Commit Graph

119 Commits

Author SHA1 Message Date
Amog Kamsetty 92718de40c [SGD] Better support for custom DDP (#11771) 2020-11-04 13:58:51 -08:00
Barak Michener 05c4e3fb2a [build] Build wheels with manylinux2014 (#11621)
* necessary changes

* Split bazel install

* manylinux2014

* change references to manylinux2014

* Fix lint

* port alex's docker build changes

* fix config issue

* remove extra manylinux2010 requirement script

* revert SHA overwrite

* wip

* incompatible_linklibs

* fix nits
2020-11-03 19:36:32 -08:00
Philipp Moritz 39ce0eadbe Ray PDB support (#11739) 2020-11-03 09:49:23 -08:00
Richard Liaw 1b357533b1 [tune] Try to enable PTL, SKlearn tests (#11542) 2020-10-24 01:08:46 -07:00
Clark Zinzow 0979589c7c [dask-on-ray] Convert tuple of object refs to list before ray.get() call. (#11582) 2020-10-23 16:39:22 -07:00
DK.Pino 9f804ade5f [Placement Group]Add get all placement group api (#11460)
* add get all interface for placement group

* add get all interface for placement group

* make it work

* fix lint

* fix lint

* fix comment

* add cpp test

* fix python lint
2020-10-23 11:46:48 -07:00
Amog Kamsetty 38eb61442b [SGD] Callback API for SGD+Tune (#11316)
Co-authored-by: Richard Liaw <rliaw@berkeley.edu>
2020-10-15 15:22:14 -07:00
SangBin Cho 6cb00208f7 [Placement Group] Export bundle reservation check method only once. (#11153)
* Export bundle reservation check method only once.

* Addressed code review.
2020-10-08 16:08:28 -07:00
Amog Kamsetty 1027bfd4b8 [Tune, Ray SGD] Update PTL integrations (#11271) 2020-10-08 13:43:07 -07:00
Sumanth Ratna 14d8826e43 Fix overriden typo (#11227) 2020-10-07 19:11:07 -07:00
Ian Rodney b6314dd15c [metrics] Cleanup unused item in Docstring (#11254)
* Use

* remove unused tag
2020-10-07 11:39:38 -07:00
Sumanth Ratna 0630242522 [Metrics] Use isinstance for type-checking keys of tags for metrics (#11202)
* Use isinstance instead of type for type-checking keys of tags for metrics

* Fix lint error
2020-10-05 14:24:22 -07:00
Amog Kamsetty 6325a973a2 [Ray SGD] Support num_steps continue training (#11142) 2020-10-02 23:43:26 -07:00
Amog Kamsetty 874da9a25f [Ray SGD] LightningModule integration + MNIST Example (#11042)
Co-authored-by: Richard Liaw <rliaw@berkeley.edu>
2020-10-01 20:07:32 -07:00
SangBin Cho 474af36591 [Core] Fix random bugs found. (#11128) 2020-09-30 00:11:57 -07:00
SangBin Cho 1e39c40370 [Placement Group] Capture child tasks by default. (#11025)
* In progress.

* Finished up.

* Improve comment.

* Addressed code review.

* Fix test failure.

* Fix ci failures.

* Fix CI issues.
2020-09-27 19:33:00 -07:00
SangBin Cho 29663d89f1 [Placement Group] Remove warning msg for placement groups. (#11034)
* Done.

* Addressed code review.

* Fixed typo.

* Addressed code review.
2020-09-25 20:53:42 -07:00
SangBin Cho 109481afd9 [Metric] custom metrics refinement (#10861)
* In progress

* In Progress.

* Addressed code review.

* Add unit tests.

* Add a simple doc.

* Fixed test failure.

* Fix all test failures from serve.

* Addressed code review.
2020-09-25 09:10:28 -07:00
Eric Liang 609c1b8acd Start moving ray internal files to _private module (#10994) 2020-09-24 22:46:35 -07:00
Amog Kamsetty 3716fb1ca9 [Ray SGD] Fix hotfix test (#11015) 2020-09-24 14:14:03 -07:00
Amog Kamsetty 07bdf062b9 [Ray SGD] [Hotfix] Worker group hotfix (#11008) 2020-09-24 12:21:30 -07:00
SangBin Cho 5e6b887f2d [Placement Group] Capture Child Task Part 1 (#10968)
* In progress.

