Commit Graph
96 Commits
Author SHA1 Message Date
55ae567f7a [tune] Fix and enable SigOpt tests (#12877)
Co-authored-by: Richard Liaw <rliaw@berkeley.edu>
2020-12-18 01:33:12 -08:00
Kai FrickeandGitHub 426f8a8d15 [tune] Fix tutorial training on GPU (#12914) 2020-12-18 01:31:40 -08:00
3d72000826 [tune] Add points_to_evaluate to BasicVariantGenerator (#12916)
Co-authored-by: Richard Liaw <rliaw@berkeley.edu>
2020-12-17 19:16:03 -08:00
ea1228074d [tune] enable points_to_eval for all search algorithms (#12790)
Co-authored-by: Richard Liaw <rliaw@berkeley.edu>
2020-12-15 11:51:53 -08:00
5f04ade6ef [tune] add more stoppers and stopper documentation (#12750)
* Add new stoppers & docs

* Add tests for maximum iteration stopper and trial plateau stopper

* Update python/ray/tune/stopper.py

Co-authored-by: Richard Liaw <rliaw@berkeley.edu>

* Update doc/source/tune/api_docs/stoppers.rst

Co-authored-by: Richard Liaw <rliaw@berkeley.edu>

* Update doc/source/tune/api_docs/stoppers.rst

Co-authored-by: Richard Liaw <rliaw@berkeley.edu>

* Apply suggestions from code review

* Apply suggestions from code review

* Update python/ray/tune/stopper.py

Co-authored-by: Richard Liaw <rliaw@berkeley.edu>
2020-12-12 01:47:19 -08:00
Kai FrickeandGitHub 905652cdd6 [tune] migrate xgboost callback api (#12745)
* Migrate to new-style xgboost callbacks

* Fix flaky progress reporter test

* Fix import error

* Take last value (not first)
2020-12-12 01:42:20 -08:00
42c70be073 [tune] Hyperopt: Directly accept category variables instead of indices (#12715)
* [tune] Hyperopt: Directly accept category variables instead of indices

