diff --git a/.gitignore b/.gitignore index 9268fa948c4e..56b81a8f39c8 100644 --- a/.gitignore +++ b/.gitignore @@ -119,6 +119,7 @@ scripts/nodes.txt /doc/_build /doc/source/_static/thumbs /doc/source/tune/generated_guides/ +/doc/source/**/doc/ # User-specific stuff: .idea/**/workspace.xml diff --git a/doc/source/_toc.yml b/doc/source/_toc.yml index 2e82bb4cab8a..a5b63efb6462 100644 --- a/doc/source/_toc.yml +++ b/doc/source/_toc.yml @@ -282,7 +282,7 @@ parts: - file: tune/examples/exercises title: "Exercises" - file: tune/faq - - file: tune/api_docs/overview.rst + - file: tune/api/api.rst - file: serve/index title: Ray Serve diff --git a/doc/source/ray-air/package-ref.rst b/doc/source/ray-air/package-ref.rst index a1e90d55b2cc..2fa48650e45a 100644 --- a/doc/source/ray-air/package-ref.rst +++ b/doc/source/ray-air/package-ref.rst @@ -159,6 +159,7 @@ Training Session .. automodule:: ray.air.session :members: + :noindex: Trainer Configs ############### @@ -364,26 +365,8 @@ Reinforcement Learning (RLlib) .. _air-builtin-callbacks: -Monitoring Integrations +Integrations ~~~~~~~~~~~~~~~~~~~~~~~ -Comet -##### - -.. autoclass:: ray.air.integrations.comet.CometLoggerCallback - -Keras -##### - -.. autoclass:: ray.air.integrations.keras.Callback - :members: - -MLflow -###### - -.. autoclass:: ray.air.integrations.mlflow.MLflowLoggerCallback - -Weights and Biases -################## - -.. autoclass:: ray.air.integrations.wandb.WandbLoggerCallback +See :doc:`this API reference ` for AIR integrations with other libraries +such as Weights and Biases, MLFlow, Keras, and more. diff --git a/doc/source/ray-references/api.rst b/doc/source/ray-references/api.rst index fb55ac2015c2..55cf2e3dd29e 100644 --- a/doc/source/ray-references/api.rst +++ b/doc/source/ray-references/api.rst @@ -8,7 +8,7 @@ API References ../ray-air/package-ref.rst ../data/api/api.rst ../train/api.rst - ../tune/api_docs/overview.rst + ../tune/api/api.rst ../serve/api/index.rst ../rllib/package_ref/index.rst ../workflows/api/api.rst diff --git a/doc/source/tune/api_docs/overview.rst b/doc/source/tune/api/api.rst similarity index 97% rename from doc/source/tune/api_docs/overview.rst rename to doc/source/tune/api/api.rst index a2581cf88495..6d2e3d36c4b7 100644 --- a/doc/source/tune/api_docs/overview.rst +++ b/doc/source/tune/api/api.rst @@ -14,15 +14,16 @@ on `Github`_. :maxdepth: 2 execution.rst + result_grid.rst trainable.rst search_space.rst suggestion.rst schedulers.rst stoppers.rst - result_grid.rst reporters.rst syncing.rst logging.rst + callbacks.rst env.rst sklearn.rst integration.rst diff --git a/doc/source/tune/api/callbacks.rst b/doc/source/tune/api/callbacks.rst new file mode 100644 index 000000000000..5cee421c13ba --- /dev/null +++ b/doc/source/tune/api/callbacks.rst @@ -0,0 +1,55 @@ +.. _tune-callbacks-docs: + +Tune Callbacks (tune.Callback) +============================== + +See :doc:`this user guide ` for more details. + +.. seealso:: + + :doc:`Tune's built-in loggers ` use the ``Callback`` interface. + + +Callback Interface +------------------ + +Callback Initialization and Setup +~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ + +.. currentmodule:: ray.tune +.. autosummary:: + :toctree: doc/ + + Callback + Callback.setup + + +Callback Hooks +~~~~~~~~~~~~~~ + +.. autosummary:: + :toctree: doc/ + + Callback.on_checkpoint + Callback.on_experiment_end + Callback.on_step_begin + Callback.on_step_end + Callback.on_trial_complete + Callback.on_trial_error + Callback.on_trial_restore + Callback.on_trial_result + Callback.on_trial_save + Callback.on_trial_start + + +Stateful Callbacks +~~~~~~~~~~~~~~~~~~ + +The following methods must be overriden for stateful callbacks to be saved/restored +properly by Tune. + +.. autosummary:: + :toctree: doc/ + + Callback.get_state + Callback.set_state diff --git a/doc/source/tune/api_docs/cli.rst b/doc/source/tune/api/cli.rst similarity index 100% rename from doc/source/tune/api_docs/cli.rst rename to doc/source/tune/api/cli.rst diff --git a/doc/source/tune/api_docs/client.rst b/doc/source/tune/api/client.rst similarity index 100% rename from doc/source/tune/api_docs/client.rst rename to doc/source/tune/api/client.rst diff --git a/doc/source/tune/api_docs/env.rst b/doc/source/tune/api/env.rst similarity index 100% rename from doc/source/tune/api_docs/env.rst rename to doc/source/tune/api/env.rst diff --git a/doc/source/tune/api/execution.rst b/doc/source/tune/api/execution.rst new file mode 100644 index 000000000000..50154462a703 --- /dev/null +++ b/doc/source/tune/api/execution.rst @@ -0,0 +1,43 @@ +Tune Execution (tune.Tuner) +=========================== + +.. _tune-run-ref: + +Tuner +----- + +.. currentmodule:: ray.tune + +.. autosummary:: + :toctree: doc/ + + Tuner + Tuner.fit + Tuner.get_results + +Tuner Configuration +~~~~~~~~~~~~~~~~~~~ + +.. autosummary:: + :toctree: doc/ + + TuneConfig + +Restoring a Tuner +~~~~~~~~~~~~~~~~~ + +.. autosummary:: + :toctree: doc/ + + Tuner.restore + Tuner.can_restore + + +tune.run_experiments +-------------------- + +.. autosummary:: + :toctree: doc/ + + run_experiments + Experiment diff --git a/doc/source/tune/api/integration.rst b/doc/source/tune/api/integration.rst new file mode 100644 index 000000000000..0ad8ff86d2a6 --- /dev/null +++ b/doc/source/tune/api/integration.rst @@ -0,0 +1,115 @@ +.. _tune-integration: + +External library integrations for Ray Tune +=========================================== + +.. TODO: Clean this up. Both tune.integration and air.integrations are +.. captured here. Most of the `tune.integration` can be deprecated soon. +.. XGBoost/LightGBM callbacks are no longer recommended - use their trainers instead +.. which will automatically report+checkpoint. +.. After PTL trainer is introduced, we can also deprecate that callback. + +.. currentmodule:: ray + +.. _tune-monitoring-integrations: + +Tune Experiment Monitoring Integrations +---------------------------------------- + +Comet (air.integrations.comet) +~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ + +:ref:`See here for an example. ` + +.. autosummary:: + :toctree: doc/ + + ~air.integrations.comet.CometLoggerCallback + + +.. _tune-integration-mlflow: + +MLflow (air.integrations.mlflow) +~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ + +:ref:`See here for an example. ` + +.. autosummary:: + :toctree: doc/ + + ~air.integrations.mlflow.MLflowLoggerCallback + ~air.integrations.mlflow.setup_mlflow + +.. _tune-integration-wandb: + +Weights and Biases (air.integrations.wandb) +~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ + +:ref:`See here for an example. ` + +.. autosummary:: + :toctree: doc/ + + ~air.integrations.wandb.WandbLoggerCallback + ~air.integrations.wandb.setup_wandb + + +Integrations with ML Libraries +-------------------------------- + +.. _tune-integration-keras: + +Keras (air.integrations.keras) +~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ + +.. autosummary:: + :toctree: doc/ + + ~air.integrations.keras.ReportCheckpointCallback + + +.. _tune-integration-mxnet: + +MXNet (tune.integration.mxnet) +~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ + +.. autosummary:: + :toctree: doc/ + + ~tune.integration.mxnet.TuneReportCallback + ~tune.integration.mxnet.TuneCheckpointCallback + + +.. _tune-integration-pytorch-lightning: + +PyTorch Lightning (tune.integration.pytorch_lightning) +~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ + +.. autosummary:: + :toctree: doc/ + + ~tune.integration.pytorch_lightning.TuneReportCallback + ~tune.integration.pytorch_lightning.TuneReportCheckpointCallback + +.. _tune-integration-xgboost: + +XGBoost (tune.integration.xgboost) +~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ + +.. autosummary:: + :toctree: doc/ + + ~tune.integration.xgboost.TuneReportCallback + ~tune.integration.xgboost.TuneReportCheckpointCallback + + +.. _tune-integration-lightgbm: + +LightGBM (tune.integration.lightgbm) +~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ + +.. autosummary:: + :toctree: doc/ + + ~tune.integration.lightgbm.TuneReportCallback + ~tune.integration.lightgbm.TuneReportCheckpointCallback diff --git a/doc/source/tune/api_docs/internals.rst b/doc/source/tune/api/internals.rst similarity index 61% rename from doc/source/tune/api_docs/internals.rst rename to doc/source/tune/api/internals.rst index 3b90dd7ed1d9..d33717ec9ef0 100644 --- a/doc/source/tune/api_docs/internals.rst +++ b/doc/source/tune/api/internals.rst @@ -3,6 +3,13 @@ Tune Internals .. _raytrialexecutor-docstring: +TunerInternal +--------------- + +.. autoclass:: ray.tune.impl.tuner_internal.TunerInternal + :members: + + RayTrialExecutor ---------------- @@ -15,6 +22,7 @@ TrialRunner ----------- .. autoclass:: ray.tune.execution.trial_runner.TrialRunner + :members: .. _trial-docstring: @@ -22,23 +30,14 @@ Trial ----- .. autoclass:: ray.tune.experiment.trial.Trial + :members: -.. _tune-callbacks-docs: - -Callbacks ---------- - -.. autoclass:: ray.tune.callback.Callback - :members: - - -.. _resources-docstring: - -PlacementGroupFactory ---------------------- +FunctionTrainable +----------------- -.. autoclass:: ray.tune.execution.placement_groups.PlacementGroupFactory +.. autoclass:: ray.tune.trainable.function_trainable.FunctionTrainable +.. autofunction:: ray.tune.trainable.function_trainable.wrap_function Registry diff --git a/doc/source/tune/api_docs/logging.rst b/doc/source/tune/api/logging.rst similarity index 62% rename from doc/source/tune/api_docs/logging.rst rename to doc/source/tune/api/logging.rst index ec9a3f9e0f90..e5eb8732b626 100644 --- a/doc/source/tune/api_docs/logging.rst +++ b/doc/source/tune/api/logging.rst @@ -31,40 +31,54 @@ relevant ones (like accuracy, loss, etc.). .. image:: ../images/ray-tune-viskit.png -TBXLogger ---------- +.. currentmodule:: ray -.. autoclass:: ray.tune.logger.TBXLoggerCallback +Tune Built-in Loggers +--------------------- -JsonLogger ----------- +.. autosummary:: + :toctree: doc/ -.. autoclass:: ray.tune.logger.JsonLoggerCallback + tune.logger.JsonLoggerCallback + tune.logger.CSVLoggerCallback + tune.logger.TBXLoggerCallback -CSVLogger ---------- -.. autoclass:: ray.tune.logger.CSVLoggerCallback - -MLFlowLogger ------------- +MLFlow Integration: MLFlowLoggerCallback +---------------------------------------- Tune also provides a logger for `MLflow `_. You can install MLflow via ``pip install mlflow``. You can see the :doc:`tutorial here `. -WandbLogger ------------ +.. autosummary:: + :toctree: doc/ + + air.integrations.mlflow.MLflowLoggerCallback + +Wandb Integration: WandbLoggerCallback +-------------------------------------- Tune also provides a logger for `Weights & Biases `_. You can install Wandb via ``pip install wandb``. -You can see the :doc:`tutorial here ` +You can see the :doc:`tutorial here `. +.. autosummary:: + :toctree: doc/ + + air.integrations.wandb.WandbLoggerCallback .. _logger-interface: -LoggerCallback --------------- +LoggerCallback Interface +------------------------ + +.. autosummary:: + :toctree: doc/ -.. autoclass:: ray.tune.logger.LoggerCallback - :members: log_trial_start, log_trial_restore, log_trial_save, log_trial_result, log_trial_end + tune.logger.LoggerCallback + tune.logger.LoggerCallback.log_trial_start + tune.logger.LoggerCallback.log_trial_restore + tune.logger.LoggerCallback.log_trial_save + tune.logger.LoggerCallback.log_trial_result + tune.logger.LoggerCallback.log_trial_end diff --git a/doc/source/tune/api_docs/reporters.rst b/doc/source/tune/api/reporters.rst similarity index 90% rename from doc/source/tune/api_docs/reporters.rst rename to doc/source/tune/api/reporters.rst index ade3263deb1c..b163d5695099 100644 --- a/doc/source/tune/api_docs/reporters.rst +++ b/doc/source/tune/api/reporters.rst @@ -88,21 +88,26 @@ The default reporting style can also be overridden more broadly by extending the results = tuner.fit() -CLIReporter ------------ +.. currentmodule:: ray.tune -.. autoclass:: ray.tune.CLIReporter - :members: add_metric_column +Reporter Interface (tune.ProgressReporter) +------------------------------------------ -JupyterNotebookReporter ------------------------ +.. autosummary:: + :toctree: doc/ + + ProgressReporter + ProgressReporter.report + ProgressReporter.should_report -.. autoclass:: ray.tune.JupyterNotebookReporter - :members: add_metric_column +Tune Built-in Reporters +----------------------- -ProgressReporter ----------------- +.. autosummary:: + :toctree: doc/ -.. autoclass:: ray.tune.ProgressReporter - :members: + CLIReporter + CLIReporter.add_metric_column + JupyterNotebookReporter + JupyterNotebookReporter.add_metric_column diff --git a/doc/source/tune/api/result_grid.rst b/doc/source/tune/api/result_grid.rst new file mode 100644 index 000000000000..abda7956af8f --- /dev/null +++ b/doc/source/tune/api/result_grid.rst @@ -0,0 +1,44 @@ +.. _tune-analysis-docs: + +.. _result-grid-docstring: + +Tune Experiment Results (tune.ResultGrid) +========================================= + +ResultGrid (tune.ResultGrid) +---------------------------- + +.. currentmodule:: ray + +.. autosummary:: + :toctree: doc/ + + tune.ResultGrid + tune.ResultGrid.get_best_result + tune.ResultGrid.get_dataframe + +.. _result-docstring: + +Result (air.Result) +------------------- + +.. autosummary:: + :toctree: doc/ + + air.Result + +.. _exp-analysis-docstring: + + +ExperimentAnalysis (tune.ExperimentAnalysis) +-------------------------------------------- + +.. note:: + + An experiment analysis is the output of the ``tune.run`` API. + It's now recommended to use ``Tuner.fit``, which outputs a ``ResultGrid`` object. + +.. autosummary:: + :toctree: doc/ + + tune.ExperimentAnalysis diff --git a/doc/source/tune/api_docs/schedulers.rst b/doc/source/tune/api/schedulers.rst similarity index 81% rename from doc/source/tune/api_docs/schedulers.rst rename to doc/source/tune/api/schedulers.rst index e53353e42399..a7ee7772092a 100644 --- a/doc/source/tune/api_docs/schedulers.rst +++ b/doc/source/tune/api/schedulers.rst @@ -13,9 +13,26 @@ Trainable and is maximized or minimized according to ``mode``. .. code-block:: python from ray import tune - tuner = tune.Tuner( ... , tune_config=tune.TuneConfig(scheduler=Scheduler(metric="accuracy", mode="max"))) + from ray.air import session + from tune.schedulers import ASHAScheduler + + def train_fn(config): + # This objective function is just for demonstration purposes + session.report({"loss": config["param"]}) + + tuner = tune.Tuner( + train_fn, + tune_config=tune.TuneConfig( + scheduler=ASHAScheduler(), + metric="loss", + mode="min", + num_samples=10, + ), + param_space={"param": tune.uniform(0, 1)}, + ) results = tuner.fit() +.. currentmodule:: ray.tune.schedulers .. _tune-scheduler-hyperband: @@ -28,15 +45,21 @@ setting the ``scheduler`` parameter of ``tune.TuneConfig``, which is taken in by .. code-block:: python from ray import tune + from tune.schedulers import ASHAScheduler + asha_scheduler = ASHAScheduler( time_attr='training_iteration', - metric='episode_reward_mean', - mode='max', + metric='loss', + mode='min', max_t=100, grace_period=10, reduction_factor=3, - brackets=1) - tuner = tune.Tuner( ... , tune_config=tune.TuneConfig(scheduler=asha_scheduler)) + brackets=1, + ) + tuner = tune.Tuner( + train_fn, + tune_config=tune.TuneConfig(scheduler=asha_scheduler), + ) results = tuner.fit() Compared to the original version of HyperBand, this implementation provides better @@ -48,9 +71,11 @@ Even though the original paper mentions a bracket count of 3, discussions with t that the value should be left to 1 bracket. This is the default used if no value is provided for the ``brackets`` argument. -.. autoclass:: ray.tune.schedulers.AsyncHyperBandScheduler +.. autosummary:: + :toctree: doc/ -.. autoclass:: ray.tune.schedulers.ASHAScheduler + AsyncHyperBandScheduler + ASHAScheduler .. _tune-original-hyperband: @@ -60,7 +85,10 @@ HyperBand (tune.schedulers.HyperBandScheduler) Tune implements the `standard version of HyperBand `__. **We recommend using the ASHA Scheduler over the standard HyperBand scheduler.** -.. autoclass:: ray.tune.schedulers.HyperBandScheduler +.. autosummary:: + :toctree: doc/ + + HyperBandScheduler HyperBand Implementation Details @@ -105,7 +133,10 @@ Median Stopping Rule (tune.schedulers.MedianStoppingRule) The Median Stopping Rule implements the simple strategy of stopping a trial if its performance falls below the median of other trials at similar points in time. -.. autoclass:: ray.tune.schedulers.MedianStoppingRule +.. autosummary:: + :toctree: doc/ + + MedianStoppingRule .. _tune-scheduler-pbt: @@ -117,23 +148,25 @@ This can be enabled by setting the ``scheduler`` parameter of ``tune.TuneConfig` .. code-block:: python + from ray import tune + from ray.tune.schedulers import PopulationBasedTraining + pbt_scheduler = PopulationBasedTraining( time_attr='training_iteration', - metric='mean_accuracy', - mode='max', - perturbation_interval=600.0, + metric='loss', + mode='min', + perturbation_interval=1, hyperparam_mutations={ "lr": [1e-3, 5e-4, 1e-4, 5e-5, 1e-5], - "alpha": lambda: random.uniform(0.0, 1.0), - ... + "alpha": tune.uniform(0.0, 1.0), } ) tuner = tune.Tuner( - ..., + train_fn, tune_config=tune.TuneConfig( num_samples=4, - scheduler=pbt_scheduler - ) + scheduler=pbt_scheduler, + ), ) tuner.fit() @@ -150,7 +183,10 @@ Take a look at :doc:`/tune/examples/pbt_visualization/pbt_visualization` to get of how PBT operates. :doc:`/tune/examples/pbt_guide` gives more examples of PBT usage. -.. autoclass:: ray.tune.schedulers.PopulationBasedTraining +.. autosummary:: + :toctree: doc/ + + PopulationBasedTraining .. _tune-scheduler-pbt-replay: @@ -165,20 +201,26 @@ config according to the obtained schedule. .. code-block:: python + from ray import tune + from ray.tune.schedulers import PopulationBasedTrainingReplay + replay = PopulationBasedTrainingReplay( experiment_dir="~/ray_results/pbt_experiment/", - trial_id="XXXXX_00001") + trial_id="XXXXX_00001" + ) tuner = tune.Tuner( - ..., + train_fn, tune_config=tune.TuneConfig(scheduler=replay) - ) + ) results = tuner.fit() See :ref:`here for an example ` on how to use the replay utility in practice. -.. autoclass:: ray.tune.schedulers.PopulationBasedTrainingReplay +.. autosummary:: + :toctree: doc/ + PopulationBasedTrainingReplay .. _tune-scheduler-pb2: @@ -203,15 +245,16 @@ PB2 can be enabled by setting the ``scheduler`` parameter of ``tune.TuneConfig`` from ray.tune.schedulers.pb2 import PB2 pb2_scheduler = PB2( - time_attr='time_total_s', - metric='mean_accuracy', - mode='max', - perturbation_interval=600.0, - hyperparam_bounds={ - "lr": [1e-3, 1e-5], - "alpha": [0.0, 1.0], - ... - }) + time_attr='time_total_s', + metric='mean_accuracy', + mode='max', + perturbation_interval=600.0, + hyperparam_bounds={ + "lr": [1e-3, 1e-5], + "alpha": [0.0, 1.0], + ... + } + ) tuner = tune.Tuner( ... , tune_config=tune.TuneConfig(scheduler=pb2_scheduler)) results = tuner.fit() @@ -227,7 +270,10 @@ With that in mind, you can run this :doc:`PB2 PPO example ` for package requirements, examples, and d An example of this in use can be found here: :doc:`/tune/examples/includes/bohb_example`. -.. autoclass:: ray.tune.schedulers.HyperBandForBOHB + +.. autosummary:: + :toctree: doc/ + + HyperBandForBOHB .. _tune-resource-changing-scheduler: @@ -265,28 +315,32 @@ It wraps around another scheduler and uses its decisions. An example of this in use can be found here: :doc:`/tune/examples/includes/xgboost_dynamic_resources_example`. -.. autoclass:: ray.tune.schedulers.ResourceChangingScheduler +.. autosummary:: + :toctree: doc/ -DistributeResources -~~~~~~~~~~~~~~~~~~~ + ResourceChangingScheduler + resource_changing_scheduler.DistributeResources + resource_changing_scheduler.DistributeResourcesToTopJob -.. autoclass:: ray.tune.schedulers.resource_changing_scheduler.DistributeResources +FIFOScheduler (Default Scheduler) +--------------------------------- -DistributeResourcesToTopJob -~~~~~~~~~~~~~~~~~~~~~~~~~~~ +.. autosummary:: + :toctree: doc/ -.. autoclass:: ray.tune.schedulers.resource_changing_scheduler.DistributeResourcesToTopJob + FIFOScheduler -FIFOScheduler -------------- +TrialScheduler Interface +------------------------ -.. autoclass:: ray.tune.schedulers.FIFOScheduler +.. autosummary:: + :toctree: doc/ -TrialScheduler --------------- + TrialScheduler + TrialScheduler.choose_trial_to_run + TrialScheduler.on_trial_result + TrialScheduler.on_trial_complete -.. autoclass:: ray.tune.schedulers.TrialScheduler - :members: Shim Instantiation (tune.create_scheduler) ------------------------------------------ @@ -295,4 +349,7 @@ There is also a shim function that constructs the scheduler based on the provide This can be useful if the scheduler you want to use changes often (e.g., specifying the scheduler via a CLI option or config file). -.. automethod:: ray.tune.create_scheduler +.. autosummary:: + :toctree: doc/ + + create_scheduler diff --git a/doc/source/tune/api_docs/search_space.rst b/doc/source/tune/api/search_space.rst similarity index 76% rename from doc/source/tune/api_docs/search_space.rst rename to doc/source/tune/api/search_space.rst index 0ab08b9edb16..21c2e7520f18 100644 --- a/doc/source/tune/api_docs/search_space.rst +++ b/doc/source/tune/api/search_space.rst @@ -3,11 +3,6 @@ Tune Search Space API ===================== -.. _tune-sample-docs: - -Random Distributions API ------------------------- - This section covers the functions you can use to define your search spaces. .. caution:: @@ -19,7 +14,7 @@ This section covers the functions you can use to define your search spaces. .. tip:: Avoid passing large objects as values in the search space, as that will incur a performance overhead. - Use :ref:`tune-with-parameters` to pass large objects in or load them inside your trainable + Use :func:`tune.with_parameters ` to pass large objects in or load them inside your trainable from disk (making sure that all nodes have access to the files) or cloud storage. See :ref:`tune-bottlenecks` for more information. @@ -83,70 +78,35 @@ For a high-level overview, see this example: "grid": tune.grid_search([32, 64, 128]) } -tune.uniform -~~~~~~~~~~~~ - -.. autofunction:: ray.tune.uniform - -tune.quniform -~~~~~~~~~~~~~ - -.. autofunction:: ray.tune.quniform - -tune.loguniform -~~~~~~~~~~~~~~~ - -.. autofunction:: ray.tune.loguniform - -tune.qloguniform -~~~~~~~~~~~~~~~~ - -.. autofunction:: ray.tune.qloguniform - -tune.randn -~~~~~~~~~~ +.. currentmodule:: ray -.. autofunction:: ray.tune.randn - -tune.qrandn -~~~~~~~~~~~ - -.. autofunction:: ray.tune.qrandn - -tune.randint -~~~~~~~~~~~~ - -.. autofunction:: ray.tune.randint - -tune.qrandint -~~~~~~~~~~~~~ - -.. autofunction:: ray.tune.qrandint - -tune.lograndint -~~~~~~~~~~~~~~~ - -.. autofunction:: ray.tune.lograndint - -tune.qlograndint -~~~~~~~~~~~~~~~~ - -.. autofunction:: ray.tune.qlograndint +Random Distributions API +------------------------ -tune.choice -~~~~~~~~~~~ +.. autosummary:: + :toctree: doc/ -.. autofunction:: ray.tune.choice + tune.uniform + tune.quniform + tune.loguniform + tune.qloguniform + tune.randn + tune.qrandn + tune.randint + tune.qrandint + tune.lograndint + tune.qlograndint + tune.choice -tune.sample_from -~~~~~~~~~~~~~~~~ -.. autofunction:: ray.tune.sample_from +Grid Search and Custom Function APIs +------------------------------------ -Grid Search API ---------------- +.. autosummary:: + :toctree: doc/ -.. autofunction:: ray.tune.grid_search + tune.grid_search + tune.sample_from References ---------- diff --git a/doc/source/tune/api_docs/sklearn.rst b/doc/source/tune/api/sklearn.rst similarity index 71% rename from doc/source/tune/api_docs/sklearn.rst rename to doc/source/tune/api/sklearn.rst index 2214303f77c4..3bb733dc8187 100644 --- a/doc/source/tune/api_docs/sklearn.rst +++ b/doc/source/tune/api/sklearn.rst @@ -8,7 +8,9 @@ Tune Scikit-Learn API (tune.sklearn) TuneGridSearchCV ---------------- -.. autoclass:: ray.tune.sklearn.TuneGridSearchCV +.. currentmodule:: ray.tune.sklearn + +.. autoclass:: TuneGridSearchCV :members: .. _tunesearchcv-docs: @@ -16,5 +18,5 @@ TuneGridSearchCV TuneSearchCV ------------ -.. autoclass:: ray.tune.sklearn.TuneSearchCV +.. autoclass:: TuneSearchCV :members: diff --git a/doc/source/tune/api/stoppers.rst b/doc/source/tune/api/stoppers.rst new file mode 100644 index 000000000000..28dfa38d8044 --- /dev/null +++ b/doc/source/tune/api/stoppers.rst @@ -0,0 +1,41 @@ +.. _tune-stoppers: + +Tune Stopping Mechanisms (tune.stopper) +======================================= + +In addition to Trial Schedulers like :ref:`ASHA `, where a number of +trials are stopped if they perform subpar, Ray Tune also supports custom stopping mechanisms to stop trials early. They can also stop the entire experiment after a condition is met. +For instance, stopping mechanisms can specify to stop trials when they reached a plateau and the metric +doesn't change anymore. + +Ray Tune comes with several stopping mechanisms out of the box. For custom stopping behavior, you can +inherit from the :class:`Stopper ` class. + +Other stopping behaviors are described :ref:`in the user guide `. + + +.. _tune-stop-ref: + +Stopper Interface (tune.Stopper) +-------------------------------- + +.. currentmodule:: ray.tune.stopper + +.. autosummary:: + :toctree: doc/ + + Stopper + Stopper.__call__ + Stopper.stop_all + +Tune Built-in Stoppers +---------------------- + +.. autosummary:: + :toctree: doc/ + + MaximumIterationStopper + ExperimentPlateauStopper + TrialPlateauStopper + TimeoutStopper + CombinedStopper diff --git a/doc/source/tune/api_docs/suggestion.rst b/doc/source/tune/api/suggestion.rst similarity index 74% rename from doc/source/tune/api_docs/suggestion.rst rename to doc/source/tune/api/suggestion.rst index 63d949106c01..bce657c9ecc5 100644 --- a/doc/source/tune/api_docs/suggestion.rst +++ b/doc/source/tune/api/suggestion.rst @@ -12,66 +12,87 @@ You can utilize these search algorithms as follows: .. code-block:: python from ray import tune - from ray.tune.search.hyperopt import HyperOptSearch - tuner = tune.Tuner(my_function, tune_config=tune.TuneConfig(search_alg=HyperOptSearch(...))) + from ray.air import session + from ray.tune.search.optuna import OptunaSearch + + def train_fn(config): + # This objective function is just for demonstration purposes + session.report({"loss": config["param"]}) + + tuner = tune.Tuner( + train_fn, + tune_config=tune.TuneConfig( + search_alg=OptunaSearch(), + num_samples=100, + metric="loss", + mode="min", + ), + param_space={"param": tune.uniform(0, 1)}, + ) results = tuner.fit() -Saving and Restoring Tune Runs ------------------------------- +Saving and Restoring Tune Search Algorithms +------------------------------------------- .. TODO: what to do about this section? It doesn't really belong here and is not worth its own guide. .. TODO: at least check that this pseudo-code runs. Certain search algorithms have ``save/restore`` implemented, -allowing reuse of learnings across multiple tuning runs. +allowing reuse of searchers that are fitted on the results of multiple tuning runs. .. code-block:: python search_alg = HyperOptSearch() tuner_1 = tune.Tuner( - trainable, - tune_config=tune.TuneConfig(search_alg=search_alg)) + train_fn, + tune_config=tune.TuneConfig(search_alg=search_alg) + ) results_1 = tuner_1.fit() search_alg.save("./my-checkpoint.pkl") - # Restore the saved state onto another search algorithm + # Restore the saved state onto another search algorithm, + # in a new tuning script search_alg2 = HyperOptSearch() search_alg2.restore("./my-checkpoint.pkl") tuner_2 = tune.Tuner( - trainable, - tune_config=tune.TuneConfig(search_alg=search_alg2)) + train_fn, + tune_config=tune.TuneConfig(search_alg=search_alg2) + ) results_2 = tuner_2.fit() -Tune automatically saves its state inside the current experiment folder ("Result Dir") during tuning. +Tune automatically saves searcher state inside the current experiment folder during tuning. +See ``Result logdir: ...`` in the output logs for this location. Note that if you have two Tune runs with the same experiment folder, the previous state checkpoint will be overwritten. You can avoid this by making sure ``air.RunConfig(name=...)`` is set to a unique -identifier. +identifier: .. code-block:: python search_alg = HyperOptSearch() tuner_1 = tune.Tuner( - cost, + train_fn, tune_config=tune.TuneConfig( num_samples=5, - search_alg=search_alg), + search_alg=search_alg, + ), run_config=air.RunConfig( - verbose=0, name="my-experiment-1", - local_dir="~/my_results" - )) + local_dir="~/my_results", + ) + ) results = tuner_1.fit() search_alg2 = HyperOptSearch() search_alg2.restore_from_dir( - os.path.join("~/my_results", "my-experiment-1")) + os.path.join("~/my_results", "my-experiment-1") + ) .. _tune-basicvariant: @@ -86,25 +107,32 @@ The :class:`BasicVariantGenerator `. -.. autoclass:: ray.tune.search.basic_variant.BasicVariantGenerator +.. currentmodule:: ray.tune.search + +.. autosummary:: + :toctree: doc/ + + basic_variant.BasicVariantGenerator .. _tune-ax: Ax (tune.search.ax.AxSearch) ---------------------------- -.. autoclass:: ray.tune.search.ax.AxSearch +.. autosummary:: + :toctree: doc/ + + ax.AxSearch .. _bayesopt: Bayesian Optimization (tune.search.bayesopt.BayesOptSearch) ----------------------------------------------------------- +.. autosummary:: + :toctree: doc/ -.. autoclass:: ray.tune.search.bayesopt.BayesOptSearch - :members: save, restore - -.. _`BayesianOptimization search space specification`: https://github.com/fmfn/BayesianOptimization/blob/master/examples/advanced-tour.ipynb + bayesopt.BayesOptSearch .. _suggest-TuneBOHB: @@ -125,7 +153,10 @@ In order to use this search algorithm, you will need to install ``HpBandSter`` a See the `BOHB paper `_ for more details. -.. autoclass:: ray.tune.search.bohb.TuneBOHB +.. autosummary:: + :toctree: doc/ + + bohb.TuneBOHB .. _BlendSearch: @@ -144,7 +175,10 @@ In order to use this search algorithm, you will need to install ``flaml``: See the `BlendSearch paper `_ and documentation in FLAML `BlendSearch documentation `_ for more details. -.. autoclass:: ray.tune.search.flaml.BlendSearch +.. autosummary:: + :toctree: doc/ + + flaml.BlendSearch .. _CFO: @@ -164,50 +198,60 @@ In order to use this search algorithm, you will need to install ``flaml``: See the `CFO paper `_ and documentation in FLAML `CFO documentation `_ for more details. -.. autoclass:: ray.tune.search.flaml.CFO +.. autosummary:: + :toctree: doc/ + + flaml.CFO .. _Dragonfly: Dragonfly (tune.search.dragonfly.DragonflySearch) ------------------------------------------------- -.. autoclass:: ray.tune.search.dragonfly.DragonflySearch - :members: save, restore +.. autosummary:: + :toctree: doc/ + + dragonfly.DragonflySearch .. _tune-hebo: HEBO (tune.search.hebo.HEBOSearch) ---------------------------------- -.. autoclass:: ray.tune.search.hebo.HEBOSearch - :members: save, restore +.. autosummary:: + :toctree: doc/ + + hebo.HEBOSearch .. _tune-hyperopt: HyperOpt (tune.search.hyperopt.HyperOptSearch) ---------------------------------------------- -.. autoclass:: ray.tune.search.hyperopt.HyperOptSearch - :members: save, restore +.. autosummary:: + :toctree: doc/ + + hyperopt.HyperOptSearch .. _nevergrad: Nevergrad (tune.search.nevergrad.NevergradSearch) ------------------------------------------------- -.. autoclass:: ray.tune.search.nevergrad.NevergradSearch - :members: save, restore +.. autosummary:: + :toctree: doc/ -.. _`Nevergrad README's Optimization section`: https://github.com/facebookresearch/nevergrad/blob/master/docs/optimization.rst#choosing-an-optimizer + nevergrad.NevergradSearch .. _tune-optuna: Optuna (tune.search.optuna.OptunaSearch) ---------------------------------------- -.. autoclass:: ray.tune.search.optuna.OptunaSearch +.. autosummary:: + :toctree: doc/ -.. _`Optuna samplers`: https://optuna.readthedocs.io/en/stable/reference/samplers.html + optuna.OptunaSearch .. _sigopt: @@ -217,25 +261,30 @@ SigOpt (tune.search.sigopt.SigOptSearch) You will need to use the `SigOpt experiment and space specification `__ to specify your search space. -.. autoclass:: ray.tune.search.sigopt.SigOptSearch +.. autosummary:: + :toctree: doc/ + + sigopt.SigOptSearch .. _skopt: Scikit-Optimize (tune.search.skopt.SkOptSearch) ----------------------------------------------- -.. autoclass:: ray.tune.search.skopt.SkOptSearch - :members: save, restore +.. autosummary:: + :toctree: doc/ -.. _`skopt Optimizer object`: https://scikit-optimize.github.io/stable/modules/generated/skopt.Optimizer.html#skopt.Optimizer + skopt.SkOptSearch .. _zoopt: ZOOpt (tune.search.zoopt.ZOOptSearch) ------------------------------------- -.. autoclass:: ray.tune.search.zoopt.ZOOptSearch - :members: save, restore +.. autosummary:: + :toctree: doc/ + + zoopt.ZOOptSearch .. _repeater: @@ -255,7 +304,10 @@ will run ``repeat`` trials of the configuration. It will then average the .. warning:: It is recommended to not use ``Repeater`` with a TrialScheduler. Early termination can negatively affect the average reported metric. -.. autoclass:: ray.tune.search.Repeater +.. autosummary:: + :toctree: doc/ + + Repeater .. _limiter: @@ -265,7 +317,10 @@ ConcurrencyLimiter (tune.search.ConcurrencyLimiter) Use ``ray.tune.search.ConcurrencyLimiter`` to limit the amount of concurrency when using a search algorithm. This is useful when a given optimization algorithm does not parallelize very well (like a naive Bayesian Optimization). -.. autoclass:: ray.tune.search.ConcurrencyLimiter +.. autosummary:: + :toctree: doc/ + + ConcurrencyLimiter .. _byo-algo: @@ -274,11 +329,15 @@ Custom Search Algorithms (tune.search.Searcher) If you are interested in implementing or contributing a new Search Algorithm, provide the following interface: -.. autoclass:: ray.tune.search.Searcher - :members: - :private-members: - :show-inheritance: +.. autosummary:: + :toctree: doc/ + Searcher + Searcher.suggest + Searcher.save + Searcher.restore + Searcher.on_trial_result + Searcher.on_trial_complete If contributing, make sure to add test cases and an entry in the function described below. @@ -290,4 +349,7 @@ There is also a shim function that constructs the search algorithm based on the This can be useful if the search algorithm you want to use changes often (e.g., specifying the search algorithm via a CLI option or config file). -.. automethod:: ray.tune.create_searcher +.. autosummary:: + :toctree: doc/ + + create_searcher diff --git a/doc/source/tune/api/syncing.rst b/doc/source/tune/api/syncing.rst new file mode 100644 index 000000000000..d7f24fd44920 --- /dev/null +++ b/doc/source/tune/api/syncing.rst @@ -0,0 +1,46 @@ +Syncing in Tune (tune.SyncConfig, tune.Syncer) +============================================== + +.. seealso:: + + See :doc:`this user guide ` for more details and examples. + + +.. currentmodule:: ray.tune.syncer + +.. _tune-sync-config: + +Tune Syncing Configuration +-------------------------- + +.. autosummary:: + :toctree: doc/ + + SyncConfig + +.. _tune-syncer: + +Remote Storage Syncer Interface (tune.Syncer) +--------------------------------------------- + +.. autosummary:: + :toctree: doc/ + + Syncer + Syncer.sync_up + Syncer.sync_down + Syncer.delete + Syncer.wait + Syncer.wait_or_retry + + +Tune Built-in Syncers +--------------------- + +.. autosummary:: + :toctree: doc/ + + SyncerCallback + _DefaultSyncer + _BackgroundSyncer + diff --git a/doc/source/tune/api_docs/trainable.rst b/doc/source/tune/api/trainable.rst similarity index 90% rename from doc/source/tune/api_docs/trainable.rst rename to doc/source/tune/api/trainable.rst index 177fc60b6296..3b46f22694c1 100644 --- a/doc/source/tune/api_docs/trainable.rst +++ b/doc/source/tune/api/trainable.rst @@ -240,7 +240,7 @@ Trainables can themselves be distributed. If your trainable function / class cre that also consume CPU / GPU resources, you will want to add more bundles to the :class:`PlacementGroupFactory` to reserve extra resource slots. For example, if a trainable class requires 1 GPU itself, but also launches 4 actors, each using another GPU, -then you should use :ref:`tune-with-resources` like this: +then you should use :func:`tune.with_resources ` like this: .. code-block:: python :emphasize-lines: 4-10 @@ -267,56 +267,72 @@ It is also possible to specify memory (``"memory"``, in bytes) and custom resour session (Function API) ---------------------- -.. autofunction:: ray.air.session.report - :noindex: +.. currentmodule:: ray -.. autofunction:: ray.air.session.get_checkpoint - :noindex: +.. autosummary:: + :toctree: doc/ -.. autofunction:: ray.air.session.get_trial_name + air.session.report :noindex: - -.. autofunction:: ray.air.session.get_trial_id + air.session.get_checkpoint :noindex: - -.. autofunction:: ray.air.session.get_trial_resources + air.session.get_trial_name :noindex: - -.. autofunction:: ray.air.session.get_trial_dir + air.session.get_trial_id + :noindex: + air.session.get_trial_resources + :noindex: + air.session.get_trial_dir :noindex: .. _tune-trainable-docstring: -tune.Trainable (Class API) --------------------------- +Trainable (Class API) +--------------------- + +.. autosummary:: + :toctree: doc/ + + tune.Trainable + tune.Trainable.setup + tune.Trainable.save_checkpoint + tune.Trainable.load_checkpoint + tune.Trainable.step + tune.Trainable.reset_config + tune.Trainable.cleanup + tune.Trainable.default_resource_request -.. autoclass:: ray.tune.Trainable - :member-order: groupwise - :private-members: - :members: .. _tune-util-ref: -Utilities ---------- +Tune Trainable Utilities +------------------------- + +Tune Data Ingestion Utilities +~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ -.. autofunction:: ray.tune.utils.wait_for_gpu +.. autosummary:: + :toctree: doc/ -.. autofunction:: ray.tune.utils.diagnose_serialization + tune.with_parameters -.. autofunction:: ray.tune.utils.validate_save_restore +Tune Resource Assignment Utilities +~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ -.. _tune-with-parameters: +.. autosummary:: + :toctree: doc/ -tune.with_parameters --------------------- + tune.with_resources + ~tune.execution.placement_groups.PlacementGroupFactory + tune.utils.wait_for_gpu -.. autofunction:: ray.tune.with_parameters -.. _tune-with-resources: +Tune Trainable Debugging Utilities +~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ -tune.with_resources --------------------- +.. autosummary:: + :toctree: doc/ -.. autofunction:: ray.tune.with_resources \ No newline at end of file + tune.utils.diagnose_serialization + tune.utils.validate_save_restore diff --git a/doc/source/tune/api_docs/execution.rst b/doc/source/tune/api_docs/execution.rst deleted file mode 100644 index 834b5c669f78..000000000000 --- a/doc/source/tune/api_docs/execution.rst +++ /dev/null @@ -1,26 +0,0 @@ -Tune Execution (Tuner, tune.Experiment) -======================================= - -.. _tune-run-ref: - -Tuner ------ - -.. autofunction:: ray.tune.Tuner - -tune.run_experiments --------------------- - -.. autofunction:: ray.tune.run_experiments - -tune.Experiment ---------------- - -.. autofunction:: ray.tune.Experiment - -.. _tune-sync-config: - -tune.SyncConfig ---------------- - -.. autofunction:: ray.tune.SyncConfig diff --git a/doc/source/tune/api_docs/integration.rst b/doc/source/tune/api_docs/integration.rst deleted file mode 100644 index 63daaa76b4f0..000000000000 --- a/doc/source/tune/api_docs/integration.rst +++ /dev/null @@ -1,91 +0,0 @@ -.. _tune-integration: - -External library integrations for Ray Tune (tune.integration) -============================================================= - -.. contents:: - :local: - :depth: 1 - - -Comet (tune.integration.comet) -------------------------------------------- - -:ref:`See also here `. - -.. autoclass:: ray.air.integrations.comet.CometLoggerCallback - :noindex: - -.. _tune-integration-keras: - -Keras (tune.integration.keras) ------------------------------------------------------- - -.. autoclass:: ray.tune.integration.keras.TuneReportCallback - -.. autoclass:: ray.tune.integration.keras.TuneReportCheckpointCallback - - -.. _tune-integration-mlflow: - -MLflow (tune.integration.mlflow) --------------------------------- - -:ref:`See also here `. - -.. autoclass:: ray.air.integrations.mlflow.MLflowLoggerCallback - :noindex: - -.. autofunction:: ray.air.integrations.mlflow.setup_mlflow - - -.. _tune-integration-mxnet: - -MXNet (tune.integration.mxnet) ------------------------------- - -.. autoclass:: ray.tune.integration.mxnet.TuneReportCallback - -.. autoclass:: ray.tune.integration.mxnet.TuneCheckpointCallback - - -.. _tune-integration-pytorch-lightning: - -PyTorch Lightning (tune.integration.pytorch_lightning) ------------------------------------------------------- - -.. autoclass:: ray.tune.integration.pytorch_lightning.TuneReportCallback - -.. autoclass:: ray.tune.integration.pytorch_lightning.TuneReportCheckpointCallback - -.. _tune-integration-wandb: - -Weights and Biases (tune.integration.wandb) -------------------------------------------- - -:ref:`See also here `. - -.. autoclass:: ray.air.integrations.wandb.WandbLoggerCallback - :noindex: - -.. autofunction:: ray.air.integrations.wandb.setup_wandb - - -.. _tune-integration-xgboost: - -XGBoost (tune.integration.xgboost) ----------------------------------- - -.. autoclass:: ray.tune.integration.xgboost.TuneReportCallback - -.. autoclass:: ray.tune.integration.xgboost.TuneReportCheckpointCallback - - -.. _tune-integration-lightgbm: - -LightGBM (tune.integration.lightgbm) ------------------------------------- - -.. autoclass:: ray.tune.integration.lightgbm.TuneReportCallback - -.. autoclass:: ray.tune.integration.lightgbm.TuneReportCheckpointCallback diff --git a/doc/source/tune/api_docs/result_grid.rst b/doc/source/tune/api_docs/result_grid.rst deleted file mode 100644 index 9333494237bb..000000000000 --- a/doc/source/tune/api_docs/result_grid.rst +++ /dev/null @@ -1,25 +0,0 @@ -.. _tune-analysis-docs: - -.. _result-grid-docstring: - -ResultGrid (tune.ResultGrid) ----------------------------- - -.. autoclass:: ray.tune.ResultGrid - :members: - -.. _result-docstring: - -Result (air.Result) -------------------- - -.. autoclass:: ray.air.Result - :members: - -.. _exp-analysis-docstring: - -ExperimentAnalysis (tune.ExperimentAnalysis) --------------------------------------------- - -.. autoclass:: ray.tune.ExperimentAnalysis - :members: diff --git a/doc/source/tune/api_docs/stoppers.rst b/doc/source/tune/api_docs/stoppers.rst deleted file mode 100644 index 30edfa793e35..000000000000 --- a/doc/source/tune/api_docs/stoppers.rst +++ /dev/null @@ -1,52 +0,0 @@ -.. _tune-stoppers: - -Tune Stopping mechanisms (tune.stopper) -======================================= - -In addition to Trial Schedulers like :ref:`ASHA `, where a number of -trials are stopped if they perform subpar, Ray Tune also supports custom stopping mechanisms to stop trials early. They can also stop the entire experiment after a condition is met. -For instance, stopping mechanisms can specify to stop trials when they reached a plateau and the metric -doesn't change anymore. - -Ray Tune comes with several stopping mechanisms out of the box. For custom stopping behavior, you can -inherit from the :class:`Stopper ` class. - -Other stopping behaviors are described :ref:`in the user guide `. - -.. contents:: - :local: - :depth: 1 - - -.. _tune-stop-ref: - -Stopper (tune.Stopper) ----------------------- - -.. autoclass:: ray.tune.Stopper - :members: __call__, stop_all - -MaximumIterationStopper (tune.stopper.MaximumIterationStopper) --------------------------------------------------------------- - -.. autoclass:: ray.tune.stopper.MaximumIterationStopper - -ExperimentPlateauStopper (tune.stopper.ExperimentPlateauStopper) ----------------------------------------------------------------- - -.. autoclass:: ray.tune.stopper.ExperimentPlateauStopper - -TrialPlateauStopper (tune.stopper.TrialPlateauStopper) ------------------------------------------------------- - -.. autoclass:: ray.tune.stopper.TrialPlateauStopper - -TimeoutStopper (tune.stopper.TimeoutStopper) --------------------------------------------- - -.. autoclass:: ray.tune.stopper.TimeoutStopper - -CombinedStopper (tune.stopper.CombinedStopper) ----------------------------------------------- - -.. autoclass:: ray.tune.stopper.CombinedStopper diff --git a/doc/source/tune/api_docs/syncing.rst b/doc/source/tune/api_docs/syncing.rst deleted file mode 100644 index 17e030794e26..000000000000 --- a/doc/source/tune/api_docs/syncing.rst +++ /dev/null @@ -1,20 +0,0 @@ -Syncing in Tune (tune.SyncConfig, tune.Syncer) -============================================== - -.. _tune-syncconfig: - -SyncConfig ----------- - -.. autoclass:: ray.tune.syncer.SyncConfig - :members: - - - -.. _tune-syncer: - -Syncer ------- - -.. autoclass:: ray.tune.syncer.Syncer - :members: diff --git a/doc/source/tune/examples/tune-xgboost.ipynb b/doc/source/tune/examples/tune-xgboost.ipynb index b732c1879fb9..edf640b0b30c 100644 --- a/doc/source/tune/examples/tune-xgboost.ipynb +++ b/doc/source/tune/examples/tune-xgboost.ipynb @@ -1,1255 +1,1255 @@ { - "cells": [ - { - "cell_type": "markdown", - "id": "edce67b9", - "metadata": {}, - "source": [ - "# Tuning XGBoost hyperparameters with Ray Tune\n", - "\n", - "(tune-xgboost-ref)=\n", - "\n", - "XGBoost is currently one of the most popular machine learning algorithms. It performs\n", - "very well on a large selection of tasks, and was the key to success in many Kaggle\n", - "competitions.\n", - "\n", - "```{image} /images/xgboost_logo.png\n", - ":align: center\n", - ":alt: XGBoost\n", - ":target: https://xgboost.readthedocs.io/en/latest/\n", - ":width: 200px\n", - "```\n", - "\n", - "This tutorial will give you a quick introduction to XGBoost, show you how\n", - "to train an XGBoost model, and then guide you on how to optimize XGBoost\n", - "parameters using Tune to get the best performance. We tackle the following topics:\n", - "\n", - "```{contents}\n", - ":depth: 2\n", - "```\n", - "\n", - ":::{note}\n", - "To run this tutorial, you will need to install the following:\n", - "\n", - "```bash\n", - "$ pip install xgboost\n", - "```\n", - ":::\n", - "\n", - "## What is XGBoost\n", - "\n", - "XGBoost is an acronym for e**X**treme **G**radient **Boost**ing. Internally,\n", - "XGBoost uses [decision trees](https://en.wikipedia.org/wiki/Decision_tree). Instead\n", - "of training just one large decision tree, XGBoost and other related algorithms train\n", - "many small decision trees. The intuition behind this is that even though single\n", - "decision trees can be inaccurate and suffer from high variance,\n", - "combining the output of a large number of these weak learners can actually lead to\n", - "strong learner, resulting in better predictions and less variance.\n", - "\n", - ":::{figure} /images/tune-xgboost-ensemble.svg\n", - ":alt: Single vs. ensemble learning\n", - "\n", - "A single decision tree (left) might be able to get to an accuracy of 70%\n", - "for a binary classification task. By combining the output of several small\n", - "decision trees, an ensemble learner (right) might end up with a higher accuracy\n", - "of 90%.\n", - ":::\n", - "\n", - "Boosting algorithms start with a single small decision tree and evaluate how well\n", - "it predicts the given examples. When building the next tree, those samples that have\n", - "been misclassified before have a higher chance of being used to generate the tree.\n", - "This is useful because it avoids overfitting to samples that can be easily classified\n", - "and instead tries to come up with models that are able to classify hard examples, too.\n", - "Please see [here for a more thorough introduction to bagging and boosting algorithms](https://towardsdatascience.com/ensemble-methods-bagging-boosting-and-stacking-c9214a10a205).\n", - "\n", - "There are many boosting algorithms. In their core, they are all very similar. XGBoost\n", - "uses second-level derivatives to find splits that maximize the *gain* (the inverse of\n", - "the *loss*) - hence the name. In practice, there really is no drawback in using\n", - "XGBoost over other boosting algorithms - in fact, it usually shows the best performance.\n", - "\n", - "## Training a simple XGBoost classifier\n", - "\n", - "Let's first see how a simple XGBoost classifier can be trained. We'll use the\n", - "`breast_cancer`-Dataset included in the `sklearn` dataset collection. This is\n", - "a binary classification dataset. Given 30 different input features, our task is to\n", - "learn to identify subjects with breast cancer and those without.\n", - "\n", - "Here is the full code to train a simple XGBoost model:" - ] - }, - { - "cell_type": "code", - "execution_count": 1, - "id": "77b3c71c", - "metadata": {}, - "outputs": [ - { - "name": "stdout", - "output_type": "stream", - "text": [ - "Accuracy: 0.9650\n" - ] - } - ], - "source": [ - "import sklearn.datasets\n", - "import sklearn.metrics\n", - "from sklearn.model_selection import train_test_split\n", - "import xgboost as xgb\n", - "\n", - "\n", - "def train_breast_cancer(config):\n", - " # Load dataset\n", - " data, labels = sklearn.datasets.load_breast_cancer(return_X_y=True)\n", - " # Split into train and test set\n", - " train_x, test_x, train_y, test_y = train_test_split(data, labels, test_size=0.25)\n", - " # Build input matrices for XGBoost\n", - " train_set = xgb.DMatrix(train_x, label=train_y)\n", - " test_set = xgb.DMatrix(test_x, label=test_y)\n", - " # Train the classifier\n", - " results = {}\n", - " bst = xgb.train(\n", - " config,\n", - " train_set,\n", - " evals=[(test_set, \"eval\")],\n", - " evals_result=results,\n", - " verbose_eval=False,\n", - " )\n", - " return results\n", - "\n", - "\n", - "if __name__ == \"__main__\":\n", - " results = train_breast_cancer(\n", - " {\"objective\": \"binary:logistic\", \"eval_metric\": [\"logloss\", \"error\"]}\n", - " )\n", - " accuracy = 1.0 - results[\"eval\"][\"error\"][-1]\n", - " print(f\"Accuracy: {accuracy:.4f}\")\n" - ] - }, - { - "cell_type": "markdown", - "id": "ec2a13f8", - "metadata": {}, - "source": [ - "As you can see, the code is quite simple. First, the dataset is loaded and split\n", - "into a `test` and `train` set. The XGBoost model is trained with `xgb.train()`.\n", - "XGBoost automatically evaluates metrics we specified on the test set. In our case\n", - "it calculates the *logloss* and the prediction *error*, which is the percentage of\n", - "misclassified examples. To calculate the accuracy, we just have to subtract the error\n", - "from `1.0`. Even in this simple example, most runs result\n", - "in a good accuracy of over `0.90`.\n", - "\n", - "Maybe you have noticed the `config` parameter we pass to the XGBoost algorithm. This\n", - "is a {class}`dict` in which you can specify parameters for the XGBoost algorithm. In this\n", - "simple example, the only parameters we passed are the `objective` and `eval_metric` parameters.\n", - "The value `binary:logistic` tells XGBoost that we aim to train a logistic regression model for\n", - "a binary classification task. You can find an overview over all valid objectives\n", - "[here in the XGBoost documentation](https://xgboost.readthedocs.io/en/latest/parameter.html#learning-task-parameters).\n", - "\n", - "## XGBoost Hyperparameters\n", - "\n", - "Even with the default settings, XGBoost was able to get to a good accuracy on the\n", - "breast cancer dataset. However, as in many machine learning algorithms, there are\n", - "many knobs to tune which might lead to even better performance. Let's explore some of\n", - "them below.\n", - "\n", - "### Maximum tree depth\n", - "\n", - "Remember that XGBoost internally uses many decision tree models to come up with\n", - "predictions. When training a decision tree, we need to tell the algorithm how\n", - "large the tree may get. The parameter for this is called the tree *depth*.\n", - "\n", - ":::{figure} /images/tune-xgboost-depth.svg\n", - ":align: center\n", - ":alt: Decision tree depth\n", - "\n", - "In this image, the left tree has a depth of 2, and the right tree a depth of 3.\n", - "Note that with each level, $2^{(d-1)}$ splits are added, where *d* is the depth\n", - "of the tree.\n", - ":::\n", - "\n", - "Tree depth is a property that concerns the model complexity. If you only allow short\n", - "trees, the models are likely not very precise - they underfit the data. If you allow\n", - "very large trees, the single models are likely to overfit to the data. In practice,\n", - "a number between `2` and `6` is often a good starting point for this parameter.\n", - "\n", - "XGBoost's default value is `3`.\n", - "\n", - "### Minimum child weight\n", - "\n", - "When a decision tree creates new leaves, it splits up the remaining data at one node\n", - "into two groups. If there are only few samples in one of these groups, it often\n", - "doesn't make sense to split it further. One of the reasons for this is that the\n", - "model is harder to train when we have fewer samples.\n", - "\n", - ":::{figure} /images/tune-xgboost-weight.svg\n", - ":align: center\n", - ":alt: Minimum child weight\n", - "\n", - "In this example, we start with 100 examples. At the first node, they are split\n", - "into 4 and 96 samples, respectively. In the next step, our model might find\n", - "that it doesn't make sense to split the 4 examples more. It thus only continues\n", - "to add leaves on the right side.\n", - ":::\n", - "\n", - "The parameter used by the model to decide if it makes sense to split a node is called\n", - "the *minimum child weight*. In the case of linear regression, this is just the absolute\n", - "number of nodes requried in each child. In other objectives, this value is determined\n", - "using the weights of the examples, hence the name.\n", - "\n", - "The larger the value, the more constrained the trees are and the less deep they will be.\n", - "This parameter thus also affects the model complexity. Values can range between 0\n", - "and infinity and are dependent on the sample size. For our ca. 500 examples in the\n", - "breast cancer dataset, values between `0` and `10` should be sensible.\n", - "\n", - "XGBoost's default value is `1`.\n", - "\n", - "### Subsample size\n", - "\n", - "Each decision tree we add is trained on a subsample of the total training dataset.\n", - "The probabilities for the samples are weighted according to the XGBoost algorithm,\n", - "but we can decide on which fraction of the samples we want to train each decision\n", - "tree on.\n", - "\n", - "Setting this value to `0.7` would mean that we randomly sample `70%` of the\n", - "training dataset before each training iteration.\n", - "\n", - "XGBoost's default value is `1`.\n", - "\n", - "### Learning rate / Eta\n", - "\n", - "Remember that XGBoost sequentially trains many decision trees, and that later trees\n", - "are more likely trained on data that has been misclassified by prior trees. In effect\n", - "this means that earlier trees make decisions for easy samples (i.e. those samples that\n", - "can easily be classified) and later trees make decisions for harder samples. It is then\n", - "sensible to assume that the later trees are less accurate than earlier trees.\n", - "\n", - "To address this fact, XGBoost uses a parameter called *Eta*, which is sometimes called\n", - "the *learning rate*. Don't confuse this with learning rates from gradient descent!\n", - "The original [paper on stochastic gradient boosting](https://www.sciencedirect.com/science/article/abs/pii/S0167947301000652)\n", - "introduces this parameter like so:\n", - "\n", - "$$\n", - "F_m(x) = F_{m-1}(x) + \\eta \\cdot \\gamma_{lm} \\textbf{1}(x \\in R_{lm})\n", - "$$\n", - "\n", - "This is just a complicated way to say that when we train we new decision tree,\n", - "represented by $\\gamma_{lm} \\textbf{1}(x \\in R_{lm})$, we want to dampen\n", - "its effect on the previous prediction $F_{m-1}(x)$ with a factor\n", - "$\\eta$.\n", - "\n", - "Typical values for this parameter are between `0.01` and `` 0.3` ``.\n", - "\n", - "XGBoost's default value is `0.3`.\n", - "\n", - "### Number of boost rounds\n", - "\n", - "Lastly, we can decide on how many boosting rounds we perform, which means how\n", - "many decision trees we ultimately train. When we do heavy subsampling or use small\n", - "learning rate, it might make sense to increase the number of boosting rounds.\n", - "\n", - "XGBoost's default value is `10`.\n", - "\n", - "### Putting it together\n", - "\n", - "Let's see how this looks like in code! We just need to adjust our `config` dict:" - ] - }, - { - "cell_type": "code", - "execution_count": 2, - "id": "35073e88", - "metadata": {}, - "outputs": [ - { - "name": "stdout", - "output_type": "stream", - "text": [ - "Accuracy: 0.9790\n" - ] - } - ], - "source": [ - "if __name__ == \"__main__\":\n", - " config = {\n", - " \"objective\": \"binary:logistic\",\n", - " \"eval_metric\": [\"logloss\", \"error\"],\n", - " \"max_depth\": 2,\n", - " \"min_child_weight\": 0,\n", - " \"subsample\": 0.8,\n", - " \"eta\": 0.2,\n", - " }\n", - " results = train_breast_cancer(config)\n", - " accuracy = 1.0 - results[\"eval\"][\"error\"][-1]\n", - " print(f\"Accuracy: {accuracy:.4f}\")\n" - ] - }, - { - "cell_type": "markdown", - "id": "69cf0c13", - "metadata": {}, - "source": [ - "The rest stays the same. Please note that we do not adjust the `num_boost_rounds` here.\n", - "The result should also show a high accuracy of over 90%.\n", - "\n", - "## Tuning the configuration parameters\n", - "\n", - "XGBoosts default parameters already lead to a good accuracy, and even our guesses in the\n", - "last section should result in accuracies well above 90%. However, our guesses were\n", - "just that: guesses. Often we do not know what combination of parameters would actually\n", - "lead to the best results on a machine learning task.\n", - "\n", - "Unfortunately, there are infinitely many combinations of hyperparameters we could try\n", - "out. Should we combine `max_depth=3` with `subsample=0.8` or with `subsample=0.9`?\n", - "What about the other parameters?\n", - "\n", - "This is where hyperparameter tuning comes into play. By using tuning libraries such as\n", - "Ray Tune we can try out combinations of hyperparameters. Using sophisticated search\n", - "strategies, these parameters can be selected so that they are likely to lead to good\n", - "results (avoiding an expensive *exhaustive search*). Also, trials that do not perform\n", - "well can be preemptively stopped to reduce waste of computing resources. Lastly, Ray Tune\n", - "also takes care of training these runs in parallel, greatly increasing search speed.\n", - "\n", - "Let's start with a basic example on how to use Tune for this. We just need to make\n", - "a few changes to our code-block:" - ] - }, - { - "cell_type": "code", - "execution_count": 3, - "id": "ff856a82", - "metadata": {}, - "outputs": [ - { - "name": "stderr", - "output_type": "stream", - "text": [ - "2022-07-22 15:52:52,004\tINFO services.py:1483 -- View the Ray dashboard at \u001b[1m\u001b[32mhttp://127.0.0.1:8268\u001b[39m\u001b[22m\n", - "2022-07-22 15:52:55,858\tWARNING function_trainable.py:619 -- Function checkpointing is disabled. This may result in unexpected behavior when using checkpointing features or certain schedulers. To enable, set the train function arguments to be `func(config, checkpoint_dir=None)`.