* In progers.

* Done.

* Addressed code review.

* Increase timeout to make a test less flaky.

* Addressed code review.

* Addressed code review.
2020-09-24 09:02:03 -07:00
Amog Kamsetty 52e1495e30 [Ray SGD] TorchTrainable pre 0.8.7 deprecation warning (#10984)
* torch trainable add pre 0.8.7 backwards compat

* raise instead

* Update python/ray/util/sgd/torch/torch_trainer.py
2020-09-23 18:19:43 -07:00
Ian Rodney 4c3f09094a [docs] redis-port -> port (#10937) 2020-09-23 17:04:13 -07:00
Amog Kamsetty 7dbd0ff824 fix example (#10964) 2020-09-23 10:33:19 -07:00
Amog Kamsetty d1d4743702 [Ray SGD] FP16 Hotfix (#10931) 2020-09-21 13:10:10 -07:00
SangBin Cho 1fdb7ef6c3 [docs] Placement group documentation (#10555)
Co-authored-by: Richard Liaw <rliaw@berkeley.edu>
2020-09-16 16:07:55 -07:00
Amog Kamsetty d5a7c53908 [Ray SGD] use_local flag + Worker group abstraction (#10539)
Co-authored-by: Richard Liaw <rliaw@berkeley.edu>
2020-09-15 11:58:57 -07:00
Ian Rodney 5bc2ba38fd [docker] Detect CPUs in container correctly (#10507)
Co-authored-by: simon-mo <simon.mo@hey.com>
Co-authored-by: Richard Liaw <rliaw@berkeley.edu>
Co-authored-by: Alex Wu <itswu.alex@gmail.com>
2020-09-13 23:40:48 -07:00
Alex Wu d2778f6fbb [hotfix] accelerator_types (#10725)
* .

* .
2020-09-11 12:43:43 -07:00
Barak Michener c6b1ed7f8f release process: bump version number to 1.1.0.dev0 everywhere (#10686) 2020-09-10 16:00:21 -07:00
Amog Kamsetty 415be78cc0 [RaySGD] Simplify Builder Process (#10321)
Co-authored-by: Richard Liaw <rliaw@berkeley.edu>
2020-09-08 15:19:40 -07:00
Clark Zinzow 36e1f20e9c Add Dask-Ray scheduler callbacks. (#10519)
Improve Dask-on-Ray documentation.

Move to RayCallback(s) namedtuples, and use top-level CBS tuple as source-of-truth for callback methods.
2020-09-08 13:00:58 -07:00
Simon Mo fdd3acd492 Promote ray.experimental.queue to ray.util (#10624) 2020-09-08 12:56:53 -07:00
Alex Wu d6a9f0e2e4 [Core] Accelerator type API (#10561) 2020-09-06 20:58:40 -07:00
Eric Liang 8ee7c182f5 [1.0] move placement groups from experimental to util. Note they are still undocumented. (#10554)
* move files

* Update __init__.py

* remove

* Update __init__.py
2020-09-04 19:01:24 -07:00
Eric Liang da83bbd764 [1.0] Move dask scheduler from experimental to util (#10553)
* move dask

* fix dask
2020-09-04 12:16:32 -07:00
krfricke 06af62ba91 [tune] refactor tune search space (#10444)
* Added basic functionality and tests

* Feature parity with old tune search space config

* Convert Optuna search spaces

* Introduced quantized values

* Updated Optuna resolving

* Added HyperOpt search space conversion

* Convert search spaces to AxSearch

* Convert search spaces to BayesOpt

* Added basic functionality and tests

* Feature parity with old tune search space config

* Convert Optuna search spaces

* Introduced quantized values

* Updated Optuna resolving

* Added HyperOpt search space conversion

* Convert search spaces to AxSearch

* Convert search spaces to BayesOpt

* Re-factored samplers into domain classes

* Re-added base classes

* Re-factored into list comprehensions

* Added `from_config` classmethod for config conversion

* Applied suggestions from code review

* Removed truncated normal distribution

* Set search properties in tune.run

* Added test for tune.run search properties

* Move sampler initializers to base classes

* Add tune API sampling test, fixed includes, fixed resampling bug

* Add to API docs

* Fix docs

* Update metric and mode only when set. Set default metric and mode to experiment analysis object.

