 +08:00
Sam Toyer
663e92ab3f [rllib] TD3/DDPG improvements and MuJoCo benchmarks (#4694)
* [rllib] Separate optimisers for DDPG actor & crit.

* [rllib] Better names for DDPG variables & options

Config changes:

- noise_scale -> exploration_ou_noise_scale
- exploration_theta -> exploration_ou_theta
- exploration_sigma -> exploration_ou_sigma
- act_noise -> exploration_gaussian_sigma
- noise_clip -> target_noise_clip

* [rllib] Make DDPG less class-y

Used functions to replace three classes with only an __init__ method & a
handful of unrelated attributes.

* [rllib] Refactor DDPG noise

* [rllib] Unify DDPG exploration annealing

Added option "exploration_should_anneal" to enable linear annealing of
exploration noise. By default this is off, for consistency with DDPG &
TD3 papers. Also renamed "exploration_final_eps" to
"exploration_final_scale" (that name seems to have been carried over
from DQN, and doesn't really make sense here). Finally, tried to rename
"eps" to "noise_scale" wherever possible.
2019-04-26 17:49:53 -07:00
Eric Liang
47cca971b5
Don't delete files in rsync up, and also shorten timeout (#4688) 2019-04-25 12:18:42 -07:00
Qing Wang
f39b6747e5 Refactor command line argument parsing with gflags (#4676) 2019-04-24 14:53:07 +08:00
William Ma
c99e3caaca Change resource bookkeeping to account for machine precision. (#4533) 2019-04-23 11:59:53 -07:00
justinwyang
8dfc833a8b Change all instances of JobID to DriverID. (#4431) 2019-04-22 16:28:09 -07:00
Andrew
06c768823c [rllib] train-eval loop implementation for rllib.Trainer class (#4647) 2019-04-21 12:08:04 -07:00
Devin Petersohn
d5df91b031 Bump version to 0.7.0dev3 (#4671) 2019-04-19 17:06:14 -07:00
Vlad Firoiu
39a09fa457 Turn replay into a circular queue. (#4667) 2019-04-19 11:42:00 -07:00
Wang Qing
9d481cc2e6 [hotfix] Missing import breaks Travis builds 2019-04-18 23:12:44 -07:00
Eric Liang
5a562bbf12
[rllib] Fix num_gpus cast and raise error on large batch (#4652) 2019-04-18 15:23:29 -07:00
Eric Liang
6848dfd179
[rllib] Replace ray.get() with ray_get_and_free() to optimize memory usage (#4586) 2019-04-17 20:30:03 -04:00
Eric Liang
3fd9dea721
[rllib] Fix tune.run(Agent class) (#4630)
* update

* Update __init__.py
2019-04-15 09:12:23 -07:00
Richard Liaw
776a7308c8
[tune] Better ASHA defaults (#4623)
## What do these changes do?
Sets ASHA defaults to paper defaults.


## Related issue number


## Linter

- [ ] I've run `scripts/format.sh` to lint the changes in this PR.
2019-04-15 01:45:43 -07:00
Vlad Firoiu
f600591468 Cast MultiCategorical num_outputs to int. (#4629) 2019-04-14 19:51:37 -07:00
Robert Nishihara
967e8aad9d Make def test_submitting_many_actors_to_one less stressful. (#4622) 2019-04-14 12:19:57 -07:00
Andrew Tan
57af1c6819 Update volume size to 100 (#4616) 2019-04-14 11:40:16 -07:00
Zachary Barry
3838548356 Custom SSH socket directories (#4299)
* ssh_control_path added as an auth option.

* revamped default ssh options to take in control path, nodeupdater checks auth config to see if a custom SSH sockets path was specified, otherwise the original hardcoded path is used. control path is now a nodeupdater instance variable

* revert socketdir in auth config and change method for determining dir

* new ssh dir method

* Lint

* ' -> " lint

* changed using USER env to getpass.getuser()
2019-04-13 23:55:41 -07:00