\n" - ] + "cells": [ + { + "cell_type": "markdown", + "id": "edce67b9", + "metadata": {}, + "source": [ + "# Tuning XGBoost hyperparameters with Ray Tune\n", + "\n", + "(tune-xgboost-ref)=\n", + "\n", + "XGBoost is currently one of the most popular machine learning algorithms. It performs\n", + "very well on a large selection of tasks, and was the key to success in many Kaggle\n", + "competitions.\n", + "\n", + "```{image} /images/xgboost_logo.png\n", + ":align: center\n", + ":alt: XGBoost\n", + ":target: https://xgboost.readthedocs.io/en/latest/\n", + ":width: 200px\n", + "```\n", + "\n", + "This tutorial will give you a quick introduction to XGBoost, show you how\n", + "to train an XGBoost model, and then guide you on how to optimize XGBoost\n", + "parameters using Tune to get the best performance. We tackle the following topics:\n", + "\n", + "```{contents}\n", + ":depth: 2\n", + "```\n", + "\n", + ":::{note}\n", + "To run this tutorial, you will need to install the following:\n", + "\n", + "```bash\n", + "$ pip install xgboost\n", + "```\n", + ":::\n", + "\n", + "## What is XGBoost\n", + "\n", + "XGBoost is an acronym for e**X**treme **G**radient **Boost**ing. Internally,\n", + "XGBoost uses [decision trees](https://en.wikipedia.org/wiki/Decision_tree). Instead\n", + "of training just one large decision tree, XGBoost and other related algorithms train\n", + "many small decision trees. The intuition behind this is that even though single\n", + "decision trees can be inaccurate and suffer from high variance,\n", + "combining the output of a large number of these weak learners can actually lead to\n", + "strong learner, resulting in better predictions and less variance.\n", + "\n", + ":::{figure} /images/tune-xgboost-ensemble.svg\n", + ":alt: Single vs. ensemble learning\n", + "\n", + "A single decision tree (left) might be able to get to an accuracy of 70%\n", + "for a binary classification task. By combining the output of several small\n", + "decision trees, an ensemble learner (right) might end up with a higher accuracy\n", + "of 90%.\n", + ":::\n", + "\n", + "Boosting algorithms start with a single small decision tree and evaluate how well\n", + "it predicts the given examples. When building the next tree, those samples that have\n", + "been misclassified before have a higher chance of being used to generate the tree.\n", + "This is useful because it avoids overfitting to samples that can be easily classified\n", + "and instead tries to come up with models that are able to classify hard examples, too.\n", + "Please see [here for a more thorough introduction to bagging and boosting algorithms](https://towardsdatascience.com/ensemble-methods-bagging-boosting-and-stacking-c9214a10a205).\n", + "\n", + "There are many boosting algorithms. In their core, they are all very similar. XGBoost\n", + "uses second-level derivatives to find splits that maximize the *gain* (the inverse of\n", + "the *loss*) - hence the name. In practice, there really is no drawback in using\n", + "XGBoost over other boosting algorithms - in fact, it usually shows the best performance.\n", + "\n", + "## Training a simple XGBoost classifier\n", + "\n", + "Let's first see how a simple XGBoost classifier can be trained. We'll use the\n", + "`breast_cancer`-Dataset included in the `sklearn` dataset collection. This is\n", + "a binary classification dataset. Given 30 different input features, our task is to\n", + "learn to identify subjects with breast cancer and those without.\n", + "\n", + "Here is the full code to train a simple XGBoost model:" + ] + }, + { + "cell_type": "code", + "execution_count": 1, + "id": "77b3c71c", + "metadata": {}, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "Accuracy: 0.9650\n" + ] + } + ], + "source": [ + "import sklearn.datasets\n", + "import sklearn.metrics\n", + "from sklearn.model_selection import train_test_split\n", + "import xgboost as xgb\n", + "\n", + "\n", + "def train_breast_cancer(config):\n", + " # Load dataset\n", + " data, labels = sklearn.datasets.load_breast_cancer(return_X_y=True)\n", + " # Split into train and test set\n", + " train_x, test_x, train_y, test_y = train_test_split(data, labels, test_size=0.25)\n", + " # Build input matrices for XGBoost\n", + " train_set = xgb.DMatrix(train_x, label=train_y)\n", + " test_set = xgb.DMatrix(test_x, label=test_y)\n", + " # Train the classifier\n", + " results = {}\n", + " bst = xgb.train(\n", + " config,\n", + " train_set,\n", + " evals=[(test_set, \"eval\")],\n", + " evals_result=results,\n", + " verbose_eval=False,\n", + " )\n", + " return results\n", + "\n", + "\n", + "if __name__ == \"__main__\":\n", + " results = train_breast_cancer(\n", + " {\"objective\": \"binary:logistic\", \"eval_metric\": [\"logloss\", \"error\"]}\n", + " )\n", + " accuracy = 1.0 - results[\"eval\"][\"error\"][-1]\n", + " print(f\"Accuracy: {accuracy:.4f}\")\n" + ] + }, + { + "cell_type": "markdown", + "id": "ec2a13f8", + "metadata": {}, + "source": [ + "As you can see, the code is quite simple. First, the dataset is loaded and split\n", + "into a `test` and `train` set. The XGBoost model is trained with `xgb.train()`.\n", + "XGBoost automatically evaluates metrics we specified on the test set. In our case\n", + "it calculates the *logloss* and the prediction *error*, which is the percentage of\n", + "misclassified examples. To calculate the accuracy, we just have to subtract the error\n", + "from `1.0`. Even in this simple example, most runs result\n", + "in a good accuracy of over `0.90`.\n", + "\n", + "Maybe you have noticed the `config` parameter we pass to the XGBoost algorithm. This\n", + "is a {class}`dict` in which you can specify parameters for the XGBoost algorithm. In this\n", + "simple example, the only parameters we passed are the `objective` and `eval_metric` parameters.\n", + "The value `binary:logistic` tells XGBoost that we aim to train a logistic regression model for\n", + "a binary classification task. You can find an overview over all valid objectives\n", + "[here in the XGBoost documentation](https://xgboost.readthedocs.io/en/latest/parameter.html#learning-task-parameters).\n", + "\n", + "## XGBoost Hyperparameters\n", + "\n", + "Even with the default settings, XGBoost was able to get to a good accuracy on the\n", + "breast cancer dataset. However, as in many machine learning algorithms, there are\n", + "many knobs to tune which might lead to even better performance. Let's explore some of\n", + "them below.\n", + "\n", + "### Maximum tree depth\n", + "\n", + "Remember that XGBoost internally uses many decision tree models to come up with\n", + "predictions. When training a decision tree, we need to tell the algorithm how\n", + "large the tree may get. The parameter for this is called the tree *depth*.\n", + "\n", + ":::{figure} /images/tune-xgboost-depth.svg\n", + ":align: center\n", + ":alt: Decision tree depth\n", + "\n", + "In this image, the left tree has a depth of 2, and the right tree a depth of 3.\n", + "Note that with each level, $2^{(d-1)}$ splits are added, where *d* is the depth\n", + "of the tree.\n", + ":::\n", + "\n", + "Tree depth is a property that concerns the model complexity. If you only allow short\n", + "trees, the models are likely not very precise - they underfit the data. If you allow\n", + "very large trees, the single models are likely to overfit to the data. In practice,\n", + "a number between `2` and `6` is often a good starting point for this parameter.\n", + "\n", + "XGBoost's default value is `3`.\n", + "\n", + "### Minimum child weight\n", + "\n", + "When a decision tree creates new leaves, it splits up the remaining data at one node\n", + "into two groups. If there are only few samples in one of these groups, it often\n", + "doesn't make sense to split it further. One of the reasons for this is that the\n", + "model is harder to train when we have fewer samples.\n", + "\n", + ":::{figure} /images/tune-xgboost-weight.svg\n", + ":align: center\n", + ":alt: Minimum child weight\n", + "\n", + "In this example, we start with 100 examples. At the first node, they are split\n", + "into 4 and 96 samples, respectively. In the next step, our model might find\n", + "that it doesn't make sense to split the 4 examples more. It thus only continues\n", + "to add leaves on the right side.\n", + ":::\n", + "\n", + "The parameter used by the model to decide if it makes sense to split a node is called\n", + "the *minimum child weight*. In the case of linear regression, this is just the absolute\n", + "number of nodes requried in each child. In other objectives, this value is determined\n", + "using the weights of the examples, hence the name.\n", + "\n", + "The larger the value, the more constrained the trees are and the less deep they will be.\n", + "This parameter thus also affects the model complexity. Values can range between 0\n", + "and infinity and are dependent on the sample size. For our ca. 500 examples in the\n", + "breast cancer dataset, values between `0` and `10` should be sensible.\n", + "\n", + "XGBoost's default value is `1`.\n", + "\n", + "### Subsample size\n", + "\n", + "Each decision tree we add is trained on a subsample of the total training dataset.\n", + "The probabilities for the samples are weighted according to the XGBoost algorithm,\n", + "but we can decide on which fraction of the samples we want to train each decision\n", + "tree on.\n", + "\n", + "Setting this value to `0.7` would mean that we randomly sample `70%` of the\n", + "training dataset before each training iteration.\n", + "\n", + "XGBoost's default value is `1`.\n", + "\n", + "### Learning rate / Eta\n", + "\n", + "Remember that XGBoost sequentially trains many decision trees, and that later trees\n", + "are more likely trained on data that has been misclassified by prior trees. In effect\n", + "this means that earlier trees make decisions for easy samples (i.e. those samples that\n", + "can easily be classified) and later trees make decisions for harder samples. It is then\n", + "sensible to assume that the later trees are less accurate than earlier trees.\n", + "\n", + "To address this fact, XGBoost uses a parameter called *Eta*, which is sometimes called\n", + "the *learning rate*. Don't confuse this with learning rates from gradient descent!\n", + "The original [paper on stochastic gradient boosting](https://www.sciencedirect.com/science/article/abs/pii/S0167947301000652)\n", + "introduces this parameter like so:\n", + "\n", + "$$\n", + "F_m(x) = F_{m-1}(x) + \\eta \\cdot \\gamma_{lm} \\textbf{1}(x \\in R_{lm})\n", + "$$\n", + "\n", + "This is just a complicated way to say that when we train we new decision tree,\n", + "represented by $\\gamma_{lm} \\textbf{1}(x \\in R_{lm})$, we want to dampen\n", + "its effect on the previous prediction $F_{m-1}(x)$ with a factor\n", + "$\\eta$.\n", + "\n", + "Typical values for this parameter are between `0.01` and `` 0.3` ``.\n", + "\n", + "XGBoost's default value is `0.3`.\n", + "\n", + "### Number of boost rounds\n", + "\n", + "Lastly, we can decide on how many boosting rounds we perform, which means how\n", + "many decision trees we ultimately train. When we do heavy subsampling or use small\n", + "learning rate, it might make sense to increase the number of boosting rounds.\n", + "\n", + "XGBoost's default value is `10`.\n", + "\n", + "### Putting it together\n", + "\n", + "Let's see how this looks like in code! We just need to adjust our `config` dict:" + ] + }, + { + "cell_type": "code", + "execution_count": 2, + "id": "35073e88", + "metadata": {}, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "Accuracy: 0.9790\n" + ] + } + ], + "source": [ + "if __name__ == \"__main__\":\n", + " config = {\n", + " \"objective\": \"binary:logistic\",\n", + " \"eval_metric\": [\"logloss\", \"error\"],\n", + " \"max_depth\": 2,\n", + " \"min_child_weight\": 0,\n", + " \"subsample\": 0.8,\n", + " \"eta\": 0.2,\n", + " }\n", + " results = train_breast_cancer(config)\n", + " accuracy = 1.0 - results[\"eval\"][\"error\"][-1]\n", + " print(f\"Accuracy: {accuracy:.4f}\")\n" + ] + }, + { + "cell_type": "markdown", + "id": "69cf0c13", + "metadata": {}, + "source": [ + "The rest stays the same. Please note that we do not adjust the `num_boost_rounds` here.\n", + "The result should also show a high accuracy of over 90%.\n", + "\n", + "## Tuning the configuration parameters\n", + "\n", + "XGBoosts default parameters already lead to a good accuracy, and even our guesses in the\n", + "last section should result in accuracies well above 90%. However, our guesses were\n", + "just that: guesses. Often we do not know what combination of parameters would actually\n", + "lead to the best results on a machine learning task.\n", + "\n", + "Unfortunately, there are infinitely many combinations of hyperparameters we could try\n", + "out. Should we combine `max_depth=3` with `subsample=0.8` or with `subsample=0.9`?\n", + "What about the other parameters?\n", + "\n", + "This is where hyperparameter tuning comes into play. By using tuning libraries such as\n", + "Ray Tune we can try out combinations of hyperparameters. Using sophisticated search\n", + "strategies, these parameters can be selected so that they are likely to lead to good\n", + "results (avoiding an expensive *exhaustive search*). Also, trials that do not perform\n", + "well can be preemptively stopped to reduce waste of computing resources. Lastly, Ray Tune\n", + "also takes care of training these runs in parallel, greatly increasing search speed.\n", + "\n", + "Let's start with a basic example on how to use Tune for this. We just need to make\n", + "a few changes to our code-block:" + ] + }, + { + "cell_type": "code", + "execution_count": 3, + "id": "ff856a82", + "metadata": {}, + "outputs": [ + { + "name": "stderr", + "output_type": "stream", + "text": [ + "2022-07-22 15:52:52,004\tINFO services.py:1483 -- View the Ray dashboard at \u001b[1m\u001b[32mhttp://127.0.0.1:8268\u001b[39m\u001b[22m\n", + "2022-07-22 15:52:55,858\tWARNING function_trainable.py:619 -- Function checkpointing is disabled. This may result in unexpected behavior when using checkpointing features or certain schedulers. To enable, set the train function arguments to be `func(config, checkpoint_dir=None)`.\n" + ] + }, + { + "data": { + "text/html": [ + "== Status ==
Current time: 2022-07-22 15:53:04 (running for 00:00:07.77)
Memory usage on this node: 10.5/16.0 GiB
Using FIFO scheduling algorithm.
Resources requested: 0/16 CPUs, 0/0 GPUs, 0.0/4.57 GiB heap, 0.0/2.0 GiB objects
Result logdir: /Users/kai/ray_results/train_breast_cancer_2022-07-22_15-52-48
Number of trials: 10/10 (10 TERMINATED)
\n", + "\n", + "\n", + "\n", + "\n", + "\n", + "\n", + "\n", + "\n", + "\n", + "\n", + "\n", + "\n", + "\n", + "\n", + "\n", + "
Trial name status loc eta max_depth min_child_weight subsample acc iter total time (s)
train_breast_cancer_f8669_00000TERMINATED127.0.0.1:488520.0069356 5 3 0.8235040.944056 1 0.0316169
train_breast_cancer_f8669_00001TERMINATED127.0.0.1:488570.00145619 6 3 0.8329470.958042 1 0.0328588
train_breast_cancer_f8669_00002TERMINATED127.0.0.1:488580.00108208 7 3 0.9873190.944056 1 0.0319381
train_breast_cancer_f8669_00003TERMINATED127.0.0.1:488590.00530429 8 2 0.6156910.923077 1 0.028388
train_breast_cancer_f8669_00004TERMINATED127.0.0.1:488600.000721843 8 1 0.6509730.958042 1 0.0299618
train_breast_cancer_f8669_00005TERMINATED127.0.0.1:488610.0074509 1 1 0.7383410.874126 1 0.0193682
train_breast_cancer_f8669_00006TERMINATED127.0.0.1:488620.0879882 8 2 0.6715760.944056 1 0.0267372
train_breast_cancer_f8669_00007TERMINATED127.0.0.1:488630.0765404 7 2 0.7081570.965035 1 0.0276129
train_breast_cancer_f8669_00008TERMINATED127.0.0.1:488640.000627649 6 1 0.81121 0.951049 1 0.0310998
train_breast_cancer_f8669_00009TERMINATED127.0.0.1:488650.000383711 2 3 0.9905790.93007 1 0.0274954


" + ], + "text/plain": [ + "" + ] + }, + "metadata": {}, + "output_type": "display_data" + }, + { + "name": "stderr", + "output_type": "stream", + "text": [ + "2022-07-22 15:52:57,385\tINFO plugin_schema_manager.py:52 -- Loading the default runtime env schemas: ['/Users/kai/coding/ray/python/ray/_private/runtime_env/../../runtime_env/schemas/working_dir_schema.json', '/Users/kai/coding/ray/python/ray/_private/runtime_env/../../runtime_env/schemas/pip_schema.json'].\n" + ] + }, + { + "name": "stdout", + "output_type": "stream", + "text": [ + "Result for train_breast_cancer_f8669_00000:\n", + " date: 2022-07-22_15-53-00\n", + " done: true\n", + " experiment_id: 07d10c5f31e74133b53272b7ccf9c528\n", + " hostname: Kais-MacBook-Pro.local\n", + " iterations_since_restore: 1\n", + " mean_accuracy: 0.9440559440559441\n", + " node_ip: 127.0.0.1\n", + " pid: 48852\n", + " time_since_restore: 0.031616926193237305\n", + " time_this_iter_s: 0.031616926193237305\n", + " time_total_s: 0.031616926193237305\n", + " timestamp: 1658501580\n", + " timesteps_since_restore: 0\n", + " training_iteration: 1\n", + " trial_id: f8669_00000\n", + " warmup_time: 0.0027849674224853516\n", + " \n", + "Result for train_breast_cancer_f8669_00009:\n", + " date: 2022-07-22_15-53-04\n", + " done: true\n", + " experiment_id: bc0d5dd2d079432b859faac8a18928f0\n", + " hostname: Kais-MacBook-Pro.local\n", + " iterations_since_restore: 1\n", + " mean_accuracy: 0.9300699300699301\n", + " node_ip: 127.0.0.1\n", + " pid: 48865\n", + " time_since_restore: 0.027495384216308594\n", + " time_this_iter_s: 0.027495384216308594\n", + " time_total_s: 0.027495384216308594\n", + " timestamp: 1658501584\n", + " timesteps_since_restore: 0\n", + " training_iteration: 1\n", + " trial_id: f8669_00009\n", + " warmup_time: 0.005235910415649414\n", + " \n", + "Result for train_breast_cancer_f8669_00001:\n", + " date: 2022-07-22_15-53-04\n", + " done: true\n", + " experiment_id: 4b10d350d4374a0d9e7d0c3b1d4e3203\n", + " hostname: Kais-MacBook-Pro.local\n", + " iterations_since_restore: 1\n", + " mean_accuracy: 0.958041958041958\n", + " node_ip: 127.0.0.1\n", + " pid: 48857\n", + " time_since_restore: 0.032858848571777344\n", + " time_this_iter_s: 0.032858848571777344\n", + " time_total_s: 0.032858848571777344\n", + " timestamp: 1658501584\n", + " timesteps_since_restore: 0\n", + " training_iteration: 1\n", + " trial_id: f8669_00001\n", + " warmup_time: 0.004731178283691406\n", + " \n", + "Result for train_breast_cancer_f8669_00008:\n", + " date: 2022-07-22_15-53-04\n", + " done: true\n", + " experiment_id: 91c25cbbeb6f409d93e1d6537cb8e1ee\n", + " hostname: Kais-MacBook-Pro.local\n", + " iterations_since_restore: 1\n", + " mean_accuracy: 0.951048951048951\n", + " node_ip: 127.0.0.1\n", + " pid: 48864\n", + " time_since_restore: 0.031099796295166016\n", + " time_this_iter_s: 0.031099796295166016\n", + " time_total_s: 0.031099796295166016\n", + " timestamp: 1658501584\n", + " timesteps_since_restore: 0\n", + " training_iteration: 1\n", + " trial_id: f8669_00008\n", + " warmup_time: 0.003270864486694336\n", + " \n", + "Result for train_breast_cancer_f8669_00005:\n", + " date: 2022-07-22_15-53-04\n", + " done: true\n", + " experiment_id: d225b0fb59e14da7adba952456ccf1d5\n", + " hostname: Kais-MacBook-Pro.local\n", + " iterations_since_restore: 1\n", + " mean_accuracy: 0.8741258741258742\n", + " node_ip: 127.0.0.1\n", + " pid: 48861\n", + " time_since_restore: 0.01936817169189453\n", + " time_this_iter_s: 0.01936817169189453\n", + " time_total_s: 0.01936817169189453\n", + " timestamp: 1658501584\n", + " timesteps_since_restore: 