* Fix interrupt test

* Update python/ray/tune/suggest/hyperopt.py

Co-authored-by: Richard Liaw <rliaw@berkeley.edu>

* Apply suggestions from code review

* Update python/ray/tune/suggest/hyperopt.py

Co-authored-by: Richard Liaw <rliaw@berkeley.edu>

* lint

Co-authored-by: Richard Liaw <rliaw@berkeley.edu>
2020-12-12 01:40:53 -08:00
Kai FrickeandGitHub df10b84113 [Release] release tests yamls for Tune & GPU (#12496) 2020-12-08 10:15:07 -08:00
Kai FrickeandGitHub 1c0d10f67e [tune] Add xgboost_ray integration (#12572) 2020-12-04 13:59:20 -08:00
219c445648 [tune] verbosity refactor second attempt (#12571)
Co-authored-by: Richard Liaw <rliaw@berkeley.edu>
2020-12-04 13:56:26 -08:00
Kai FrickeandGitHub 0a12eba603 Revert "Fix race condition between failure detection and references going out of scope (#12548)" (#12570)
This reverts commit 8801e87a
2020-12-02 10:20:17 -05:00
Kai FrickeandGitHub 63b85df828 [xgb] update docs (#12549) 2020-12-01 23:17:23 -08:00
Kai FrickeandGitHub 1d0ade1b93 Revert "[Core] zero-copy serializer for pytorch (#12344)" (#12469)
This reverts commit 0a505ca8
2020-11-28 10:06:02 -08:00
b94bfdfa99 [tune] use default anonymous metric _metric if at least a mode is set (#12159)
Co-authored-by: Richard Liaw <rliaw@berkeley.edu>
2020-11-23 20:09:33 -08:00
Kai FrickeandGitHub 608d0378c4 [tune] Add test for infinite trials (#12156) 2020-11-21 12:54:01 -08:00
f1ace386db [tune] detect docker and kubernetes syncers (#12108)
Co-authored-by: Richard Liaw <rliaw@berkeley.edu>
2020-11-19 12:17:17 -08:00
Kai FrickeandGitHub 6999075c75 [tune] Add seed parameter to BOHB (#12160) 2020-11-19 10:27:16 -08:00
Kai FrickeandGitHub 2b60c5774b [tune] cache checkpoint serialization (#12064) 2020-11-18 09:03:53 -08:00
Kai FrickeandGitHub 9f5986ee58 [tune] logger migration to ExperimentLogger classes (#11984) 2020-11-16 15:08:37 -08:00
8609e2dd90 [tune] refactor verbosity levels (#11767)
Co-authored-by: Richard Liaw <rliaw@berkeley.edu>
2020-11-16 10:32:53 -08:00
02c02369ca [tune] Fix hpo randint limits (#11946)
Co-authored-by: Sumanth Ratna <sumanthratna@gmail.com>
2020-11-12 08:45:49 -08:00
Kai FrickeandGitHub 287aba6dc3 [tune] schedulers: Add test for context finalization (#11889) 2020-11-09 11:37:05 -08:00
Kai FrickeandGitHub 88be1ea20b [tune] Handle infinite and NaN values (#11835) 2020-11-09 11:18:31 -08:00
Kai FrickeandGitHub 603accf1c2 [tune] logger refactor part 3: Add ExperimentLogger class (#11749) 2020-11-05 08:55:38 -08:00
007634fd1b [tune] logger refactor part 2: Add SyncerCallback (#11748)
Co-authored-by: Richard Liaw <rliaw@berkeley.edu>
2020-11-03 21:04:40 -08:00
Kai FrickeandGitHub f7b19c41e3 [tune] logger refactor part 1: move classes and utilities to own files (#11746)
* [tune] logger refactor part 1: move classes and utilities to own files

* Fix circular dependency

* Remove uneeded pretty print copy

* Apply suggestions from code review
2020-11-03 07:48:09 -08:00
Kai FrickeandGitHub ba63ded311 [tune] better error when metric or mode unset in search algorithms (#11646) 2020-10-28 13:17:59 -07:00
Kai FrickeandGitHub 1a1ff28d18 [tune] allow tune search spaces to be passed to search algorithms (#11503) 2020-10-26 12:33:13 -07:00
Kai FrickeandGitHub 8ee4f7eca3 [tune] fix pbt ptl example (#11573)
* [tune] fix pbt ptl example

* wider smoke test
2020-10-23 12:42:13 -07:00
Kai FrickeandGitHub 6d11fb8bc6 [tune] validate function callable in tune.with_parameters() (#11504) 2020-10-20 16:03:24 -07:00
Kai FrickeandGitHub ed81010b0d [tune] support hierarchical search spaces for hyperopt (#11431)
* support hierarchical search spaces for hyperopt

* Reduce num samples

* Fix prefix
2020-10-19 16:00:08 -07:00
Kai FrickeandGitHub 2f74fe5b71 [tune/docs] Add PTL example to tune docs/examples (#11474) 2020-10-19 14:47:58 -07:00
Kai FrickeandGitHub f890808c14 [docs] Add xgboost_ray to community integrations (#11428) 2020-10-16 10:15:28 -07:00
Kai FrickeandGitHub f7120d2a18 [tune] Make metrics parameter optional in pytorch lightning integration (#11402) 2020-10-14 17:50:34 -07:00
b450cb030a [tune] reuse actors for function API (#11230)
Co-authored-by: Kristian Hartikainen <kristian.hartikainen@gmail.com>
2020-10-08 16:15:02 -07:00
Kai FrickeandGitHub e58613b5e5 [tune/dashboard] Fix Tune dashboard to work with all trainables (#11232) 2020-10-06 14:03:31 -07:00
Kai FrickeandGitHub 681c24754a [tune] Example for using huggingface hyperparamer_search API (#11158) 2020-10-01 16:00:57 -07:00
bdf647c4ec [tune] docker syncer (#11035)
* Add DockerSyncer

