Ion and Philipp Moritz
d8ae9de99c
Caching task resource requirements. ( #3231 )
...
* caching resource requirements
* small fixes
* avoid copying the resource map
2018-11-05 15:14:09 -08:00
Eric Liang and Richard Liaw
813f51769f
[rllib] Fix rllib rollouts script and add test ( #3211 )
...
## What do these changes do?
Clean up the checkpointing to handle the new checkpoint dirs. Add a test for rollout.py
## Related issue number
https://github.com/ray-project/ray/issues/3206
https://github.com/ray-project/ray/issues/3204
2018-11-05 00:33:25 -08:00
Philipp Moritz and Robert Nishihara
99bac44375
Update CMake to support Mac OS X 10.14 ( #3218 )
2018-11-04 16:32:58 -08:00
xutianming and Philipp Moritz
fb6ac28b44
single sourcing the package version ( #3220 )
2018-11-04 13:53:55 -08:00
Eric Liang and GitHub
369cb833fe
[rllib] Implement custom metrics ( #3144 )
2018-11-03 18:48:32 -07:00
Eric Liang and GitHub
7d69c77a19
[rllib] Decouple ape-x sampling and learning speed
2018-11-03 18:07:39 -07:00
Philipp Moritz and Robert Nishihara
0da15b1c1f
Fix build system dependency for local_scheduler_client ( #3215 )
2018-11-03 13:19:02 -07:00
Eric Liang and Robert Nishihara
9a0f0db070
Add ray stack tool for debugging ( #3213 )
2018-11-03 13:13:02 -07:00
Wang Qing and Robert Nishihara
ca7d4c2cf5
Enable to specify driver id by user. ( #3084 )
2018-11-02 19:01:50 -07:00
Si-Yuan and Philipp Moritz
5ce7ed7dad
Fix 'tempfile' docs ( #3180 )
...
* Fix docs.
* Update doc/source/tempfile.rst
Co-Authored-By: suquark <suquark@gmail.com >
* Remove doc for raylet socket.
2018-11-02 16:50:55 -07:00
Eric Liang and Robert Nishihara
8c03683573
Add warning about using latest wheels ( #3207 )
2018-11-02 15:41:10 -07:00
Robert Nishihara and Philipp Moritz
e495ab5e7c
Fix some paths /tmp/raylogs -> /tmp/ray. ( #3189 )
2018-11-02 12:10:53 -07:00
Robert Nishihara and Philipp Moritz
5822aa2388
Rename get_task -> worker_idle in timeline. ( #3179 )
...
* Rename get_task -> worker_idle in timeline.
* Fix test.
2018-11-02 12:08:46 -07:00
Eric Liang and GitHub
2bef9844bf
Revert "[autoscaler] Also grant roles to worker nodes" ( #3199 )
...
This reverts commit 55d161b49f .
2018-11-01 23:23:06 -07:00
Robert Nishihara and Devin Petersohn
e612e26103
Add use_raylet option for backwards compatibility. ( #3176 )
...
* Add use_raylet option for backwards compatibility.
* Update message.
2018-11-01 14:16:04 -07:00
Robert Nishihara and Philipp Moritz
57d6e98302
Update actor fault tolerance documentation. ( #3175 )
2018-11-01 11:52:05 -07:00
Robert Nishihara and Richard Liaw
60f28040ea
Document fractional resources. ( #3174 )
2018-11-01 10:50:56 -07:00
Eric Liang and GitHub
b2caed9651
[minor] fix a3c pytorch example dim 80 => 84
2018-10-31 22:00:14 -07:00
Eric Liang and GitHub
cd284bb487
[rllib] Document env compatibility, Ape-X support for multi-agent ( #3147 )
2018-10-31 21:59:34 -07:00
Richard Liaw and GitHub
2086a57e61
[tune] Add Fractional GPU example/docs ( #3169 )
...
* Add example for fractional GPU support
* Update tune_mnist_keras.py
* Update doc/source/tune-usage.rst
2018-10-31 18:53:16 -07:00
Robert Nishihara and Philipp Moritz
1f29a960f4
Update task_table and object_table API. ( #3161 )
...
* Update task_table and object_table API.
* Fix
2018-10-31 12:52:50 -07:00
Dennis Chung and Richard Liaw
9df2e6e6f4
[tune] Modify stop criteria in hyperopt example ( #3102 )
...
Modify `training_iteraion` to `timesteps_total` because only `timesteps_total` is inside the reporter.