0\n", + " training_iteration: 1\n", + " trial_id: f8669_00005\n", + " warmup_time: 0.003901958465576172\n", + " \n", + "Result for train_breast_cancer_f8669_00004:\n", + " date: 2022-07-22_15-53-04\n", + " done: true\n", + " experiment_id: 322484af6ea5422f8aaf8ff6a91af4f7\n", + " hostname: Kais-MacBook-Pro.local\n", + " iterations_since_restore: 1\n", + " mean_accuracy: 0.958041958041958\n", + " node_ip: 127.0.0.1\n", + " pid: 48860\n", + " time_since_restore: 0.029961824417114258\n", + " time_this_iter_s: 0.029961824417114258\n", + " time_total_s: 0.029961824417114258\n", + " timestamp: 1658501584\n", + " timesteps_since_restore: 0\n", + " training_iteration: 1\n", + " trial_id: f8669_00004\n", + " warmup_time: 0.003547191619873047\n", + " \n", + "Result for train_breast_cancer_f8669_00002:\n", + " date: 2022-07-22_15-53-04\n", + " done: true\n", + " experiment_id: 3f588954160b42ce8ce200f68127ebcd\n", + " hostname: Kais-MacBook-Pro.local\n", + " iterations_since_restore: 1\n", + " mean_accuracy: 0.9440559440559441\n", + " node_ip: 127.0.0.1\n", + " pid: 48858\n", + " time_since_restore: 0.03193807601928711\n", + " time_this_iter_s: 0.03193807601928711\n", + " time_total_s: 0.03193807601928711\n", + " timestamp: 1658501584\n", + " timesteps_since_restore: 0\n", + " training_iteration: 1\n", + " trial_id: f8669_00002\n", + " warmup_time: 0.003523111343383789\n", + " \n", + "Result for train_breast_cancer_f8669_00003:\n", + " date: 2022-07-22_15-53-04\n", + " done: true\n", + " experiment_id: a39ea777ce2d4ebca51b3d7a4179dae5\n", + " hostname: Kais-MacBook-Pro.local\n", + " iterations_since_restore: 1\n", + " mean_accuracy: 0.9230769230769231\n", + " node_ip: 127.0.0.1\n", + " pid: 48859\n", + " time_since_restore: 0.028388023376464844\n", + " time_this_iter_s: 0.028388023376464844\n", + " time_total_s: 0.028388023376464844\n", + " timestamp: 1658501584\n", + " timesteps_since_restore: 0\n", + " training_iteration: 1\n", + " trial_id: f8669_00003\n", + " warmup_time: 0.0035560131072998047\n", + " \n", + "Result for train_breast_cancer_f8669_00006:\n", + " date: 2022-07-22_15-53-04\n", + " done: true\n", + " experiment_id: f97c6b9674854f8d89ec26ba58cc1618\n", + " hostname: Kais-MacBook-Pro.local\n", + " iterations_since_restore: 1\n", + " mean_accuracy: 0.9440559440559441\n", + " node_ip: 127.0.0.1\n", + " pid: 48862\n", + " time_since_restore: 0.026737213134765625\n", + " time_this_iter_s: 0.026737213134765625\n", + " time_total_s: 0.026737213134765625\n", + " timestamp: 1658501584\n", + " timesteps_since_restore: 0\n", + " training_iteration: 1\n", + " trial_id: f8669_00006\n", + " warmup_time: 0.003425121307373047\n", + " \n", + "Result for train_breast_cancer_f8669_00007:\n", + " date: 2022-07-22_15-53-04\n", + " done: true\n", + " experiment_id: ff172037065a4d55998ed72f51bdc5df\n", + " hostname: Kais-MacBook-Pro.local\n", + " iterations_since_restore: 1\n", + " mean_accuracy: 0.965034965034965\n", + " node_ip: 127.0.0.1\n", + " pid: 48863\n", + " time_since_restore: 0.027612924575805664\n", + " time_this_iter_s: 0.027612924575805664\n", + " time_total_s: 0.027612924575805664\n", + " timestamp: 1658501584\n", + " timesteps_since_restore: 0\n", + " training_iteration: 1\n", + " trial_id: f8669_00007\n", + " warmup_time: 0.0031311511993408203\n", + " \n" + ] + }, + { + "name": "stderr", + "output_type": "stream", + "text": [ + "2022-07-22 15:53:04,846\tINFO tune.py:738 -- Total run time: 8.99 seconds (7.74 seconds for the tuning loop).\n" + ] + } + ], + "source": [ + "import sklearn.datasets\n", + "import sklearn.metrics\n", + "\n", + "from ray import air, tune\n", + "from ray.air import session\n", + "\n", + "\n", + "def train_breast_cancer(config):\n", + " # Load dataset\n", + " data, labels = sklearn.datasets.load_breast_cancer(return_X_y=True)\n", + " # Split into train and test set\n", + " train_x, test_x, train_y, test_y = train_test_split(data, labels, test_size=0.25)\n", + " # Build input matrices for XGBoost\n", + " train_set = xgb.DMatrix(train_x, label=train_y)\n", + " test_set = xgb.DMatrix(test_x, label=test_y)\n", + " # Train the classifier\n", + " results = {}\n", + " xgb.train(\n", + " config,\n", + " train_set,\n", + " evals=[(test_set, \"eval\")],\n", + " evals_result=results,\n", + " verbose_eval=False,\n", + " )\n", + " # Return prediction accuracy\n", + " accuracy = 1.0 - results[\"eval\"][\"error\"][-1]\n", + " session.report({\"mean_accuracy\": accuracy, \"done\": True})\n", + "\n", + "\n", + "if __name__ == \"__main__\":\n", + " config = {\n", + " \"objective\": \"binary:logistic\",\n", + " \"eval_metric\": [\"logloss\", \"error\"],\n", + " \"max_depth\": tune.randint(1, 9),\n", + " \"min_child_weight\": tune.choice([1, 2, 3]),\n", + " \"subsample\": tune.uniform(0.5, 1.0),\n", + " \"eta\": tune.loguniform(1e-4, 1e-1),\n", + " }\n", + " tuner = tune.Tuner(\n", + " train_breast_cancer,\n", + " tune_config=tune.TuneConfig(\n", + " num_samples=10,\n", + " ),\n", + " param_space=config,\n", + " )\n", + " results = tuner.fit()\n" + ] + }, + { + "cell_type": "markdown", + "id": "4999e858", + "metadata": {}, + "source": [ + "As you can see, the changes in the actual training function are minimal. Instead of\n", + "returning the accuracy value, we report it back to Tune using `session.report()`.\n", + "Our `config` dictionary only changed slightly. Instead of passing hard-coded\n", + "parameters, we tell Tune to choose values from a range of valid options. There are\n", + "a number of options we have here, all of which are explained in\n", + "{ref}`the Tune docs `.\n", + "\n", + "For a brief explanation, this is what they do:\n", + "\n", + "- `tune.randint(min, max)` chooses a random integer value between *min* and *max*.\n", + " Note that *max* is exclusive, so it will not be sampled.\n", + "- `tune.choice([a, b, c])` chooses one of the items of the list at random. Each item\n", + " has the same chance to be sampled.\n", + "- `tune.uniform(min, max)` samples a floating point number between *min* and *max*.\n", + " Note that *max* is exclusive here, too.\n", + "- `tune.loguniform(min, max, base=10)` samples a floating point number between *min* and *max*,\n", + " but applies a logarithmic transformation to these boundaries first. Thus, this makes\n", + " it easy to sample values from different orders of magnitude.\n", + "\n", + "The `num_samples=10` option we pass to the `TuneConfig()` means that we sample 10 different\n", + "hyperparameter configurations from this search space.\n", + "\n", + "The output of our training run coud look like this:\n", + "\n", + "```{code-block} bash\n", + ":emphasize-lines: 14\n", + "\n", + " Number of trials: 10/10 (10 TERMINATED)\n", + " +---------------------------------+------------+-------+-------------+-------------+--------------------+-------------+----------+--------+------------------+\n", + " | Trial name | status | loc | eta | max_depth | min_child_weight | subsample | acc | iter | total time (s) |\n", + " |---------------------------------+------------+-------+-------------+-------------+--------------------+-------------+----------+--------+------------------|\n", + " | train_breast_cancer_b63aa_00000 | TERMINATED | | 0.000117625 | 2 | 2 | 0.616347 | 0.916084 | 1 | 0.0306492 |\n", + " | train_breast_cancer_b63aa_00001 | TERMINATED | | 0.0382954 | 8 | 2 | 0.581549 | 0.937063 | 1 | 0.0357082 |\n", + " | train_breast_cancer_b63aa_00002 | TERMINATED | | 0.000217926 | 1 | 3 | 0.528428 | 0.874126 | 1 | 0.0264609 |\n", + " | train_breast_cancer_b63aa_00003 | TERMINATED | | 0.000120929 | 8 | 1 | 0.634508 | 0.958042 | 1 | 0.036406 |\n", + " | train_breast_cancer_b63aa_00004 | TERMINATED | | 0.00839715 | 5 | 1 | 0.730624 | 0.958042 | 1 | 0.0389378 |\n", + " | train_breast_cancer_b63aa_00005 | TERMINATED | | 0.000732948 | 8 | 2 | 0.915863 | 0.958042 | 1 | 0.0382841 |\n", + " | train_breast_cancer_b63aa_00006 | TERMINATED | | 0.000856226 | 4 | 1 | 0.645209 | 0.916084 | 1 | 0.0357089 |\n", + " | train_breast_cancer_b63aa_00007 | TERMINATED | | 0.00769908 | 7 | 1 | 0.729443 | 0.909091 | 1 | 0.0390737 |\n", + " | train_breast_cancer_b63aa_00008 | TERMINATED | | 0.00186339 | 5 | 3 | 0.595744 | 0.944056 | 1 | 0.0343912 |\n", + " | train_breast_cancer_b63aa_00009 | TERMINATED | | 0.000950272 | 3 | 2 | 0.835504 | 0.965035 | 1 | 0.0348201 |\n", + " +---------------------------------+------------+-------+-------------+-------------+--------------------+-------------+----------+--------+------------------+\n", + "```\n", + "\n", + "The best configuration we found used `eta=0.000950272`, `max_depth=3`,\n", + "`min_child_weight=2`, `subsample=0.835504` and reached an accuracy of\n", + "`0.965035`.\n", + "\n", + "## Early stopping\n", + "\n", + "Currently, Tune samples 10 different hyperparameter configurations and trains a full\n", + "XGBoost on all of them. In our small example, training is very fast. However,\n", + "if training takes longer, a significant amount of computer resources is spent on trials\n", + "that will eventually show a bad performance, e.g. a low accuracy. It would be good\n", + "if we could identify these trials early and stop them, so we don't waste any resources.\n", + "\n", + "This is where Tune's *Schedulers* shine. A Tune `TrialScheduler` is responsible\n", + "for starting and stopping trials. Tune implements a number of different schedulers, each\n", + "described {ref}`in the Tune documentation `.\n", + "For our example, we will use the `AsyncHyperBandScheduler` or `ASHAScheduler`.\n", + "\n", + "The basic idea of this scheduler: We sample a number of hyperparameter configurations.\n", + "Each of these configurations is trained for a specific number of iterations.\n", + "After these iterations, only the best performing hyperparameters are retained. These\n", + "are selected according to some loss metric, usually an evaluation loss. This cycle is\n", + "repeated until we end up with the best configuration.\n", + "\n", + "The `ASHAScheduler` needs to know three things:\n", + "\n", + "1. Which metric should be used to identify badly performing trials?\n", + "2. Should this metric be maximized or minimized?\n", + "3. How many iterations does each trial train for?\n", + "\n", + "There are more parameters, which are explained in the\n", + "{ref}`documentation `.\n", + "\n", + "Lastly, we have to report the loss metric to Tune. We do this with a `Callback` that\n", + "XGBoost accepts and calls after each evaluation round. Ray Tune comes\n", + "with {ref}`two XGBoost callbacks `\n", + "we can use for this. The `TuneReportCallback` just reports the evaluation\n", + "metrics back to Tune. The `TuneReportCheckpointCallback` also saves\n", + "checkpoints after each evaluation round. We will just use the latter in this\n", + "example so that we can retrieve the saved model later.\n", + "\n", + "These parameters from the `eval_metrics` configuration setting are then automatically\n", + "reported to Tune via the callback. Here, the raw error will be reported, not the accuracy.\n", + "To display the best reached accuracy, we will inverse it later.\n", + "\n", + "We will also load the best checkpointed model so that we can use it for predictions.\n", + "The best model is selected with respect to the `metric` and `mode` parameters we\n", + "pass to the `TunerConfig()`." + ] + }, + { + "cell_type": "code", + "execution_count": 9, + "id": "d08b5b0a", + "metadata": {}, + "outputs": [ + { + "data": { + "text/html": [ + "== Status ==
Current time: 2022-07-22 16:56:01 (running for 00:00:10.38)
Memory usage on this node: 10.3/16.0 GiB
Using AsyncHyperBand: num_stopped=10\n", + "Bracket: Iter 8.000: -0.5107275277792991 | Iter 4.000: -0.5876629346317344 | Iter 2.000: -0.6544494184997531 | Iter 1.000: -0.6859214191253369
Resources requested: 0/16 CPUs, 0/0 GPUs, 0.0/4.57 GiB heap, 0.0/2.0 GiB objects
Current best trial: c28a3_00003 with eval-logloss=0.38665050018083796 and parameters={'objective': 'binary:logistic', 'eval_metric': ['logloss', 'error'], 'max_depth': 2, 'min_child_weight': 3, 'subsample': 0.782626252548841, 'eta': 0.06385952388342125}
Result logdir: /Users/kai/ray_results/train_breast_cancer_2022-07-22_16-55-50
Number of trials: 10/10 (10 TERMINATED)
\n", + "\n", + "\n", + "\n", + "\n", + "\n", + "\n", + "\n", + "\n", + "\n", + "\n", + "\n", + "\n", + "\n", + "\n", + "\n", + "
Trial name status loc eta max_depth min_child_weight subsample iter total time (s) eval-logloss eval-error
train_breast_cancer_c28a3_00000TERMINATED127.0.0.1:544160.0186954 2 2 0.516916 10 0.22218 0.571496 0.0629371
train_breast_cancer_c28a3_00001TERMINATED127.0.0.1:544400.0304404 8 2 0.745969 2 0.135674 0.650353 0.0629371
train_breast_cancer_c28a3_00002TERMINATED127.0.0.1:544410.0217157 8 3 0.764138 2 0.173076 0.658545 0.041958
train_breast_cancer_c28a3_00003TERMINATED127.0.0.1:544420.0638595 2 3 0.782626 10 0.281865 0.386651 0.041958
train_breast_cancer_c28a3_00004TERMINATED127.0.0.1:544430.00442794 7 2 0.792359 1 0.0270212 0.689577 0.0699301
train_breast_cancer_c28a3_00005TERMINATED127.0.0.1:544440.00222624 3 1 0.536331 1 0.0238512 0.691446 0.0839161
train_breast_cancer_c28a3_00006TERMINATED127.0.0.1:544450.000825129 1 1 0.82472 1 0.015312 0.692624 0.118881
train_breast_cancer_c28a3_00007TERMINATED127.0.0.1:544460.000770826 7 2 0.947268 1 0.0175898 0.692598 0.132867
train_breast_cancer_c28a3_00008TERMINATED127.0.0.1:544470.000429759 7 1 0.88524 1 0.0193739 0.692785 0.0559441
train_breast_cancer_c28a3_00009TERMINATED127.0.0.1:544480.0149863 2 1 0.722738 1 0.0165932 0.682266 0.111888


" + ], + "text/plain": [ + "" + ] + }, + "metadata": {}, + "output_type": "display_data" + }, + { + "name": "stdout", + "output_type": "stream", + "text": [ + "Result for train_breast_cancer_c28a3_00000:\n", + " date: 2022-07-22_16-55-55\n", + " done: false\n", + " eval-error: 0.08391608391608392\n", + " eval-logloss: 0.6790360066440556\n", + " experiment_id: 2a3189442db341519836a07fb2d65dd9\n", + " hostname: Kais-MacBook-Pro.local\n", + " iterations_since_restore: 1\n", + " node_ip: 127.0.0.1\n", + " pid: 54416\n", + " time_since_restore: 0.01624011993408203\n", + " time_this_iter_s: 0.01624011993408203\n", + " time_total_s: 0.01624011993408203\n", + " timestamp: 1658505355\n", + " timesteps_since_restore: 0\n", + " training_iteration: 1\n", + " trial_id: c28a3_00000\n", + " warmup_time: 0.0035409927368164062\n", + " \n", + "Result for train_breast_cancer_c28a3_00000:\n", + " date: 2022-07-22_16-55-56\n", + " done: true\n", + " eval-error: 0.06293706293706294\n", + " eval-logloss: 0.5714958122560194\n", + " experiment_id: 2a3189442db341519836a07fb2d65dd9\n", + " hostname: Kais-MacBook-Pro.local\n", + " iterations_since_restore: 10\n", + " node_ip: 127.0.0.1\n", + " pid: 54416\n", + " time_since_restore: 0.22218012809753418\n", + " time_this_iter_s: 0.007044076919555664\n", + " time_total_s: 0.22218012809753418\n", + " timestamp: 1658505356\n", + " timesteps_since_restore: 0\n", + " training_iteration: 10\n", + " trial_id: c28a3_00000\n", + " warmup_time: 0.0035409927368164062\n", + " \n", + "Result for train_breast_cancer_c28a3_00003:\n", + " date: 2022-07-22_16-56-01\n", + " done: false\n", + " eval-error: 0.08391608391608392\n", + " eval-logloss: 0.6472820101918041\n", + " experiment_id: 7ff6133237404b4ea4755b9f8cd114f2\n", + " hostname: Kais-MacBook-Pro.local\n", + " iterations_since_restore: 1\n", + " node_ip: 127.0.0.1\n", + " pid: 54442\n", + " time_since_restore: 0.023206233978271484\n", + " time_this_iter_s: 0.023206233978271484\n", + " time_total_s: 0.023206233978271484\n", + " timestamp: 1658505361\n", + " timesteps_since_restore: 0\n", + " training_iteration: 1\n", + " trial_id: c28a3_00003\n", + " warmup_time: 0.006722211837768555\n", + " \n", + "Result for train_breast_cancer_c28a3_00005:\n", + " date: 2022-07-22_16-56-01\n", + " done: true\n", + " eval-error: 0.08391608391608392\n", + " eval-logloss: 0.6914464114429234\n", + " experiment_id: 344762ab6d574b63a9374e19526d0510\n", + " hostname: Kais-MacBook-Pro.local\n", + " iterations_since_restore: 1\n", + " node_ip: 127.0.0.1\n", + " pid: 54444\n", + " time_since_restore: 0.02385115623474121\n", + " time_this_iter_s: 0.02385115623474121\n", + " time_total_s: 0.02385115623474121\n", + " timestamp: 1658505361\n", + " timesteps_since_restore: 0\n", + " training_iteration: 1\n", + " trial_id: c28a3_00005\n", + " warmup_time: 0.008936882019042969\n", + " \n", + "Result for train_breast_cancer_c28a3_00009:\n", + " date: 2022-07-22_16-56-01\n", + " done: true\n", + " eval-error: 0.11188811188811189\n", + " eval-logloss: 0.6822656309688008\n", + " experiment_id: 133901655fa64bf79f2dcc4e8e5e41b1\n", + " hostname: Kais-MacBook-Pro.local\n", + " iterations_since_restore: 1\n", + " node_ip: 127.0.0.1\n", + " pid: 54448\n", + " time_since_restore: 0.016593217849731445\n", + " time_this_iter_s: 0.016593217849731445\n", + " time_total_s: 0.016593217849731445\n", + " timestamp: 1658505361\n", + " timesteps_since_restore: 0\n", + " training_iteration: 1\n", + " trial_id: c28a3_00009\n", + " warmup_time: 0.004940032958984375\n", + " \n", + "Result for train_breast_cancer_c28a3_00007:\n", + " date: 2022-07-22_16-56-01\n", + " done: true\n", + " eval-error: 0.13286713286713286\n", + " eval-logloss: 0.6925980357023386\n", + " experiment_id: b4331027cbaf442ab905b2e51797dbbd\n", + " hostname: Kais-MacBook-Pro.local\n", + " iterations_since_restore: 1\n", + " node_ip: 127.0.0.1\n", + " pid: 54446\n", + " time_since_restore: 0.017589807510375977\n", + " time_this_iter_s: 0.017589807510375977\n", + " time_total_s: 0.017589807510375977\n", + " timestamp: 1658505361\n", + " timesteps_since_restore: 0\n", + " training_iteration: 1\n", + " trial_id: c28a3_00007\n", + " warmup_time: 0.003782033920288086\n", + " \n", + "Result for train_breast_cancer_c28a3_00006:\n", + " date: 2022-07-22_16-56-01\n", + " done: true\n", + " eval-error: 0.11888111888111888\n", + " eval-logloss: 0.6926244418104212\n", + " experiment_id: d3906de5943a4e05a4cc782382f67d24\n", + " hostname: Kais-MacBook-Pro.local\n", + " iterations_since_restore: 1\n", + " node_ip: 127.0.0.1\n", + " pid: 54445\n", + " time_since_restore: 0.015311956405639648\n", + " time_this_iter_s: 0.015311956405639648\n", + " time_total_s: 0.015311956405639648\n", + " timestamp: 1658505361\n", + " timesteps_since_restore: 0\n", + " training_iteration: 1\n", + " trial_id: c28a3_00006\n", + " warmup_time: 0.005506038665771484\n", + " \n", + "Result for train_breast_cancer_c28a3_00002:\n", + " date: 2022-07-22_16-56-01\n", + " done: false\n", + " eval-error: 0.04895104895104895\n", + " eval-logloss: 0.6752762102580571\n", + " experiment_id: a3645fc2d43145d88a1f5b7cc94df703\n", + " hostname: Kais-MacBook-Pro.local\n", + " iterations_since_restore: 1\n", + " node_ip: 127.0.0.1\n", + " pid: 54441\n", + " time_since_restore: 0.027367830276489258\n", + " time_this_iter_s: 0.027367830276489258\n", + " time_total_s: 0.027367830276489258\n", + " timestamp: 1658505361\n", + " timesteps_since_restore: 0\n", + " training_iteration: 1\n", + " trial_id: c28a3_00002\n", + " warmup_time: 0.0062830448150634766\n", + " \n", + "Result for train_breast_cancer_c28a3_00001:\n", + " date: 2022-07-22_16-56-01\n", + " done: false\n", + " eval-error: 0.07692307692307693\n", + " eval-logloss: 0.6698804135089154\n", + " experiment_id: 85766fe4d9fa482a91e396a8fd509a19\n", + " hostname: Kais-MacBook-Pro.local\n", + " iterations_since_restore: 1\n", + " node_ip: 127.0.0.1\n", + " pid: 54440\n", + " time_since_restore: 0.017169952392578125\n", + " time_this_iter_s: 0.017169952392578125\n", + " time_total_s: 0.017169952392578125\n", + " timestamp: 1658505361\n", + " timesteps_since_restore: 0\n", + " training_iteration: 1\n", + " trial_id: c28a3_00001\n", + " warmup_time: 