Sven Mika
5254d2fb36
[RLlib] Support parallelizing evaluation and training (optional). ( #15040 )
2021-04-13 09:53:35 +02:00
Clark Zinzow
05d99c9432
[dask-on-ray] Don't leak a global enabling of client mode in Dask callback test. ( #15257 )
...
* Don't leak a global enabling of client mode in Dask callback test.
* Enable and disable client_mode_enabled, not _client_hook_enabled.
2021-04-12 22:00:30 -07:00
SangBin Cho
9197552802
Temporarily disable flaky tests. ( #15250 )
2021-04-12 14:09:07 -07:00
Simon Mo
8bf4b37877
[Hotfix] Pin dm-tree package version ( #15249 )
...
`dm-tree` released a new version https://pypi.org/project/dm-tree/#history
and it depends on `bazel` to build from source, and it conflicts with
our current bazel setup (this conflict is non-trivial to fix).
2021-04-12 13:50:49 -07:00
Clark Zinzow
95659987a4
[Core] Event loop instrumentation - manual instrumentation hooks, instrumentation for deadline timer and local stream socket. ( #15144 )
...
* Added manual hooks in event loop instrumentation.
* Added instrumentation of the deadline timer in the periodical runner.
* Added instrumentation of the local stream socket in the ClientConnection.
* Addressed feedback except for opaque handle.
* Switch to opaque stats handle API.
* Add opaque stats handle destructor check to ensure that RecordExecution is called.
* Revert "Add opaque stats handle destructor check to ensure that RecordExecution is called."
This reverts commit 62cf8fca670d78c1160f0a9526b6cbe6e3a25725.
* Apply suggestions from code review
Co-authored-by: SangBin Cho <rkooo567@gmail.com>
* Other feedback, fixes for code suggestions.
* Prevent handler stats from leaking queueing stats when handler execution is never recorded.
* Enable event loop instrumentation.
* Revert "Enable event loop instrumentation."
This reverts commit df90c504e45e1963dc2ef6c3197dc5c965bc19e7.
* Reorg GCS client and IO context member fields to prevent use-after-free.
Co-authored-by: SangBin Cho <rkooo567@gmail.com>
2021-04-12 13:39:00 -07:00
Ameer Haj Ali
580b2bb9cc
[metrics.Histogram] improve error message ( #15240 )
2021-04-12 21:28:21 +03:00
Tao Wang
4c9eee609c
Revert "Revert "[GCS]Increase heartbeat interval to reduce pressure o… ( #15207 )
...
* Revert "Revert "[GCS]Increase heartbeat interval to reduce pressure on gcs server (#14203 )" (#15194 )"
This reverts commit a9ac4ad890
2021-04-12 10:45:42 -07:00
Richard Liaw
56c95075d1
Revert "[tune] enable mnist test v3 ( #15198 )" ( #15242 )
...
This reverts commit d913f32126
2021-04-12 09:27:55 -07:00
Hao Chen
10ff2f3b4a
Fix duplicate destruction of CoreWorkerProcess instance ( #15245 )
2021-04-12 21:01:21 +08:00
qicosmos
e54dfd8cc5
[C++ worker] Ray actor task for RAY_REMOTE ( #15039 )
2021-04-12 15:40:35 +08:00
Sven Mika
9c5a0cfd7a
[RLlib] Issue 14385: Policy.compute_actions_from_input_dict does not properly track accessed fields for Policy's view requirements. ( #14386 )
2021-04-11 18:20:04 +02:00
Sven Mika
dfc116ea27
[RLlib] Discussion 681: Metrics prepends newest episodes instead of appending. ( #15236 )
2021-04-11 15:31:43 +02:00
Sven Mika
1c9701e9cb
[RLlib] Discussion 1513: on_episode_step() callback called after very first reset (should not). ( #15218 )
2021-04-11 13:16:17 +02:00
Sven Mika
b267f1f1ba
[RLlib] Add support for Int-Box action spaces. ( #15012 )
2021-04-11 13:16:01 +02:00
Clark Zinzow
1b62e9f844
[dask-on-ray][client] Support ClientObjectRefs in the Dask-on-Ray scheduler. ( #15237 )
...
* Support ClientObjectRefs in the Dask-on-Ray scheduler.
2021-04-11 10:44:38 +03:00
Qing Wang
0f444b1a59
Fix unexpected error when handling the process that has exited in memory monitor. ( #14932 )
...
Co-authored-by: Qing Wang <jovany.wq@antgroup.com>
2021-04-11 00:57:10 +08:00
Richard Liaw
0136ae10f8
[tune] run new test ( #15119 )
...
* add-runtest
Signed-off-by: Richard Liaw <rliaw@berkeley.edu>
* ok
Signed-off-by: Richard Liaw <rliaw@berkeley.edu>
2021-04-10 01:12:51 -07:00
chenk008
6709560ef6
fix setproctitle break /proc/PID/environ ( #15056 )
...
* fix setproctitle break /proc/PID/environ
* bugfix
* add ut
* fix ut
* fix ut
* fix ut
* improve comment
* improve comment
* fix ut lint
* fix ut lint
* revert init.py
Co-authored-by: wuhua.ck <wuhua.ck@alibaba-inc.com>
2021-04-09 15:45:19 -07:00
Sam O
be62444bc5
[Log monitor] Resolves the stacktrace ( #15199 )
2021-04-09 11:32:04 -07:00
Jayce Li
b2f9c48647
doc: Fix ray serve in Deploying on Kubernetes ( #15208 )
2021-04-09 09:08:05 -07:00
Siyuan (Ryans) Zhuang
af9e38fd1c
Cloudpickle workaround for false positive cases ( #15202 )
...