2018-10-30 13:26:40 -07:00
Stephanie Wang and GitHub
aacbd007a0
[xray] Implement faster flush policy for lineage cache ( #3071 )
...
* Policy that flushes the lineage stash immediately
* Fix bug where remote tasks in uncommitted lineage weren't getting subscribed to, add reg test
* test
* Fix bug where waiting task was getting subscribed
* Cleanup
* Update src/ray/raylet/lineage_cache.cc
Co-Authored-By: stephanie-wang <swang@cs.berkeley.edu >
* Update src/ray/raylet/lineage_cache.cc
Co-Authored-By: stephanie-wang <swang@cs.berkeley.edu >
* cleanup
* cleanup
* Add another test for task with many parents
* fix, unsubscribe to new waiting tasks
* Unsubscribe as soon as the commit notification is handled
2018-10-30 09:59:50 -07:00
Eric Liang and GitHub
a221f55b0d
[rllib] Add custom value functions, fix up and document multi-agent variable sharing ( #3151 )
2018-10-29 19:37:27 -07:00
Robert Nishihara and Eric Liang
e49839c73f
Fix linting. ( #3155 )
2018-10-28 20:43:29 -07:00
Robert Nishihara and Philipp Moritz
32f0d6b77e
Deprecate num_workers argument to ray.init and ray start. ( #3114 )
...
* Remove num_workers argument.
* Fix
* Fix
2018-10-28 20:12:49 -07:00
Robert Nishihara and Philipp Moritz
9868af4c7c
Use /tmp instead of /dev/shm for object store on Linux if /dev/shm is too small. ( #3149 )
...
* Use /tmp instead of /dev/shm for object store on Linux if /dev/shm is too small.
* Add logging statement and address comments.
* Fix
2018-10-28 20:09:06 -07:00
Robert Nishihara and Philipp Moritz
08fc9e5bcd
Add more description to setup.py. ( #3153 )
2018-10-28 19:49:52 -07:00
Robert Nishihara and Philipp Moritz
fd854ff090
Allow the node manager port and object manager port to be set through… ( #3130 )
...
* Allow the node manager port and object manager port to be set through ray start.
* Linting
* Fix Java test
* Address comments.
2018-10-28 17:28:41 -07:00
Eric Liang and GitHub
a404401dc6
Update agent.py to fix lint error
2018-10-28 15:28:08 -07:00
Jones Wong and Eric Liang
d6bf890648
Solve hang caused by ray.get in collect_metrics ( #3096 )
2018-10-28 11:52:18 -07:00
Eric Liang and GitHub
af0c1174cd
[sgd] Merge sharded param server based SGD implementation ( #3033 )
...
This includes most of the TF code used for the OSDI experiment. Perf sanity check on p3.16xl instances: Overall scaling looks ok, with the multi-node results within 5% of OSDI final numbers. This seems reasonable given that hugepages are not enabled here, and the param server shards are placed randomly.
$ RAY_USE_XRAY=1 ./test_sgd.py --gpu --batch-size=64 --num-workers=N \
--devices-per-worker=M --strategy=<simple|ps> \
--warmup --object-store-memory=10000000000
Images per second total
gpus total | simple | ps
========================================
1 | 218
2 (1 worker) | 388
4 (1 worker) | 759
4 (2 workers) | 176 | 623
8 (1 worker) | 985
8 (2 workers) | 349 | 1031
16 (2 nodes, 2 workers) | 600 | 1661
16 (2 nodes, 4 workers) | 468 | 1712 <--- OSDI perf was 1817
2018-10-27 21:25:02 -07:00
Yuhong Guo and Robert Nishihara
befbf78048
Delete empty pubsub keys ( #3146 )
...
We found that there are large amount of pub-sub keys with no content in it (This case is worse when wait-id is used in the key name.).
This logic of deleting empty pub-sub keys from GCS was in legacy ray but not in raylet.
2018-10-27 11:58:39 -07:00
Eric Liang and Richard Liaw
6531eed2d0
[rllib] Better error message when action space dim too high ( #3119 )
2018-10-26 16:55:00 -07:00
Robert Nishihara and Philipp Moritz
658c14282c
Remove legacy Ray code. ( #3121 )
...
* Remove legacy Ray code.
* Fix cmake and simplify monitor.
* Fix linting
* Updates
* Fix
* Implement some methods.
* Remove more plasma manager references.