0.006204843521118164\n", + " \n", + "Result for train_breast_cancer_c28a3_00008:\n", + " date: 2022-07-22_16-56-01\n", + " done: true\n", + " eval-error: 0.05594405594405594\n", + " eval-logloss: 0.692784742458717\n", + " experiment_id: 2c7d8bc38ad04536b1dec76819a2b3bf\n", + " hostname: Kais-MacBook-Pro.local\n", + " iterations_since_restore: 1\n", + " node_ip: 127.0.0.1\n", + " pid: 54447\n", + " time_since_restore: 0.01937389373779297\n", + " time_this_iter_s: 0.01937389373779297\n", + " time_total_s: 0.01937389373779297\n", + " timestamp: 1658505361\n", + " timesteps_since_restore: 0\n", + " training_iteration: 1\n", + " trial_id: c28a3_00008\n", + " warmup_time: 0.004342079162597656\n", + " \n", + "Result for train_breast_cancer_c28a3_00001:\n", + " date: 2022-07-22_16-56-01\n", + " done: true\n", + " eval-error: 0.06293706293706294\n", + " eval-logloss: 0.6503534216980834\n", + " experiment_id: 85766fe4d9fa482a91e396a8fd509a19\n", + " hostname: Kais-MacBook-Pro.local\n", + " iterations_since_restore: 2\n", + " node_ip: 127.0.0.1\n", + " pid: 54440\n", + " time_since_restore: 0.13567376136779785\n", + " time_this_iter_s: 0.11850380897521973\n", + " time_total_s: 0.13567376136779785\n", + " timestamp: 1658505361\n", + " timesteps_since_restore: 0\n", + " training_iteration: 2\n", + " trial_id: c28a3_00001\n", + " warmup_time: 0.006204843521118164\n", + " \n", + "Result for train_breast_cancer_c28a3_00004:\n", + " date: 2022-07-22_16-56-01\n", + " done: true\n", + " eval-error: 0.06993006993006994\n", + " eval-logloss: 0.689577207281873\n", + " experiment_id: ef4fdc645c444112985b4957ab8a84e9\n", + " hostname: Kais-MacBook-Pro.local\n", + " iterations_since_restore: 1\n", + " node_ip: 127.0.0.1\n", + " pid: 54443\n", + " time_since_restore: 0.027021169662475586\n", + " time_this_iter_s: 0.027021169662475586\n", + " time_total_s: 0.027021169662475586\n", + " timestamp: 1658505361\n", + " timesteps_since_restore: 0\n", + " training_iteration: 1\n", + " trial_id: c28a3_00004\n", + " warmup_time: 0.0063669681549072266\n", + " \n", + "Result for train_breast_cancer_c28a3_00002:\n", + " date: 2022-07-22_16-56-01\n", + " done: true\n", + " eval-error: 0.04195804195804196\n", + " eval-logloss: 0.658545415301423\n", + " experiment_id: a3645fc2d43145d88a1f5b7cc94df703\n", + " hostname: Kais-MacBook-Pro.local\n", + " iterations_since_restore: 2\n", + " node_ip: 127.0.0.1\n", + " pid: 54441\n", + " time_since_restore: 0.17307591438293457\n", + " time_this_iter_s: 0.1457080841064453\n", + " time_total_s: 0.17307591438293457\n", + " timestamp: 1658505361\n", + " timesteps_since_restore: 0\n", + " training_iteration: 2\n", + " trial_id: c28a3_00002\n", + " warmup_time: 0.0062830448150634766\n", + " \n", + "Result for train_breast_cancer_c28a3_00003:\n", + " date: 2022-07-22_16-56-01\n", + " done: true\n", + " eval-error: 0.04195804195804196\n", + " eval-logloss: 0.38665050018083796\n", + " experiment_id: 7ff6133237404b4ea4755b9f8cd114f2\n", + " hostname: Kais-MacBook-Pro.local\n", + " iterations_since_restore: 10\n", + " node_ip: 127.0.0.1\n", + " pid: 54442\n", + " time_since_restore: 0.28186488151550293\n", + " time_this_iter_s: 0.03063178062438965\n", + " time_total_s: 0.28186488151550293\n", + " timestamp: 1658505361\n", + " timesteps_since_restore: 0\n", + " training_iteration: 10\n", + " trial_id: c28a3_00003\n", + " warmup_time: 0.006722211837768555\n", + " \n" + ] + }, + { + "name": "stderr", + "output_type": "stream", + "text": [ + "2022-07-22 16:56:01,498\tINFO tune.py:738 -- Total run time: 10.53 seconds (10.37 seconds for the tuning loop).\n" + ] + }, + { + "name": "stdout", + "output_type": "stream", + "text": [ + "Best model parameters: {'objective': 'binary:logistic', 'eval_metric': ['logloss', 'error'], 'max_depth': 2, 'min_child_weight': 3, 'subsample': 0.782626252548841, 'eta': 0.06385952388342125}\n", + "Best model total accuracy: 0.9580\n" + ] + } + ], + "source": [ + "import sklearn.datasets\n", + "import sklearn.metrics\n", + "import os\n", + "from ray.tune.schedulers import ASHAScheduler\n", + "from sklearn.model_selection import train_test_split\n", + "import xgboost as xgb\n", + "\n", + "from ray import air, tune\n", + "from ray.air import session\n", + "from ray.tune.integration.xgboost import TuneReportCheckpointCallback\n", + "\n", + "\n", + "def train_breast_cancer(config: dict):\n", + " # This is a simple training function to be passed into Tune\n", + " # Load dataset\n", + " data, labels = sklearn.datasets.load_breast_cancer(return_X_y=True)\n", + " # Split into train and test set\n", + " train_x, test_x, train_y, test_y = train_test_split(data, labels, test_size=0.25)\n", + " # Build input matrices for XGBoost\n", + " train_set = xgb.DMatrix(train_x, label=train_y)\n", + " test_set = xgb.DMatrix(test_x, label=test_y)\n", + " # Train the classifier, using the Tune callback\n", + " xgb.train(\n", + " config,\n", + " train_set,\n", + " evals=[(test_set, \"eval\")],\n", + " verbose_eval=False,\n", + " callbacks=[TuneReportCheckpointCallback(filename=\"model.xgb\")],\n", + " )\n", + "\n", + "\n", + "def get_best_model_checkpoint(results):\n", + " best_bst = xgb.Booster()\n", + " best_result = results.get_best_result()\n", + "\n", + " with best_result.checkpoint.as_directory() as best_checkpoint_dir:\n", + " best_bst.load_model(os.path.join(best_checkpoint_dir, \"model.xgb\"))\n", + " accuracy = 1.0 - best_result.metrics[\"eval-error\"]\n", + " print(f\"Best model parameters: {best_result.config}\")\n", + " print(f\"Best model total accuracy: {accuracy:.4f}\")\n", + " return best_bst\n", + "\n", + "\n", + "def tune_xgboost(smoke_test=False):\n", + " search_space = {\n", + " # You can mix constants with search space objects.\n", + " \"objective\": \"binary:logistic\",\n", + " \"eval_metric\": [\"logloss\", \"error\"],\n", + " \"max_depth\": tune.randint(1, 9),\n", + " \"min_child_weight\": tune.choice([1, 2, 3]),\n", + " \"subsample\": tune.uniform(0.5, 1.0),\n", + " \"eta\": tune.loguniform(1e-4, 1e-1),\n", + " }\n", + " # This will enable aggressive early stopping of bad trials.\n", + " scheduler = ASHAScheduler(\n", + " max_t=10, grace_period=1, reduction_factor=2 # 10 training iterations\n", + " )\n", + "\n", + " tuner = tune.Tuner(\n", + " train_breast_cancer,\n", + " tune_config=tune.TuneConfig(\n", + " metric=\"eval-logloss\",\n", + " mode=\"min\",\n", + " scheduler=scheduler,\n", + " num_samples=1 if smoke_test else 10,\n", + " ),\n", + " param_space=search_space,\n", + " )\n", + " results = tuner.fit()\n", + "\n", + " return results\n", + "\n", + "\n", + "if __name__ == \"__main__\":\n", + " import argparse\n", + "\n", + " parser = argparse.ArgumentParser()\n", + " parser.add_argument(\n", + " \"--smoke-test\", action=\"store_true\", help=\"Finish quickly for testing\"\n", + " )\n", + " args, _ = parser.parse_known_args()\n", + "\n", + " results = tune_xgboost(smoke_test=args.smoke_test)\n", + "\n", + " # Load the best model checkpoint.\n", + " best_bst = get_best_model_checkpoint(results)\n", + "\n", + " # You could now do further predictions with\n", + " # best_bst.predict(...)\n" + ] + }, + { + "cell_type": "markdown", + "id": "20732fe4", + "metadata": {}, + "source": [ + "The output of our run could look like this:\n", + "\n", + "```{code-block} bash\n", + ":emphasize-lines: 7\n", + "\n", + " Number of trials: 10/10 (10 TERMINATED)\n", + " +---------------------------------+------------+-------+-------------+-------------+--------------------+-------------+--------+------------------+----------------+--------------+\n", + " | Trial name | status | loc | eta | max_depth | min_child_weight | subsample | iter | total time (s) | eval-logloss | eval-error |\n", + " |---------------------------------+------------+-------+-------------+-------------+--------------------+-------------+--------+------------------+----------------+--------------|\n", + " | train_breast_cancer_ba275_00000 | TERMINATED | | 0.00205087 | 2 | 1 | 0.898391 | 10 | 0.380619 | 0.678039 | 0.090909 |\n", + " | train_breast_cancer_ba275_00001 | TERMINATED | | 0.000183834 | 4 | 3 | 0.924939 | 1 | 0.0228798 | 0.693009 | 0.111888 |\n", + " | train_breast_cancer_ba275_00002 | TERMINATED | | 0.0242721 | 7 | 2 | 0.501551 | 10 | 0.376154 | 0.54472 | 0.06993 |\n", + " | train_breast_cancer_ba275_00003 | TERMINATED | | 0.000449692 | 5 | 3 | 0.890212 | 1 | 0.0234981 | 0.692811 | 0.090909 |\n", + " | train_breast_cancer_ba275_00004 | TERMINATED | | 0.000376393 | 7 | 2 | 0.883609 | 1 | 0.0231569 | 0.692847 | 0.062937 |\n", + " | train_breast_cancer_ba275_00005 | TERMINATED | | 0.00231942 | 3 | 3 | 0.877464 | 2 | 0.104867 | 0.689541 | 0.083916 |\n", + " | train_breast_cancer_ba275_00006 | TERMINATED | | 0.000542326 | 1 | 2 | 0.578584 | 1 | 0.0213971 | 0.692765 | 0.083916 |\n", + " | train_breast_cancer_ba275_00007 | TERMINATED | | 0.0016801 | 1 | 2 | 0.975302 | 1 | 0.02226 | 0.691999 | 0.083916 |\n", + " | train_breast_cancer_ba275_00008 | TERMINATED | | 0.000595756 | 8 | 3 | 0.58429 | 1 | 0.0221152 | 0.692657 | 0.06993 |\n", + " | train_breast_cancer_ba275_00009 | TERMINATED | | 0.000357845 | 8 | 1 | 0.637776 | 1 | 0.022635 | 0.692859 | 0.090909 |\n", + " +---------------------------------+------------+-------+-------------+-------------+--------------------+-------------+--------+------------------+----------------+--------------+\n", + "\n", + "\n", + " Best model parameters: {'objective': 'binary:logistic', 'eval_metric': ['logloss', 'error'], 'max_depth': 7, 'min_child_weight': 2, 'subsample': 0.5015513240240503, 'eta': 0.024272050872920895}\n", + " Best model total accuracy: 0.9301\n", + "```\n", + "\n", + "As you can see, most trials have been stopped only after a few iterations. Only the\n", + "two most promising trials were run for the full 10 iterations.\n", + "\n", + "You can also ensure that all available resources are being used as the scheduler\n", + "terminates trials, freeing them up. This can be done through the\n", + "`ResourceChangingScheduler`. An example of this can be found here:\n", + "{doc}`/tune/examples/includes/xgboost_dynamic_resources_example`.\n", + "\n", + "## Using fractional GPUs\n", + "\n", + "You can often accelerate your training by using GPUs in addition to CPUs. However,\n", + "you usually don't have as many GPUs as you have trials to run. For instance, if you\n", + "run 10 Tune trials in parallel, you usually don't have access to 10 separate GPUs.\n", + "\n", + "Tune supports *fractional GPUs*. This means that each task is assigned a fraction\n", + "of the GPU memory for training. For 10 tasks, this could look like this:" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "7d1b20a3", + "metadata": {}, + "outputs": [], + "source": [ + "config = {\n", + " \"objective\": \"binary:logistic\",\n", + " \"eval_metric\": [\"logloss\", \"error\"],\n", + " \"tree_method\": \"gpu_hist\",\n", + " \"max_depth\": tune.randint(1, 9),\n", + " \"min_child_weight\": tune.choice([1, 2, 3]),\n", + " \"subsample\": tune.uniform(0.5, 1.0),\n", + " \"eta\": tune.loguniform(1e-4, 1e-1),\n", + "}\n", + "\n", + "tuner = tune.Tuner(\n", + " tune.with_resources(train_breast_cancer, resources={\"cpu\": 1, \"gpu\": 0.1}),\n", + " tune_config=tune.TuneConfig(\n", + " num_samples=10,\n", + " ),\n", + " param_space=config,\n", + ")\n", + "results = tuner.fit()\n" + ] + }, + { + "cell_type": "markdown", + "id": "ee131861", + "metadata": {}, + "source": [ + "Each task thus works with 10% of the available GPU memory. You also have to tell\n", + "XGBoost to use the `gpu_hist` tree method, so it knows it should use the GPU.\n", + "\n", + "## Conclusion\n", + "\n", + "You should now have a basic understanding on how to train XGBoost models and on how\n", + "to tune the hyperparameters to yield the best results. In our simple example,\n", + "Tuning the parameters didn't make a huge difference for the accuracy.\n", + "But in larger applications, intelligent hyperparameter tuning can make the\n", + "difference between a model that doesn't seem to learn at all, and a model\n", + "that outperforms all the other ones.\n", + "\n", + "## More XGBoost Examples\n", + "\n", + "- {doc}`/tune/examples/includes/xgboost_dynamic_resources_example`:\n", + " Trains a basic XGBoost model with Tune with the class-based API and a ResourceChangingScheduler, ensuring all resources are being used at all time.\n", + "\n", + "## Learn More\n", + "\n", + "- [XGBoost Hyperparameter Tuning - A Visual Guide](https://kevinvecmanis.io/machine%20learning/hyperparameter%20tuning/dataviz/python/2019/05/11/XGBoost-Tuning-Visual-Guide.html)\n", + "- [Notes on XGBoost Parameter Tuning](https://xgboost.readthedocs.io/en/latest/tutorials/param_tuning.html)\n", + "- [Doing XGBoost Hyperparameter Tuning the smart way](https://towardsdatascience.com/doing-xgboost-hyper-parameter-tuning-the-smart-way-part-1-of-2-f6d255a45dde)" + ] + } + ], + "metadata": { + "kernelspec": { + "display_name": "ray_dev_py38", + "language": "python", + "name": "python3" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.8.13 | packaged by conda-forge | (default, Mar 25 2022, 06:05:16) \n[Clang 12.0.1 ]" + }, + "orphan": true, + "vscode": { + "interpreter": { + "hash": "265d195fda5292fe8f69c6e37c435a5634a1ed3b6799724e66a975f68fa21517" + } + } }, - { - "data": { - "text/html": [ - "== Status ==
Current time: 2022-07-22 15:53:04 (running for 00:00:07.77)
Memory usage on this node: 10.5/16.0 GiB
Using FIFO scheduling algorithm.
Resources requested: 0/16 CPUs, 0/0 GPUs, 0.0/4.57 GiB heap, 0.0/2.0 GiB objects
Result logdir: /Users/kai/ray_results/train_breast_cancer_2022-07-22_15-52-48
Number of trials: 10/10 (10 TERMINATED)
\n", - "\n", - "\n", - "\n", - "\n", - "\n", - "\n", - "\n", - "\n", - "\n", - "\n", - "\n", - "\n", - "\n", - "\n", - "\n", - "
Trial name status loc eta max_depth min_child_weight subsample acc iter total time (s)
train_breast_cancer_f8669_00000TERMINATED127.0.0.1:488520.0069356 5 3 0.8235040.944056 1 0.0316169
train_breast_cancer_f8669_00001TERMINATED127.0.0.1:488570.00145619 6 3 0.8329470.958042 1 0.0328588
train_breast_cancer_f8669_00002TERMINATED127.0.0.1:488580.00108208 7 3 0.9873190.944056 1 0.0319381
train_breast_cancer_f8669_00003TERMINATED127.0.0.1:488590.00530429 8 2 0.6156910.923077 1 0.028388
train_breast_cancer_f8669_00004TERMINATED127.0.0.1:488600.000721843 8 1 0.6509730.958042 1 0.0299618
train_breast_cancer_f8669_00005TERMINATED127.0.0.1:488610.0074509 1 1 0.7383410.874126 1 0.0193682
train_breast_cancer_f8669_00006TERMINATED127.0.0.1:488620.0879882 8 2 0.6715760.944056 1 0.0267372
train_breast_cancer_f8669_00007TERMINATED127.0.0.1:488630.0765404 7 2 0.7081570.965035 1 0.0276129
train_breast_cancer_f8669_00008TERMINATED127.0.0.1:488640.000627649 6 1 0.81121 0.951049 1 0.0310998
train_breast_cancer_f8669_00009TERMINATED127.0.0.1:488650.000383711 2 3 0.9905790.93007 1 0.0274954


" - ], - "text/plain": [ - "" - ] - }, - "metadata": {}, - "output_type": "display_data" - }, - { - "name": "stderr", - "output_type": "stream", - "text": [ - "2022-07-22 15:52:57,385\tINFO plugin_schema_manager.py:52 -- Loading the default runtime env schemas: ['/Users/kai/coding/ray/python/ray/_private/runtime_env/../../runtime_env/schemas/working_dir_schema.json', '/Users/kai/coding/ray/python/ray/_private/runtime_env/../../runtime_env/schemas/pip_schema.json'].\n" - ] - }, - { - "name": "stdout", - "output_type": "stream", - "text": [ - "Result for train_breast_cancer_f8669_00000:\n", - " date: 2022-07-22_15-53-00\n", - " done: true\n", - " experiment_id: 07d10c5f31e74133b53272b7ccf9c528\n", - " hostname: Kais-MacBook-Pro.local\n", - " iterations_since_restore: 1\n", - " mean_accuracy: 0.9440559440559441\n", - " node_ip: 127.0.0.1\n", - " pid: 48852\n", - " time_since_restore: 0.031616926193237305\n", - " time_this_iter_s: 0.031616926193237305\n", - " time_total_s: 0.031616926193237305\n", - " timestamp: 1658501580\n", - " timesteps_since_restore: 0\n", - " training_iteration: 1\n", - " trial_id: f8669_00000\n", - " warmup_time: 0.0027849674224853516\n", - " \n", - "Result for train_breast_cancer_f8669_00009:\n", - " date: 2022-07-22_15-53-04\n", - " done: true\n", - " experiment_id: bc0d5dd2d079432b859faac8a18928f0\n", - " hostname: Kais-MacBook-Pro.local\n", - " iterations_since_restore: 1\n", - " mean_accuracy: 0.9300699300699301\n", - " node_ip: 127.0.0.1\n", - " pid: 48865\n", - " time_since_restore: 0.027495384216308594\n", - " time_this_iter_s: 0.027495384216308594\n", - " time_total_s: 0.027495384216308594\n", - " timestamp: 1658501584\n", - " timesteps_since_restore: 0\n", - " training_iteration: 1\n", - " trial_id: f8669_00009\n", - " warmup_time: 0.005235910415649414\n", - " \n", - "Result for train_breast_cancer_f8669_00001:\n", - " date: 2022-07-22_15-53-04\n", - " done: true\n", - " experiment_id: 4b10d350d4374a0d9e7d0c3b1d4e3203\n", - " hostname: Kais-MacBook-Pro.local\n", - " iterations_since_restore: 1\n", - " mean_accuracy: 0.958041958041958\n", - " node_ip: 127.0.0.1\n", - " pid: 48857\n", - " time_since_restore: 0.032858848571777344\n", - " time_this_iter_s: 0.032858848571777344\n", - " time_total_s: 0.032858848571777344\n", - " timestamp: 1658501584\n", - " timesteps_since_restore: 0\n", - " training_iteration: 1\n", - " trial_id: f8669_00001\n", - " warmup_time: 0.004731178283691406\n", - " \n", - "Result for train_breast_cancer_f8669_00008:\n", - " date: 2022-07-22_15-53-04\n", - " done: true\n", - " experiment_id: 91c25cbbeb6f409d93e1d6537cb8e1ee\n", - " hostname: Kais-MacBook-Pro.local\n", - " iterations_since_restore: 1\n", - " mean_accuracy: 0.951048951048951\n", - " node_ip: 127.0.0.1\n", - " pid: 48864\n", - " time_since_restore: 0.031099796295166016\n", - " time_this_iter_s: 0.031099796295166016\n", - " time_total_s: 0.031099796295166016\n", - " timestamp: 1658501584\n", - " timesteps_since_restore: 0\n", - " training_iteration: 1\n", - " trial_id: f8669_00008\n", - " warmup_time: 0.003270864486694336\n", - " \n", - "Result for train_breast_cancer_f8669_00005:\n", - " date: 2022-07-22_15-53-04\n", - " done: true\n", - " experiment_id: d225b0fb59e14da7adba952456ccf1d5\n", - " hostname: Kais-MacBook-Pro.local\n", - " iterations_since_restore: 1\n", - " mean_accuracy: 0.8741258741258742\n", - " node_ip: 127.0.0.1\n", - " pid: 48861\n", - " time_since_restore: 0.01936817169189453\n", - " time_this_iter_s: 0.01936817169189453\n", - " time_total_s: 0.01936817169189453\n", - " timestamp: 1658501584\n", - " timesteps_since_restore: 0\n", - " training_iteration: 1\n", - " trial_id: f8669_00005\n", - " warmup_time: 0.003901958465576172\n", - " \n", - "Result for train_breast_cancer_f8669_00004:\n", - " date: 2022-07-22_15-53-04\n", - " done: true\n", - " experiment_id: 322484af6ea5422f8aaf8ff6a91af4f7\n", - " hostname: Kais-MacBook-Pro.local\n", - " iterations_since_restore: 1\n", - " mean_accuracy: 0.958041958041958\n", - " node_ip: 127.0.0.1\n", - " pid: 48860\n", - " time_since_restore: 0.029961824417114258\n", - " time_this_iter_s: 0.029961824417114258\n", - " time_total_s: 0.029961824417114258\n", - " timestamp: 1658501584\n", - " timesteps_since_restore: 0\n", - " training_iteration: 1\n", - " trial_id: f8669_00004\n", - " warmup_time: 0.003547191619873047\n", - " \n", - "Result for train_breast_cancer_f8669_00002:\n", - " date: 2022-07-22_15-53-04\n", - " done: true\n", - " experiment_id: 3f588954160b42ce8ce200f68127ebcd\n", - " hostname: Kais-MacBook-Pro.local\n", - " iterations_since_restore: 1\n", - " mean_accuracy: 0.9440559440559441\n", - " node_ip: 127.0.0.1\n", - " pid: 48858\n", - " time_since_restore: 0.03193807601928711\n", - " time_this_iter_s: 0.03193807601928711\n", - " time_total_s: 0.03193807601928711\n", - " timestamp: 1658501584\n", - " timesteps_since_restore: 0\n", - " training_iteration: 1\n", - " trial_id: f8669_00002\n", - " warmup_time: 0.003523111343383789\n", - " \n", - "Result for train_breast_cancer_f8669_00003:\n", - " date: 2022-07-22_15-53-04\n", - " done: true\n", - " experiment_id: a39ea777ce2d4ebca51b3d7a4179dae5\n", - " hostname: Kais-MacBook-Pro.local\n", - " iterations_since_restore: 1\n", - " mean_accuracy: 0.9230769230769231\n", - " node_ip: 127.0.0.1\n", - " pid: 48859\n", - " time_since_restore: 0.028388023376464844\n", - " time_this_iter_s: 0.028388023376464844\n", - " time_total_s: 0.028388023376464844\n", - " timestamp: 1658501584\n", - " timesteps_since_restore: 0\n", - " training_iteration: 1\n", - " trial_id: f8669_00003\n", - " warmup_time: 0.0035560131072998047\n", - " \n", - "Result for train_breast_cancer_f8669_00006:\n", - " date: 2022-07-22_15-53-04\n", - " done: true\n", - " experiment_id: f97c6b9674854f8d89ec26ba58cc1618\n", - " hostname: Kais-MacBook-Pro.local\n", - " iterations_since_restore: 1\n", - " mean_accuracy: 0.9440559440559441\n", - " node_ip: 127.0.0.1\n", - " pid: 48862\n", - " time_since_restore: 0.026737213134765625\n", - " time_this_iter_s: 0.026737213134765625\n", - " time_total_s: 0.026737213134765625\n", - " timestamp: 1658501584\n", - " timesteps_since_restore: 0\n", - " training_iteration: 1\n", - " trial_id: f8669_00006\n", - " warmup_time: 0.003425121307373047\n", - " \n", - "Result for train_breast_cancer_f8669_00007:\n", - " date: 2022-07-22_15-53-04\n", - " done: true\n", - " experiment_id: ff172037065a4d55998ed72f51bdc5df\n", - " hostname: Kais-MacBook-Pro.local\n", - " iterations_since_restore: 1\n", - " mean_accuracy: 0.965034965034965\n", - " node_ip: 127.0.0.1\n", - " pid: 48863\n", - " time_since_restore: 0.027612924575805664\n", - " time_this_iter_s: 0.027612924575805664\n", - " time_total_s: 0.027612924575805664\n", - " timestamp: 1658501584\n", - " timesteps_since_restore: 0\n", - " training_iteration: 1\n", - " trial_id: f8669_00007\n", - " warmup_time: 0.0031311511993408203\n", - " \n" - ] - }, - { - "name": "stderr", - "output_type": "stream", - "text": [ - "2022-07-22 15:53:04,846\tINFO tune.py:738 -- Total run time: 8.99 seconds (7.74 seconds for the tuning loop).