* Cloudpickle workaround for false positive cases in '_is_parametrized_type_hint'.
* update comments
2021-04-09 02:22:46 -07:00
Richard Liaw
d913f32126
[tune] enable mnist test v3 ( #15198 )
2021-04-09 00:10:12 -07:00
Dmitri Gekhtman
58fbb419ea
[client][rllib] Add client_mode_hook for ray.get_gpu_ids ( #15185 )
2021-04-08 23:36:11 -07:00
Eric Liang
268409b6ad
updat warning ( #15200 )
2021-04-08 17:56:52 -07:00
Stephanie Wang
94e592004e
Prioritize worker requests for objects over queued task arguments ( #15157 )
2021-04-08 14:51:21 -07:00
Dmitri Gekhtman
4289fa8d43
[kubernetes][autoscaler][test] Kubernetes scale tests ( #15133 )
2021-04-08 11:42:53 -07:00
Eric Liang
982558a4d3
Update ray client protocol version ( #15184 )
2021-04-08 11:38:48 -07:00
SangBin Cho
a88d20729a
[Test] Skip TestConvergenceOptuna temporarily ( #15197 )
...
* Skip flaky tune test temporarily.
* Lint.
2021-04-08 11:36:10 -07:00
Edward Oakes
06f0c0b6a2
[serve] Remove test_api.py::test_shard_key ( #15195 )
2021-04-08 10:50:17 -07:00
SangBin Cho
a9ac4ad890
Revert "[GCS]Increase heartbeat interval to reduce pressure on gcs server ( #14203 )" ( #15194 )
...
This reverts commit ef195e5108
2021-04-08 09:29:13 -07:00
SangBin Cho
bd58a9a9ff
[Build] Fix symbol problems ( #15187 )
2021-04-08 09:11:15 -07:00
Edward Oakes
e4ca337e16
[serve] Change remaining tests to use deployment API ( #15167 )
2021-04-08 08:15:38 -05:00
Tomas Bures
72dbaab0c1
Bugfix for set_verbosity ignoring certain inputs ( #15088 )
...
The function set_verbosity would ignore verbosity levels given as the enum literal of Verbosity. For example `set_verbosity(Verbosity.V1_EXPERIMENT)`. This patch fixes the problem.
2021-04-08 10:59:29 +01:00
Alex Wu
e5feaee95a
[core worker] Disable async connections ( #15161 )
2021-04-07 22:32:04 -07:00
SangBin Cho
61d120557d
[Pubsub] Generalize pubsub, Move pubsub code to pubsub_lib module ( #15164 )
...
* cherry-pick-1
* cherry-pick-2
* cherry-pick-part-3
* Should work.
* Lint fix.
* Fix lint 2.
2021-04-07 20:40:39 -07:00
Tao Wang
ef195e5108
[GCS]Increase heartbeat interval to reduce pressure on gcs server ( #14203 )
2021-04-08 11:14:43 +08:00
Simon Mo
b125cd95b4
[Serve] Cleanup ThreadProxiedRouter and strictly scope endpoint to router ( #14911 )
2021-04-07 16:59:24 -07:00
architkulkarni
aa2f60b3d5
[Serve] [Doc] Add section on integrating with model registries ( #15139 )
2021-04-07 16:19:52 -05:00
Simon Mo
f0ba15c2d3
[Serve] Check for endpoint in the controller ( #15087 )
2021-04-07 14:16:19 -07:00
Edward Oakes
259813d8bd
[serve] Add list_deployments API ( #15152 )
2021-04-07 15:36:08 -05:00
Dmitri Gekhtman
158c3aa26f
[client] support ray.get_gpu_ids() ( #15173 )
...
since this is running on the driver, it will just return an empty list `[]`
2021-04-07 23:02:07 +03:00
Edward Oakes
59cdc30615
[packaging] Remove stub_file deprecation path ( #15171 )
2021-04-07 14:04:25 -05:00
Edward Oakes
eed34092f2
[serve] Defer building starlette request to backend replica ( #15169 )
2021-04-07 13:28:53 -05:00
Edward Oakes
195f818224
[serve] Move route to deployment options ( #15124 )
2021-04-07 10:50:11 -05:00
Edward Oakes
70195e2114
[serve] Make serve.start() idempotent ( #15148 )
2021-04-07 08:51:37 -05:00
Edward Oakes
65f0bae47f
[serve] Default deployment name to callable name ( #15121 )
2021-04-07 08:48:27 -05:00
SangBin Cho
1df60a61a2
Skip serve test that is not working in the master ( #15163 )
2021-04-06 23:17:22 -07:00
Dmitri Gekhtman
d0c83d5ecd
[kubernetes][autoscaler] Underscore for rsync file ( #15155 )
2021-04-06 23:04:31 -07:00
SangBin Cho
e0872083b8
[Pubsub] Generalize pubsub impl part 1 ( #15116 )
...
* Finished the implementation. Cpp tests are left.
* Fix cpp tests.
* Addressed code review.
* Addressed code review.
* Change the destruction order.
* Addressed code review part 2.
2021-04-06 20:59:32 -07:00
SangBin Cho
e624476b78
[Test] skip failing tests in test_actor_resources. ( #15151 )
2021-04-06 20:57:03 -07:00