* Add docs

* Update python/ray/tune/integration/docker.py

Co-authored-by: Richard Liaw <rliaw@berkeley.edu>

* Updated docs

* fix dir

* Added docker integration test

* added docker integration test to bazel build

* Use sdk.rsync API

Co-authored-by: Richard Liaw <rliaw@berkeley.edu>
2020-10-01 11:59:23 -07:00
Kai FrickeandGitHub de3c606b20 [tune] Use latest Ax release (#11155)
* Use latest Ax release

* specify egg

* Install latest Ax only for py3.7
2020-10-01 08:51:15 -07:00
Kai FrickeandGitHub c77cfaa5ad [tune] use dated experiment dir per default (#11104) 2020-09-30 14:43:59 -07:00
Kai FrickeandGitHub b8f344f695 [tune] add faq entry for reproducing experiments (setting seeds etc) (#11106) 2020-09-29 14:48:39 -07:00
Kai FrickeandGitHub 10015e60fb [tune] fix flaky callback test (#11102) 2020-09-29 09:44:32 -07:00
Kai FrickeandGitHub e7315b0856 [tune] Callbacks for tune runs (#11001) 2020-09-27 16:50:07 -07:00
Kai FrickeandGitHub d9c4dea7cf [tune] strict metric checking (#10972) 2020-09-24 10:00:48 -07:00
Kai FrickeandGitHub 5921e87ecd [tune] Only add new trial when there is no pending trial (#10979) 2020-09-23 11:08:12 -07:00
Kai FrickeandGitHub 2d16ab2e16 [tune] Remove unnecessary wandb group parameter (#10950) 2020-09-22 09:36:51 -07:00
6247740b94 [tune] sort running trials to top in status table (#10926)
Co-authored-by: Richard Liaw <rliaw@berkeley.edu>
2020-09-21 13:39:51 -07:00
50d63b8077 [tune] update pt tutorial docs (#10925)
Co-authored-by: Richard Liaw <rliaw@berkeley.edu>
2020-09-21 13:33:37 -07:00
Kai FrickeandGitHub 508cfa3540 [tune] Support yield and return statements (#10857)
* Support `yield` and `return` statements in Tune trainable functions

* Support anonymous metric with ``tune.report(value)``

* Raise on invalid return/yield value

* Fix end to end reporter test
2020-09-17 20:18:35 -07:00
Kai FrickeandGitHub ee99c919e3 [tune] lazy trials (#10802)
* Lazily fill trial queue

* Update interface

* Update end to end reporter test

* Removed `next_trials()` method

* Lint

* Print total number of samples to be generated in progress reporter. Allow infinite samples.