* Fix
* Linting
* Fix
* Fix
* Make sure class IDs are strings.
* Some path fixes
* Fix
* Path fixes and update arrow
* Fixes.
* linting
* Fixes
* Java fixes
* Some java fixes
* TaskLanguage -> Language
* Minor
* Fix python test and remove unused method signature.
* Fix java tests
* Fix jenkins tests
* Remove commented out code.
2018-10-26 13:36:58 -07:00
Eric Liang and GitHub
055daf17a0
[autoscaler] better message if there are more than 10 key pairs
2018-10-26 12:42:11 -07:00
bibabolynn and Hao Chen
b4614ae69a
[java] customize path of ray.conf ( #3100 )
...
users can add custom path of ray.config by using -Dray.config=/path/to/ray.conf
2018-10-26 13:36:34 +08:00
Philipp Moritz and Robert Nishihara
d3148cc3ab
[SGD] Provide better error message if model graphs have different numbers of variables ( #3139 )
2018-10-25 22:18:10 -07:00
Philipp Moritz and Robert Nishihara
d34516f1f8
Update Gemfile Jekyll version ( #3140 )
2018-10-25 21:43:08 -07:00
Robert Nishihara and Philipp Moritz
5aa29613db
Fix linting errors. ( #3127 )
2018-10-24 16:30:00 -07:00
Eric Liang and GitHub
55d161b49f
[autoscaler] Also grant roles to worker nodes
2018-10-24 13:57:36 -07:00
Hanwei Jin and Hao Chen
7c1fd19fd9
[Java] support python worker command in raylet ( #3092 )
...
<!--
Thank you for your contribution!
Please review https://github.com/ray-project/ray/blob/master/CONTRIBUTING.rst before opening a pull request.
-->
## What do these changes do?
support raylet, which is started by java runManager, to start python default_worker.py .
So when doing local test of java call python task, it helps auto start python worker.
## Related issue number
<!-- Are there any issues opened that will be resolved by merging this change? -->
2018-10-24 20:43:39 +08:00
Robert Nishihara and Philipp Moritz
9c1826ed69
Use XRay backend by default. ( #3020 )
...
* Use XRay backend by default.
* Remove irrelevant valgrind tests.
* Fix
* Move tests around.
* Fix
* Fix test
* Fix test.
* String/unicode fix.
* Fix test
* Fix unicode issue.
* Minor changes
* Fix bug in test_global_state.py.
* Fix test.
* Linting
* Try arrow change and other object manager changes.
* Use newer plasma client API
* Small updates
* Revert plasma client api change.
* Update
* Update arrow and allow SendObjectHeaders to fail.
* Update arrow
* Update python/ray/experimental/state.py
Co-Authored-By: robertnishihara <robertnishihara@gmail.com >
* Address comments.
2018-10-23 12:46:39 -07:00
Robert Nishihara and Eric Liang
9d2e864caf
Fix Python linting error. ( #3113 )
2018-10-22 23:41:42 -07:00
Robert Nishihara and Philipp Moritz
22dd7e0428
Add test for wait reconstruction. ( #3110 )
2018-10-22 23:16:54 -07:00
Eric Liang and GitHub
73a092e08c
update ( #3112 )
2018-10-22 22:55:43 -07:00
Philipp Moritz and Robert Nishihara
8d8b6e5bfa
Retry connections to redis for async and subscribe contexts ( #3105 )
...
This is fixing a problem that @devin-petersohn observed on the windows subsystem for linux.
In theory, redis should be up once the async connect is happening and there should be no retries needed for the async connect. However on the windows subsystem for linux, the async connect was failing even though the synchronous one was working. Maybe windows has a different semantics here than linux.
2018-10-22 22:31:13 -07:00
Richard Liaw and Eric Liang
eff7cb4458
[tune] Fix SearchAlg finishing early ( #3081 )
...
* Fix trial search alg finishing early
* Fix lint
* fix lint
* nit fix
2018-10-22 12:17:13 -07:00
Eric Liang and GitHub
221d1663c1
[rllib] switch to python logger ( #3098 )
...
* logg
* set rllib logger
* comment
* info
* rlib
* comment
* add format
* fix lint
* add file info
* update
* add ts
* lint
* better docs
* fix value error
* soft log level
2018-10-21 23:43:57 -07:00
Richard Liaw and Robert Nishihara
40c4148d4f
Cluster Utilities for Fault Tolerance Tests ( #3008 )
2018-10-20 22:56:29 -07:00