\n" - ] - } - ], - "source": [ - "import sklearn.datasets\n", - "import sklearn.metrics\n", - "\n", - "from ray import air, tune\n", - "from ray.air import session\n", - "\n", - "\n", - "def train_breast_cancer(config):\n", - " # Load dataset\n", - " data, labels = sklearn.datasets.load_breast_cancer(return_X_y=True)\n", - " # Split into train and test set\n", - " train_x, test_x, train_y, test_y = train_test_split(data, labels, test_size=0.25)\n", - " # Build input matrices for XGBoost\n", - " train_set = xgb.DMatrix(train_x, label=train_y)\n", - " test_set = xgb.DMatrix(test_x, label=test_y)\n", - " # Train the classifier\n", - " results = {}\n", - " xgb.train(\n", - " config,\n", - " train_set,\n", - " evals=[(test_set, \"eval\")],\n", - " evals_result=results,\n", - " verbose_eval=False,\n", - " )\n", - " # Return prediction accuracy\n", - " accuracy = 1.0 - results[\"eval\"][\"error\"][-1]\n", - " session.report({\"mean_accuracy\": accuracy, \"done\": True})\n", - "\n", - "\n", - "if __name__ == \"__main__\":\n", - " config = {\n", - " \"objective\": \"binary:logistic\",\n", - " \"eval_metric\": [\"logloss\", \"error\"],\n", - " \"max_depth\": tune.randint(1, 9),\n", - " \"min_child_weight\": tune.choice([1, 2, 3]),\n", - " \"subsample\": tune.uniform(0.5, 1.0),\n", - " \"eta\": tune.loguniform(1e-4, 1e-1),\n", - " }\n", - " tuner = tune.Tuner(\n", - " train_breast_cancer,\n", - " tune_config=tune.TuneConfig(\n", - " num_samples=10,\n", - " ),\n", - " param_space=config,\n", - " )\n", - " results = tuner.fit()\n" - ] - }, - { - "cell_type": "markdown", - "id": "4999e858", - "metadata": {}, - "source": [ - "As you can see, the changes in the actual training function are minimal. Instead of\n", - "returning the accuracy value, we report it back to Tune using `session.report()`.\n", - "Our `config` dictionary only changed slightly. Instead of passing hard-coded\n", - "parameters, we tell Tune to choose values from a range of valid options. There are\n", - "a number of options we have here, all of which are explained in\n", - "{ref}`the Tune docs `.\n", - "\n", - "For a brief explanation, this is what they do:\n", - "\n", - "- `tune.randint(min, max)` chooses a random integer value between *min* and *max*.\n", - " Note that *max* is exclusive, so it will not be sampled.\n", - "- `tune.choice([a, b, c])` chooses one of the items of the list at random. Each item\n", - " has the same chance to be sampled.\n", - "- `tune.uniform(min, max)` samples a floating point number between *min* and *max*.\n", - " Note that *max* is exclusive here, too.\n", - "- `tune.loguniform(min, max, base=10)` samples a floating point number between *min* and *max*,\n", - " but applies a logarithmic transformation to these boundaries first. Thus, this makes\n", - " it easy to sample values from different orders of magnitude.\n", - "\n", - "The `num_samples=10` option we pass to the `TuneConfig()` means that we sample 10 different\n", - "hyperparameter configurations from this search space.\n", - "\n", - "The output of our training run coud look like this:\n", - "\n", - "```{code-block} bash\n", - ":emphasize-lines: 14\n", - "\n", - " Number of trials: 10/10 (10 TERMINATED)\n", - " +---------------------------------+------------+-------+-------------+-------------+--------------------+-------------+----------+--------+------------------+\n", - " | Trial name | status | loc | eta | max_depth | min_child_weight | subsample | acc | iter | total time (s) |\n", - " |---------------------------------+------------+-------+-------------+-------------+--------------------+-------------+----------+--------+------------------|\n", - " | train_breast_cancer_b63aa_00000 | TERMINATED | | 0.000117625 | 2 | 2 | 0.616347 | 0.916084 | 1 | 0.0306492 |\n", - " | train_breast_cancer_b63aa_00001 | TERMINATED | | 0.0382954 | 8 | 2 | 0.581549 | 0.937063 | 1 | 0.0357082 |\n", - " | train_breast_cancer_b63aa_00002 | TERMINATED | | 0.000217926 | 1 | 3 | 0.528428 | 0.874126 | 1 | 0.0264609 |\n", - " | train_breast_cancer_b63aa_00003 | TERMINATED | | 0.000120929 | 8 | 1 | 0.634508 | 0.958042 | 1 | 0.036406 |\n", - " | train_breast_cancer_b63aa_00004 | TERMINATED | | 0.00839715 | 5 | 1 | 0.730624 | 0.958042 | 1 | 0.0389378 |\n", - " | train_breast_cancer_b63aa_00005 | TERMINATED | | 0.000732948 | 8 | 2 | 0.915863 | 0.958042 | 1 | 0.0382841 |\n", - " | train_breast_cancer_b63aa_00006 | TERMINATED | | 0.000856226 | 4 | 1 | 0.645209 | 0.916084 | 1 | 0.0357089 |\n", - " | train_breast_cancer_b63aa_00007 | TERMINATED | | 0.00769908 | 7 | 1 | 0.729443 | 0.909091 | 1 | 0.0390737 |\n", - " | train_breast_cancer_b63aa_00008 | TERMINATED | | 0.00186339 | 5 | 3 | 0.595744 | 0.944056 | 1 | 0.0343912 |\n", - " | train_breast_cancer_b63aa_00009 | TERMINATED | | 0.000950272 | 3 | 2 | 0.835504 | 0.965035 | 1 | 0.0348201 |\n", - " +---------------------------------+------------+-------+-------------+-------------+--------------------+-------------+----------+--------+------------------+\n", - "```\n", - "\n", - "The best configuration we found used `eta=0.000950272`, `max_depth=3`,\n", - "`min_child_weight=2`, `subsample=0.835504` and reached an accuracy of\n", - "`0.965035`.\n", - "\n", - "## Early stopping\n", - "\n", - "Currently, Tune samples 10 different hyperparameter configurations and trains a full\n", - "XGBoost on all of them. In our small example, training is very fast. However,\n", - "if training takes longer, a significant amount of computer resources is spent on trials\n", - "that will eventually show a bad performance, e.g. a low accuracy. It would be good\n", - "if we could identify these trials early and stop them, so we don't waste any resources.\n", - "\n", - "This is where Tune's *Schedulers* shine. A Tune `TrialScheduler` is responsible\n", - "for starting and stopping trials. Tune implements a number of different schedulers, each\n", - "described {ref}`in the Tune documentation `.\n", - "For our example, we will use the `AsyncHyperBandScheduler` or `ASHAScheduler`.\n", - "\n", - "The basic idea of this scheduler: We sample a number of hyperparameter configurations.\n", - "Each of these configurations is trained for a specific number of iterations.\n", - "After these iterations, only the best performing hyperparameters are retained. These\n", - "are selected according to some loss metric, usually an evaluation loss. This cycle is\n", - "repeated until we end up with the best configuration.\n", - "\n", - "The `ASHAScheduler` needs to know three things:\n", - "\n", - "1. Which metric should be used to identify badly performing trials?\n", - "2. Should this metric be maximized or minimized?\n", - "3. How many iterations does each trial train for?\n", - "\n", - "There are more parameters, which are explained in the\n", - "{ref}`documentation `.\n", - "\n", - "Lastly, we have to report the loss metric to Tune. We do this with a `Callback` that\n", - "XGBoost accepts and calls after each evaluation round. Ray Tune comes\n", - "with {ref}`two XGBoost callbacks `\n", - "we can use for this. The `TuneReportCallback` just reports the evaluation\n", - "metrics back to Tune. The `TuneReportCheckpointCallback` also saves\n", - "checkpoints after each evaluation round. We will just use the latter in this\n", - "example so that we can retrieve the saved model later.\n", - "\n", - "These parameters from the `eval_metrics` configuration setting are then automatically\n", - "reported to Tune via the callback. Here, the raw error will be reported, not the accuracy.\n", - "To display the best reached accuracy, we will inverse it later.\n", - "\n", - "We will also load the best checkpointed model so that we can use it for predictions.\n", - "The best model is selected with respect to the `metric` and `mode` parameters we\n", - "pass to the `TunerConfig()`." - ] - }, - { - "cell_type": "code", - "execution_count": 9, - "id": "d08b5b0a", - "metadata": {}, - "outputs": [ - { - "data": { - "text/html": [ - "== Status ==
Current time: 2022-07-22 16:56:01 (running for 00:00:10.38)
Memory usage on this node: 10.3/16.0 GiB
Using AsyncHyperBand: num_stopped=10\n", - "Bracket: Iter 8.000: -0.5107275277792991 | Iter 4.000: -0.5876629346317344 | Iter 2.000: -0.6544494184997531 | Iter 1.000: -0.6859214191253369
Resources requested: 0/16 CPUs, 0/0 GPUs, 0.0/4.57 GiB heap, 0.0/2.0 GiB objects
Current best trial: c28a3_00003 with eval-logloss=0.38665050018083796 and parameters={'objective': 'binary:logistic', 'eval_metric': ['logloss', 'error'], 'max_depth': 2, 'min_child_weight': 3, 'subsample': 0.782626252548841, 'eta': 0.06385952388342125}
Result logdir: /Users/kai/ray_results/train_breast_cancer_2022-07-22_16-55-50
Number of trials: 10/10 (10 TERMINATED)
\n", - "\n", - "\n", - "\n", - "\n", - "\n", - "\n", - "\n", - "\n", - "\n", - "\n", - "\n", - "\n", - "\n", - "\n", - "\n", - "
Trial name status loc eta max_depth min_child_weight subsample iter total time (s) eval-logloss eval-error
train_breast_cancer_c28a3_00000TERMINATED127.0.0.1:544160.0186954 2 2 0.516916 10 0.22218 0.571496 0.0629371
train_breast_cancer_c28a3_00001TERMINATED127.0.0.1:544400.0304404 8 2 0.745969 2 0.135674 0.650353 0.0629371
train_breast_cancer_c28a3_00002TERMINATED127.0.0.1:544410.0217157 8 3 0.764138 2 0.173076 0.658545 0.041958
train_breast_cancer_c28a3_00003TERMINATED127.0.0.1:544420.0638595 2 3 0.782626 10 0.281865 0.386651 0.041958
train_breast_cancer_c28a3_00004TERMINATED127.0.0.1:544430.00442794 7 2 0.792359 1 0.0270212 0.689577 0.0699301
train_breast_cancer_c28a3_00005TERMINATED127.0.0.1:544440.00222624 3 1 0.536331 1 0.0238512 0.691446 0.0839161
train_breast_cancer_c28a3_00006TERMINATED127.0.0.1:544450.000825129 1 1 0.82472 1 0.015312 0.692624 0.118881
train_breast_cancer_c28a3_00007TERMINATED127.0.0.1:544460.000770826 7 2 0.947268 1 0.0175898 0.692598 0.132867
train_breast_cancer_c28a3_00008TERMINATED127.0.0.1:544470.000429759 7 1 0.88524 1 0.0193739 0.692785 0.0559441
train_breast_cancer_c28a3_00009TERMINATED127.0.0.1:544480.0149863 2 1 0.722738 1 0.0165932 0.682266 0.111888


" - ], - "text/plain": [ - "" - ] - }, - "metadata": {}, - "output_type": "display_data" - }, - { - "name": "stdout", - "output_type": "stream", - "text": [ - "Result for train_breast_cancer_c28a3_00000:\n", - " date: 2022-07-22_16-55-55\n", - " done: false\n", - " eval-error: 0.08391608391608392\n", - " eval-logloss: 0.6790360066440556\n", - " experiment_id: 2a3189442db341519836a07fb2d65dd9\n", - " hostname: Kais-MacBook-Pro.local\n", - " iterations_since_restore: 1\n", - " node_ip: 127.0.0.1\n", - " pid: 54416\n", - " time_since_restore: 0.01624011993408203\n", - " time_this_iter_s: 0.01624011993408203\n", - " time_total_s: 0.01624011993408203\n", - " timestamp: 1658505355\n", - " timesteps_since_restore: 0\n", - " training_iteration: 1\n", - " trial_id: c28a3_00000\n", - " warmup_time: 0.0035409927368164062\n", - " \n", - "Result for train_breast_cancer_c28a3_00000:\n", - " date: 2022-07-22_16-55-56\n", - " done: true\n", - " eval-error: 0.06293706293706294\n", - " eval-logloss: 0.5714958122560194\n", - " experiment_id: 2a3189442db341519836a07fb2d65dd9\n", - " hostname: Kais-MacBook-Pro.local\n", - " iterations_since_restore: 10\n", - " node_ip: 127.0.0.1\n", - " pid: 54416\n", - " time_since_restore: 0.22218012809753418\n", - " time_this_iter_s: 0.007044076919555664\n", - " time_total_s: 0.22218012809753418\n", - " timestamp: 1658505356\n", - " timesteps_since_restore: 0\n", - " training_iteration: 10\n", - " trial_id: c28a3_00000\n", - " warmup_time: 0.0035409927368164062\n", - " \n", - "Result for train_breast_cancer_c28a3_00003:\n", - " date: 2022-07-22_16-56-01\n", - " done: false\n", - " eval-error: 0.08391608391608392\n", - " eval-logloss: 0.6472820101918041\n", - " experiment_id: 7ff6133237404b4ea4755b9f8cd114f2\n", - " hostname: Kais-MacBook-Pro.local\n", - " iterations_since_restore: 1\n", - " node_ip: 127.0.0.1\n", - " pid: 54442\n", - " time_since_restore: 0.023206233978271484\n", - " time_this_iter_s: 0.023206233978271484\n", - " time_total_s: 0.023206233978271484\n", - " timestamp: 1658505361\n", - " timesteps_since_restore: 0\n", - " training_iteration: 1\n", - " trial_id: c28a3_00003\n", - " warmup_time: 0.006722211837768555\n", - " \n", - "Result for train_breast_cancer_c28a3_00005:\n", - " date: 2022-07-22_16-56-01\n", - " done: true\n", - " eval-error: 0.08391608391608392\n", - " eval-logloss: 0.6914464114429234\n", - " experiment_id: 344762ab6d574b63a9374e19526d0510\n", - " hostname: Kais-MacBook-Pro.local\n", - " iterations_since_restore: 1\n", - " node_ip: 127.0.0.1\n", - " pid: 54444\n", - " time_since_restore: 0.02385115623474121\n", - " time_this_iter_s: 0.02385115623474121\n", - " time_total_s: 0.02385115623474121\n", - " timestamp: 1658505361\n", - " timesteps_since_restore: 0\n", - " training_iteration: 1\n", - " trial_id: c28a3_00005\n", - " warmup_time: 0.008936882019042969\n", - " \n", - "Result for train_breast_cancer_c28a3_00009:\n", - " date: 2022-07-22_16-56-01\n", - " done: true\n", - " eval-error: 0.11188811188811189\n", - " eval-logloss: 0.6822656309688008\n", - " experiment_id: 133901655fa64bf79f2dcc4e8e5e41b1\n", - " hostname: Kais-MacBook-Pro.local\n", - " iterations_since_restore: 1\n", - " node_ip: 127.0.0.1\n", - " pid: 54448\n", - " time_since_restore: 0.016593217849731445\n", - " time_this_iter_s: 0.016593217849731445\n", - " time_total_s: 0.016593217849731445\n", - " timestamp: 1658505361\n", - " timesteps_since_restore: 0\n", - " training_iteration: 1\n", - " trial_id: c28a3_00009\n", - " warmup_time: 0.004940032958984375\n", - " \n", - "Result for train_breast_cancer_c28a3_00007:\n", - " date: 2022-07-22_16-56-01\n", - " done: true\n", - " eval-error: 0.13286713286713286\n", - " eval-logloss: 0.6925980357023386\n", - " experiment_id: b4331027cbaf442ab905b2e51797dbbd\n", - " hostname: Kais-MacBook-Pro.local\n", - " iterations_since_restore: 1\n", - " node_ip: 127.0.0.1\n", - " pid: 54446\n", - " time_since_restore: 0.017589807510375977\n", - " time_this_iter_s: 0.017589807510375977\n", - " time_total_s: 0.017589807510375977\n", - " timestamp: 1658505361\n", - " timesteps_since_restore: 0\n", - " training_iteration: 1\n", - " trial_id: c28a3_00007\n", - " warmup_time: 0.003782033920288086\n", - " \n", - "Result for train_breast_cancer_c28a3_00006:\n", - " date: 2022-07-22_16-56-01\n", - " done: true\n", - " eval-error: 0.11888111888111888\n", - " eval-logloss: 0.6926244418104212\n", - " experiment_id: d3906de5943a4e05a4cc782382f67d24\n", - " hostname: Kais-MacBook-Pro.local\n", - " iterations_since_restore: 1\n", - " node_ip: 127.0.0.1\n", - " pid: 54445\n", - " time_since_restore: 0.015311956405639648\n", - " time_this_iter_s: 0.015311956405639648\n", - " time_total_s: 0.015311956405639648\n", - " timestamp: 1658505361\n", - " timesteps_since_restore: 0\n", - " training_iteration: 1\n", - " trial_id: c28a3_00006\n", - " warmup_time: 0.005506038665771484\n", - " \n", - "Result for train_breast_cancer_c28a3_00002:\n", - " date: 2022-07-22_16-56-01\n", - " done: false\n", - " eval-error: 0.04895104895104895\n", - " eval-logloss: 0.6752762102580571\n", - " experiment_id: a3645fc2d43145d88a1f5b7cc94df703\n", - " hostname: Kais-MacBook-Pro.local\n", - " iterations_since_restore: 1\n", - " node_ip: 127.0.0.1\n", - " pid: 54441\n", - " time_since_restore: 0.027367830276489258\n", - " time_this_iter_s: 0.027367830276489258\n", - " time_total_s: 0.027367830276489258\n", - " timestamp: 1658505361\n", - " timesteps_since_restore: 0\n", - " training_iteration: 1\n", - " trial_id: c28a3_00002\n", - " warmup_time: 0.0062830448150634766\n", - " \n", - "Result for train_breast_cancer_c28a3_00001:\n", - " date: 2022-07-22_16-56-01\n", - " done: false\n", - " eval-error: 0.07692307692307693\n", - " eval-logloss: 0.6698804135089154\n", - " experiment_id: 85766fe4d9fa482a91e396a8fd509a19\n", - " hostname: Kais-MacBook-Pro.local\n", - " iterations_since_restore: 1\n", - " node_ip: 127.0.0.1\n", - " pid: 54440\n", - " time_since_restore: 0.017169952392578125\n", - " time_this_iter_s: 0.017169952392578125\n", - " time_total_s: 0.017169952392578125\n", - " timestamp: 1658505361\n", - " timesteps_since_restore: 0\n", - " training_iteration: 1\n", - " trial_id: c28a3_00001\n", - " warmup_time: 0.006204843521118164\n", - " \n", - "Result for train_breast_cancer_c28a3_00008:\n", - " date: 2022-07-22_16-56-01\n", - " done: true\n", - " eval-error: 0.05594405594405594\n", - " eval-logloss: 0.692784742458717\n", - " experiment_id: 2c7d8bc38ad04536b1dec76819a2b3bf\n", - " hostname: Kais-MacBook-Pro.local\n", - " iterations_since_restore: 1\n", - " node_ip: 127.0.0.1\n", - " pid: 54447\n", - " time_since_restore: 0.01937389373779297\n", - " time_this_iter_s: 0.01937389373779297\n", - " time_total_s: 0.01937389373779297\n", - " timestamp: 1658505361\n", - " timesteps_since_restore: 0\n", - " training_iteration: 1\n", - " trial_id: c28a3_00008\n", - " warmup_time: 0.004342079162597656\n", - " \n", - "Result for train_breast_cancer_c28a3_00001:\n", - " date: 2022-07-22_16-56-01\n", - " done: true\n", - " eval-error: 0.06293706293706294\n", - " eval-logloss: 0.6503534216980834\n", - " experiment_id: 85766fe4d9fa482a91e396a8fd509a19\n", - " hostname: Kais-MacBook-Pro.local\n", - " iterations_since_restore: 2\n", - " node_ip: 127.0.0.1\n", - " pid: 54440\n", - " time_since_restore: 0.13567376136779785\n", - " time_this_iter_s: 0.11850380897521973\n", - " time_total_s: 0.13567376136779785\n", - " timestamp: 1658505361\n", - " timesteps_since_restore: 0\n", - " training_iteration: 2\n", - " trial_id: c28a3_00001\n", - " warmup_time: 0.006204843521118164\n", - " \n", - "Result for train_breast_cancer_c28a3_00004:\n", - " date: 2022-07-22_16-56-01\n", - " done: true\n", - " eval-error: 0.06993006993006994\n", - " eval-logloss: 0.689577207281873\n", - " experiment_id: ef4fdc645c444112985b4957ab8a84e9\n", - " hostname: Kais-MacBook-Pro.local\n", - " iterations_since_restore: 1\n", - " node_ip: 127.0.0.1\n", - " pid: 54443\n", - " time_since_restore: 0.027021169662475586\n", - " time_this_iter_s: 0.027021169662475586\n", - " time_total_s: 0.027021169662475586\n", - " timestamp: 1658505361\n", - " timesteps_since_restore: 0\n", - " training_iteration: 1\n", - " trial_id: c28a3_00004\n", - " warmup_time: 0.0063669681549072266\n", - " \n", - "Result for train_breast_cancer_c28a3_00002:\n", - " date: 2022-07-22_16-56-01\n", - " done: true\n", - " eval-error: 0.04195804195804196\n", - " eval-logloss: 0.658545415301423\n", - " experiment_id: a3645fc2d43145d88a1f5b7cc94df703\n", - " hostname: Kais-MacBook-Pro.local\n", - " iterations_since_restore: 2\n", - " node_ip: 127.0.0.1\n", - " pid: 54441\n", - " time_since_restore: 0.17307591438293457\n", - " time_this_iter_s: 0.1457080841064453\n", - " time_total_s: 0.17307591438293457\n", - " timestamp: 1658505361\n", - " timesteps_since_restore: 0\n", - " training_iteration: 2\n", - " trial_id: c28a3_00002\n", - " warmup_time: 0.0062830448150634766\n", - " \n", - "Result for train_breast_cancer_c28a3_00003:\n", - " date: 2022-07-22_16-56-01\n", - " done: true\n", - " eval-error: 0.04195804195804196\n", - " eval-logloss: 0.38665050018083796\n", - " experiment_id: 7ff6133237404b4ea4755b9f8cd114f2\n", - " hostname: Kais-MacBook-Pro.local\n", - " iterations_since_restore: 10\n", - " node_ip: 127.0.0.1\n", - " pid: 54442\n", - " time_since_restore: 0.28186488151550293\n", - " time_this_iter_s: 0.03063178062438965\n", - " time_total_s: 0.28186488151550293\n", - " timestamp: 1658505361\n", - " timesteps_since_restore: 0\n", - " training_iteration: 10\n", - " trial_id: c28a3_00003\n", - " warmup_time: 0.006722211837768555\n", - " \n" - ] - }, - { - "name": "stderr", - "output_type": "stream", - "text": [ - "2022-07-22 16:56:01,498\tINFO tune.py:738 -- Total run time: 10.53 seconds (10.37 seconds for the tuning loop).