* Nit check
2020-09-17 08:51:46 -07:00
c9fafe7733 [tune] added type hints (#10806)
Co-authored-by: Richard Liaw <rliaw@berkeley.edu>
2020-09-15 21:03:56 -07:00
Kai FrickeandGitHub 0865d68466 [tune] convert fallback representation to numbers in wandb integration (#10799) 2020-09-15 11:47:11 -07:00
Kai FrickeandGitHub 7eaf063f29 [tune] wrapper function to pass arbitrary objects through the object store to trainables (#10679) 2020-09-10 17:39:44 -07:00
67bf396ae7 [tune] wandb log cleaning to use yaml representer (#10680)
Co-authored-by: Richard Liaw <rliaw@berkeley.edu>
2020-09-10 00:07:16 -07:00
d7c7aba99c [tune] Tune experiment analysis improvements (#10645)
Co-authored-by: Richard Liaw <rliaw@berkeley.edu>
2020-09-08 21:00:52 -07:00
756a9ea641 [tune] add mode/metric parameters to tune.run (#10627)
Co-authored-by: Richard Liaw <rliaw@berkeley.edu>
2020-09-08 17:06:21 -07:00
Kai FrickeandGitHub 87c4f36f02 [tune] Added timeout parameter to tune.run(), (#10642) 2020-09-08 15:38:28 -07:00
69c1a9dd08 [tune] clean up logs before logging to wandb (#10654)
Co-authored-by: Richard Liaw <rliaw@berkeley.edu>
2020-09-08 14:48:18 -07:00
Kai FrickeandGitHub 088f8ebb69 [tune] Add algorithms for search space conversion (#10621) 2020-09-07 13:44:16 -07:00
Kai FrickeandGitHub c5e9bafe15 [tune] Fix flaky test in test_sample (#10602) 2020-09-06 10:29:48 -07:00
2fac66650d [tune] extend search space api docs (#10576)
Co-authored-by: Richard Liaw <rliaw@berkeley.edu>
2020-09-04 18:39:51 -07:00
6b6780a108 [tune] Allow trials to remain paused in BOHB (#10531)
Co-authored-by: Richard Liaw <rliaw@berkeley.edu>
2020-09-04 15:29:33 -07:00
Kai FrickeandGitHub 2e49e22f21 [tune] Add test_sample to bazel BUILD (#10566) 2020-09-04 09:09:02 -07:00
Kai FrickeandGitHub 5c3d4a6670 [tune] added MXNet integration callbacks (#10533) 2020-09-03 18:06:44 -07:00
krfrickeandGitHub 91535e9102 [tune] Refactored Keras integration callbacks (#10509) 2020-09-03 10:16:08 -07:00
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
krfrickeandGitHub 57c4183724 [tune] add xgboost callbacks to integration module (#10502) 2020-09-02 11:16:09 -07:00
krfrickeandGitHub 1dd55f4b07 [tune] remove callbacks from config in wandb logger initialization (#10441) 2020-09-01 14:26:39 -07:00
f3f698816d [tune] Added PyTorch Lightning callbacks to integrations (#10220)
Co-authored-by: Richard Liaw <rliaw@berkeley.edu>
2020-08-31 15:30:48 -07:00
krfrickeandGitHub efa1d51aea [tune] Added external PyTorch tutorial test (#10192) 2020-08-31 15:24:31 -07:00
krfrickeandGitHub 53ab228b75 [tune] Fix log to file on actor reuse (#10363) 2020-08-27 08:22:19 -07:00
5a787a8253 [tune] added FAQ to docs (#10222)
Co-authored-by: Richard Liaw <rliaw@berkeley.edu>
2020-08-24 21:51:02 -07:00
krfrickeandGitHub c31876002d [tune/rllib] made wandb compatible with rllib trainables (#10252) 2020-08-21 17:25:52 -07:00
8f0f7371a0 [tune] Added Kubernetes syncer and sync client (#10097)
Co-authored-by: Edward Oakes <ed.nmi.oakes@gmail.com>
Co-authored-by: Richard Liaw <rliaw@berkeley.edu>
Co-authored-by: Kai Fricke <kai@anyscale.com>
2020-08-16 14:09:28 -07:00
16486a8df3 [tune] Add OptunaSearcher wrapper around Optuna samplers (#10044)
Co-authored-by: Richard Liaw <rliaw@berkeley.edu>
Co-authored-by: Kai Fricke <kai@anyscale.com>
2020-08-12 16:13:22 -07:00
221fdc0774 [tune] fix flaky PBT replay test (#10047)
Co-authored-by: Kai Fricke <kai@anyscale.com>
2020-08-11 14:17:31 -07:00
7301733a1f [tune] Close logfile contexts (#10026)
Co-authored-by: Kai Fricke <kai@anyscale.com>
2020-08-10 12:40:40 -07:00
dee3322ab0 [tune] Ray Tune + Serve end-to-end integration example (#9908)
Co-authored-by: Kai Fricke <kai@anyscale.com>
2020-08-07 15:32:49 -07:00
0ef8224446 [tune] PBT replay utility scheduler (#9953)
Co-authored-by: Kai Fricke <kai@anyscale.com>
2020-08-07 12:41:49 -07:00
ef717ecda6 [tune] Prevent leak of magic keys in trial config (#9903)
Co-authored-by: Kai Fricke <kai@anyscale.com>
Co-authored-by: Richard Liaw <rliaw@berkeley.edu>
2020-08-04 11:24:01 -07:00
c741d1cf9c [tune] stdout/stderr logging redirection (#9817)
* Add `log_to_file` parameter, pass to Trainable config, redirect stdout/stderr.