* Fix experiment analysis tests

* Raise error when delimiter is used in the config keys

* Added randint/qrandint to API docs, added additional check in tune.run

* Fix tests

* Fix linting error

* Applied suggestions from code review. Re-aded tune.function for the time being

* Fix sampling tests

* Fix experiment analysis tests

* Fix tests and linting error

* Removed unnecessary default_config attribute from OptunaSearch

* Revert to set AxSearch default metric

* fix-min-max

* fix

* nits

* Added function check, enhanced loguniform error message

* fix-print

* fix

* fix

* Raise if unresolved values are in config and search space is already set

Co-authored-by: Richard Liaw <rliaw@berkeley.edu>
2020-09-03 09:06:13 -07:00
Zhe Zhang 4356abeb1c [Doc] Fix errors in ActorPool documentation (#10410)
* Fix errors in ActorPool documentation

1. map() and map_unordered() name
2. The print statement doesn't work

* Correctly change lines

* Address comments from pr
2020-08-31 13:57:10 -07:00
Richard Liaw cb438be146 [core] Move log_to_driver back to public (#10422) 2020-08-29 16:35:14 -07:00
Eric Liang 519354a39a [api] Initial API deprecations for Ray 1.0 (#10325) 2020-08-28 15:03:50 -07:00
Yu Shan 5264f888e4 fix iterable dataset (issue 9899) (#9952) 2020-08-22 19:40:38 -07:00
SangBin Cho 92664249e8 Partially Use f string (#10218)
* flynt. trial 1.

* Trial 1.

* Addressed code review.
2020-08-20 18:21:16 -07:00
Amog Kamsetty 9ff687c093 [SGD][Docs] docs for training/ validation results (#10181) 2020-08-19 17:22:28 -07:00
SangBin Cho 44826878ff [Core] Remove Legacy Raylet Code (#9936)
* Remove a flag and some methods in node manager including HandleDisconnectedActor, ResubmitTask, and HandleTaskReconstruction

* Make actor creator always required + remove raylet transport

* Remove actor reporter + remove FinishAssignedActorCreationTask

* Remove actor tasks.

* Remove finishactortask and switched it to finishactorcreation task

* Remove reconstruction policy.

* Remove lineage cache.

* Formatting.

* Remove actor frontier code.

* Removed build error.

* Revert "Remove reconstruction policy."

This reverts commit 9d25c9bced4da5fbcac5d484d51013345f16513b.

* Recover HandleReconstruction to mark expired objects as failed.
2020-08-06 16:37:50 -07:00
Richard Liaw 0c3b9ebeef [tune/sgd] Document func_trainable and add checkpoint context (#9739)
Co-authored-by: krfricke <krfricke@users.noreply.github.com>
Co-authored-by: Amog Kamsetty <amogkam@users.noreply.github.com>
2020-07-30 09:46:37 -07:00
Richard Liaw f3fdb5c5db [tune] distributed torch wrapper (#9550)
* changes

* add-working

* checkpoint

* ccleanu

* fix

* ok

* formatting

* ok

* tests

* some-good-stuff

* fix-torch

* ddp-torch

* torch-test

* sessions

* add-small-test

* fix

* remove

* gpu-working

* update-tests

* ok

* try-test

* formgat

* ok

* ok
2020-07-26 09:37:22 -07:00
krfricke 9f3570828a [tune] move jenkins tests to travis (#9609)
Co-authored-by: Richard Liaw <rliaw@berkeley.edu>
Co-authored-by: Kai Fricke <kai@anyscale.com>
2020-07-24 21:22:54 -07:00
krfricke ea4797bf38 [RaySGD] revised existing transformer example to work with transformers>=3.0 (#9661)
Co-authored-by: Kai Fricke <kai@anyscale.com>
2020-07-23 10:58:09 -07:00
Philipp Moritz a5f4659d9f Support ray task type checking (#9574) 2020-07-21 19:05:42 -07:00