* Add logging handler to root ray logger

* Added test for `log_to_file` parameter

* Added logs, reuse test

* Revert debug change

* Update logdir on reset, flush streams after each train() step

* Remove magic keys from visible config

Co-authored-by: Kai Fricke <kai@anyscale.com>
2020-08-03 11:18:34 -07:00
619e44e54a [tune] Added WandbLogger (#9725)
Co-authored-by: Richard Liaw <rliaw@berkeley.edu>
Co-authored-by: Kai Fricke <kai@anyscale.com>
2020-07-30 13:09:03 -07:00
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
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
ad0219b80d [tune] fix pbt checkpoint_freq (#9517)
* Only delete old checkpoint if it is not the same as the new one

* Return early if old checkpoint value coincides with new checkpoint value

Co-authored-by: Kai Fricke <kai@anyscale.com>
2020-07-18 00:58:16 -07:00
87630cf024 [tune] Unflattened lookup for ProgressReporter (#9525)
Co-authored-by: Kai Fricke <kai@anyscale.com>
2020-07-17 13:52:54 -07:00
5a40299d42 [tune] extend PTL template (GPU, typing fixes, tensorboard) (#9451)
Co-authored-by: Kai Fricke <kai@anyscale.com>
2020-07-15 10:30:20 -07:00
deba082cb4 [tune] PyTorch CIFAR10 example (#9338)
Co-authored-by: Richard Liaw <rliaw@berkeley.edu>
Co-authored-by: Kai Fricke <kai@anyscale.com>
2020-07-13 23:16:05 -07:00
e0b6984dce [tune] pytorch lightning template and walkthrough (#9151)
Co-authored-by: Kai Fricke <kai@anyscale.com>
2020-06-29 16:52:07 -07:00
22ea8dde84 [Tune] Added XGBoost tutorial and template (#9060)
* Added XGBoost tutorial and template

* XGBoost tutorial: Cut some clutter

* Apply suggestions from code review

Co-authored-by: Richard Liaw <rliaw@berkeley.edu>

* Added XGboost logo

* Fixed further references

Co-authored-by: Kai Fricke <kai@anyscale.com>
Co-authored-by: Richard Liaw <rliaw@berkeley.edu>
2020-06-25 15:59:54 -07:00
060e524c92 [tune] Parameter columns can now be specified in tune reporters (#8802)
Co-authored-by: Richard Liaw <rliaw@berkeley.edu>
Co-authored-by: Kai Fricke <kai@anyscale.com>
2020-06-11 11:30:25 -07:00
e62c1d2051 [tune] Use scientific notation in tune dashboard (#8782)
Co-authored-by: Kai Fricke <kai@anyscale.com>
2020-06-05 10:41:07 -07:00
f4ee3e76d8 [tune] last-n-avg
Co-authored-by: Kai Fricke <kai@anyscale.com>
2020-06-02 20:06:04 -07:00
4d0e9f3c71 [Dashboard] Sort IDLE workers to bottom in dashboard (#8708)
* Sort IDLE workers to bottom in dashboard

* Fixed linting error

Co-authored-by: Kai Fricke <kai@anyscale.com>
2020-06-02 14:00:59 -07:00
krfrickeandGitHub e5b6566d28 Remove blocking flag from serve.init() (#8654) 2020-05-29 13:25:35 -07:00
krfrickeandGitHub 4633d81c39 [tune] added average scope to experiment analysis (#8445) 2020-05-14 15:20:43 -07:00