\n" - ] - }, - { - "name": "stdout", - "output_type": "stream", - "text": [ - "Best model parameters: {'objective': 'binary:logistic', 'eval_metric': ['logloss', 'error'], 'max_depth': 2, 'min_child_weight': 3, 'subsample': 0.782626252548841, 'eta': 0.06385952388342125}\n", - "Best model total accuracy: 0.9580\n" - ] - } - ], - "source": [ - "import sklearn.datasets\n", - "import sklearn.metrics\n", - "import os\n", - "from ray.tune.schedulers import ASHAScheduler\n", - "from sklearn.model_selection import train_test_split\n", - "import xgboost as xgb\n", - "\n", - "from ray import air, tune\n", - "from ray.air import session\n", - "from ray.tune.integration.xgboost import TuneReportCheckpointCallback\n", - "\n", - "\n", - "def train_breast_cancer(config: dict):\n", - " # This is a simple training function to be passed into Tune\n", - " # Load dataset\n", - " data, labels = sklearn.datasets.load_breast_cancer(return_X_y=True)\n", - " # Split into train and test set\n", - " train_x, test_x, train_y, test_y = train_test_split(data, labels, test_size=0.25)\n", - " # Build input matrices for XGBoost\n", - " train_set = xgb.DMatrix(train_x, label=train_y)\n", - " test_set = xgb.DMatrix(test_x, label=test_y)\n", - " # Train the classifier, using the Tune callback\n", - " xgb.train(\n", - " config,\n", - " train_set,\n", - " evals=[(test_set, \"eval\")],\n", - " verbose_eval=False,\n", - " callbacks=[TuneReportCheckpointCallback(filename=\"model.xgb\")],\n", - " )\n", - "\n", - "\n", - "def get_best_model_checkpoint(results):\n", - " best_bst = xgb.Booster()\n", - " best_result = results.get_best_result()\n", - "\n", - " with best_result.checkpoint.as_directory() as best_checkpoint_dir:\n", - " best_bst.load_model(os.path.join(best_checkpoint_dir, \"model.xgb\"))\n", - " accuracy = 1.0 - best_result.metrics[\"eval-error\"]\n", - " print(f\"Best model parameters: {best_result.config}\")\n", - " print(f\"Best model total accuracy: {accuracy:.4f}\")\n", - " return best_bst\n", - "\n", - "\n", - "def tune_xgboost(smoke_test=False):\n", - " search_space = {\n", - " # You can mix constants with search space objects.\n", - " \"objective\": \"binary:logistic\",\n", - " \"eval_metric\": [\"logloss\", \"error\"],\n", - " \"max_depth\": tune.randint(1, 9),\n", - " \"min_child_weight\": tune.choice([1, 2, 3]),\n", - " \"subsample\": tune.uniform(0.5, 1.0),\n", - " \"eta\": tune.loguniform(1e-4, 1e-1),\n", - " }\n", - " # This will enable aggressive early stopping of bad trials.\n", - " scheduler = ASHAScheduler(\n", - " max_t=10, grace_period=1, reduction_factor=2 # 10 training iterations\n", - " )\n", - "\n", - " tuner = tune.Tuner(\n", - " train_breast_cancer,\n", - " tune_config=tune.TuneConfig(\n", - " metric=\"eval-logloss\",\n", - " mode=\"min\",\n", - " scheduler=scheduler,\n", - " num_samples=1 if smoke_test else 10,\n", - " ),\n", - " param_space=search_space,\n", - " )\n", - " results = tuner.fit()\n", - "\n", - " return results\n", - "\n", - "\n", - "if __name__ == \"__main__\":\n", - " import argparse\n", - "\n", - " parser = argparse.ArgumentParser()\n", - " parser.add_argument(\n", - " \"--smoke-test\", action=\"store_true\", help=\"Finish quickly for testing\"\n", - " )\n", - " args, _ = parser.parse_known_args()\n", - "\n", - " results = tune_xgboost(smoke_test=args.smoke_test)\n", - "\n", - " # Load the best model checkpoint.\n", - " best_bst = get_best_model_checkpoint(results)\n", - "\n", - " # You could now do further predictions with\n", - " # best_bst.predict(...)\n" - ] - }, - { - "cell_type": "markdown", - "id": "20732fe4", - "metadata": {}, - "source": [ - "The output of our run could look like this:\n", - "\n", - "```{code-block} bash\n", - ":emphasize-lines: 7\n", - "\n", - " Number of trials: 10/10 (10 TERMINATED)\n", - " +---------------------------------+------------+-------+-------------+-------------+--------------------+-------------+--------+------------------+----------------+--------------+\n", - " | Trial name | status | loc | eta | max_depth | min_child_weight | subsample | iter | total time (s) | eval-logloss | eval-error |\n", - " |---------------------------------+------------+-------+-------------+-------------+--------------------+-------------+--------+------------------+----------------+--------------|\n", - " | train_breast_cancer_ba275_00000 | TERMINATED | | 0.00205087 | 2 | 1 | 0.898391 | 10 | 0.380619 | 0.678039 | 0.090909 |\n", - " | train_breast_cancer_ba275_00001 | TERMINATED | | 0.000183834 | 4 | 3 | 0.924939 | 1 | 0.0228798 | 0.693009 | 0.111888 |\n", - " | train_breast_cancer_ba275_00002 | TERMINATED | | 0.0242721 | 7 | 2 | 0.501551 | 10 | 0.376154 | 0.54472 | 0.06993 |\n", - " | train_breast_cancer_ba275_00003 | TERMINATED | | 0.000449692 | 5 | 3 | 0.890212 | 1 | 0.0234981 | 0.692811 | 0.090909 |\n", - " | train_breast_cancer_ba275_00004 | TERMINATED | | 0.000376393 | 7 | 2 | 0.883609 | 1 | 0.0231569 | 0.692847 | 0.062937 |\n", - " | train_breast_cancer_ba275_00005 | TERMINATED | | 0.00231942 | 3 | 3 | 0.877464 | 2 | 0.104867 | 0.689541 | 0.083916 |\n", - " | train_breast_cancer_ba275_00006 | TERMINATED | | 0.000542326 | 1 | 2 | 0.578584 | 1 | 0.0213971 | 0.692765 | 0.083916 |\n", - " | train_breast_cancer_ba275_00007 | TERMINATED | | 0.0016801 | 1 | 2 | 0.975302 | 1 | 0.02226 | 0.691999 | 0.083916 |\n", - " | train_breast_cancer_ba275_00008 | TERMINATED | | 0.000595756 | 8 | 3 | 0.58429 | 1 | 0.0221152 | 0.692657 | 0.06993 |\n", - " | train_breast_cancer_ba275_00009 | TERMINATED | | 0.000357845 | 8 | 1 | 0.637776 | 1 | 0.022635 | 0.692859 | 0.090909 |\n", - " +---------------------------------+------------+-------+-------------+-------------+--------------------+-------------+--------+------------------+----------------+--------------+\n", - "\n", - "\n", - " Best model parameters: {'objective': 'binary:logistic', 'eval_metric': ['logloss', 'error'], 'max_depth': 7, 'min_child_weight': 2, 'subsample': 0.5015513240240503, 'eta': 0.024272050872920895}\n", - " Best model total accuracy: 0.9301\n", - "```\n", - "\n", - "As you can see, most trials have been stopped only after a few iterations. Only the\n", - "two most promising trials were run for the full 10 iterations.\n", - "\n", - "You can also ensure that all available resources are being used as the scheduler\n", - "terminates trials, freeing them up. This can be done through the\n", - "`ResourceChangingScheduler`. An example of this can be found here:\n", - "{doc}`/tune/examples/includes/xgboost_dynamic_resources_example`.\n", - "\n", - "## Using fractional GPUs\n", - "\n", - "You can often accelerate your training by using GPUs in addition to CPUs. However,\n", - "you usually don't have as many GPUs as you have trials to run. For instance, if you\n", - "run 10 Tune trials in parallel, you usually don't have access to 10 separate GPUs.\n", - "\n", - "Tune supports *fractional GPUs*. This means that each task is assigned a fraction\n", - "of the GPU memory for training. For 10 tasks, this could look like this:" - ] - }, - { - "cell_type": "code", - "execution_count": null, - "id": "7d1b20a3", - "metadata": {}, - "outputs": [], - "source": [ - "config = {\n", - " \"objective\": \"binary:logistic\",\n", - " \"eval_metric\": [\"logloss\", \"error\"],\n", - " \"tree_method\": \"gpu_hist\",\n", - " \"max_depth\": tune.randint(1, 9),\n", - " \"min_child_weight\": tune.choice([1, 2, 3]),\n", - " \"subsample\": tune.uniform(0.5, 1.0),\n", - " \"eta\": tune.loguniform(1e-4, 1e-1),\n", - "}\n", - "\n", - "tuner = tune.Tuner(\n", - " tune.with_resources(train_breast_cancer, resources={\"cpu\": 1, \"gpu\": 0.1}),\n", - " tune_config=tune.TuneConfig(\n", - " num_samples=10,\n", - " ),\n", - " param_space=config,\n", - ")\n", - "results = tuner.fit()\n" - ] - }, - { - "cell_type": "markdown", - "id": "ee131861", - "metadata": {}, - "source": [ - "Each task thus works with 10% of the available GPU memory. You also have to tell\n", - "XGBoost to use the `gpu_hist` tree method, so it knows it should use the GPU.\n", - "\n", - "## Conclusion\n", - "\n", - "You should now have a basic understanding on how to train XGBoost models and on how\n", - "to tune the hyperparameters to yield the best results. In our simple example,\n", - "Tuning the parameters didn't make a huge difference for the accuracy.\n", - "But in larger applications, intelligent hyperparameter tuning can make the\n", - "difference between a model that doesn't seem to learn at all, and a model\n", - "that outperforms all the other ones.\n", - "\n", - "## More XGBoost Examples\n", - "\n", - "- {doc}`/tune/examples/includes/xgboost_dynamic_resources_example`:\n", - " Trains a basic XGBoost model with Tune with the class-based API and a ResourceChangingScheduler, ensuring all resources are being used at all time.\n", - "\n", - "## Learn More\n", - "\n", - "- [XGBoost Hyperparameter Tuning - A Visual Guide](https://kevinvecmanis.io/machine%20learning/hyperparameter%20tuning/dataviz/python/2019/05/11/XGBoost-Tuning-Visual-Guide.html)\n", - "- [Notes on XGBoost Parameter Tuning](https://xgboost.readthedocs.io/en/latest/tutorials/param_tuning.html)\n", - "- [Doing XGBoost Hyperparameter Tuning the smart way](https://towardsdatascience.com/doing-xgboost-hyper-parameter-tuning-the-smart-way-part-1-of-2-f6d255a45dde)" - ] - } - ], - "metadata": { - "kernelspec": { - "display_name": "ray_dev_py38", - "language": "python", - "name": "python3" - }, - "language_info": { - "codemirror_mode": { - "name": "ipython", - "version": 3 - }, - "file_extension": ".py", - "mimetype": "text/x-python", - "name": "python", - "nbconvert_exporter": "python", - "pygments_lexer": "ipython3", - "version": "3.8.13 | packaged by conda-forge | (default, Mar 25 2022, 06:05:16) \n[Clang 12.0.1 ]" - }, - "orphan": true, - "vscode": { - "interpreter": { - "hash": "265d195fda5292fe8f69c6e37c435a5634a1ed3b6799724e66a975f68fa21517" - } - } - }, - "nbformat": 4, - "nbformat_minor": 5 -} + "nbformat": 4, + "nbformat_minor": 5 +} \ No newline at end of file diff --git a/doc/source/tune/faq.rst b/doc/source/tune/faq.rst index 20ff41119fc6..ca4db42ef5df 100644 --- a/doc/source/tune/faq.rst +++ b/doc/source/tune/faq.rst @@ -341,7 +341,7 @@ are efficiently stored and retrieved on your cluster machines. :func:`tune.with_parameters() ` also works with class trainables. Please see -:ref:`here for further details ` and examples. +:func:`tune.with_parameters() ` for more details and examples. How can I reproduce experiments? diff --git a/doc/source/tune/getting-started.rst b/doc/source/tune/getting-started.rst index 624fe86d5409..14e303b2dae3 100644 --- a/doc/source/tune/getting-started.rst +++ b/doc/source/tune/getting-started.rst @@ -73,7 +73,7 @@ make sure that the function is :ref:`serializable by Ray `. :start-after: __train_func_begin__ :end-before: __train_func_end__ -Let's run one trial by calling :ref:`Tuner.fit ` and :ref:`randomly sample ` +Let's run one trial by calling :ref:`Tuner.fit ` and :ref:`randomly sample ` from a uniform distribution for learning rate and momentum. .. literalinclude:: /../../python/ray/tune/tests/tutorial.py diff --git a/doc/source/tune/key-concepts.rst b/doc/source/tune/key-concepts.rst index 40944fe73a49..0606566d8444 100644 --- a/doc/source/tune/key-concepts.rst +++ b/doc/source/tune/key-concepts.rst @@ -81,10 +81,10 @@ how these values are sampled (e.g. from a uniform distribution or a normal distribution). Tune offers various functions to define search spaces and sampling methods. -:ref:`You can find the documentation of these search space definitions here `. +:ref:`You can find the documentation of these search space definitions here `. Here's an example covering all search space functions. Again, -:ref:`here is the full explanation of all these functions `. +:ref:`here is the full explanation of all these functions `. .. literalinclude:: doc_code/key_concepts.py :language: python diff --git a/doc/source/tune/tutorials/tune-resources.rst b/doc/source/tune/tutorials/tune-resources.rst index aca47a7a2f4f..540533582c5e 100644 --- a/doc/source/tune/tutorials/tune-resources.rst +++ b/doc/source/tune/tutorials/tune-resources.rst @@ -18,7 +18,7 @@ of CPUs (cores) on your machine. ) results = tuner.fit() -You can override this per trial resources with :ref:`tune-with-resources`. Here you can +You can override this per trial resources with :func:`tune.with_resources `. Here you can specify your resource requests using either a dictionary, a :class:`~ray.air.config.ScalingConfig`, or a :class:`PlacementGroupFactory ` object. In any case, Ray Tune will try to start a placement group for each trial. diff --git a/doc/source/tune/tutorials/tune-search-spaces.rst b/doc/source/tune/tutorials/tune-search-spaces.rst index d10dd84b3a6e..bb2f441b7ecd 100644 --- a/doc/source/tune/tutorials/tune-search-spaces.rst +++ b/doc/source/tune/tutorials/tune-search-spaces.rst @@ -16,7 +16,7 @@ Thereby, you can either use the ``tune.grid_search`` primitive to use grid searc results = tuner.fit() -Or you can use one of the random sampling primitives to specify distributions (:ref:`tune-sample-docs`): +Or you can use one of the random sampling primitives to specify distributions (:doc:`/tune/api/search_space`): .. code-block:: python @@ -130,7 +130,7 @@ for a total of 90 trials, each with randomly sampled values of ``alpha`` and ``b .. tip:: Avoid passing large objects as values in the search space, as that will incur a performance overhead. - Use :ref:`tune-with-parameters` to pass large objects in or load them inside your trainable + Use :func:`tune.with_parameters ` to pass large objects in or load them inside your trainable from disk (making sure that all nodes have access to the files) or cloud storage. See :ref:`tune-bottlenecks` for more information. diff --git a/doc/source/tune/tutorials/tune_get_data_in_and_out.md b/doc/source/tune/tutorials/tune_get_data_in_and_out.md index 01e12d05571a..6585366114f2 100644 --- a/doc/source/tune/tutorials/tune_get_data_in_and_out.md +++ b/doc/source/tune/tutorials/tune_get_data_in_and_out.md @@ -62,7 +62,7 @@ Objects from the outer scope of the `training_function` will also be automatical TL;DR - use the `param_space` argument to specify small, serializable constants and variables. ``` -The first way of passing inputs into Trainables is the [*search space*](tune-key-concepts-search-spaces) (it may also be called *parameter space* or *config*). In the Trainable itself, it maps to the `config` dict passed in as an argument to the function. You define the search space using the `param_space` argument of the `Tuner`. The search space is a dict and may be composed of [*distributions*](), which will sample a different value for each Trial, or of constant values. The search space may be composed of nested dictionaries, and those in turn can have distributions as well. +The first way of passing inputs into Trainables is the [*search space*](tune-key-concepts-search-spaces) (it may also be called *parameter space* or *config*). In the Trainable itself, it maps to the `config` dict passed in as an argument to the function. You define the search space using the `param_space` argument of the `Tuner`. The search space is a dict and may be composed of [*distributions*](), which will sample a different value for each Trial, or of constant values. The search space may be composed of nested dictionaries, and those in turn can have distributions as well. ```{warning} Each value in the search space will be saved directly in the Trial metadata. This means that every value in the search space **must** be serializable and take up a small amount of memory. @@ -116,7 +116,7 @@ tuner = Tuner( TL;DR - use the `tune.with_parameters` util function to specify large constant parameters. ``` -If we have large objects that are constant across Trials, we can use the [`tune.with_parameters`](tune-with-parameters) utility to pass them into the Trainable directly. The objects will be stored in the [Ray object store](serialization-guide) so that each Trial worker may access them to obtain a local copy to use in its process. +If we have large objects that are constant across Trials, we can use the {func}`tune.with_parameters ` utility to pass them into the Trainable directly. The objects will be stored in the [Ray object store](serialization-guide) so that each Trial worker may access them to obtain a local copy to use in its process. ```{tip} Objects put into the Ray object store must be serializable. diff --git a/python/ray/tune/search/bayesopt/bayesopt_search.py b/python/ray/tune/search/bayesopt/bayesopt_search.py index 160bab1918d6..d02752b52cbd 100644 --- a/python/ray/tune/search/bayesopt/bayesopt_search.py +++ b/python/ray/tune/search/bayesopt/bayesopt_search.py @@ -53,8 +53,14 @@ class BayesOptSearch(Searcher): pip install bayesian-optimization - This algorithm requires setting a search space using the - `BayesianOptimization search space specification`_. + Initializing this search algorithm with a ``space`` requires that it's + in the ``BayesianOptimization`` search space format. Otherwise, you + should instead pass in a Tune search space into ``Tuner(param_space=...)``, + and the search space will be automatically converted for you. + + See this `BayesianOptimization example notebook + `_ + for an example. Args: space: Continuous search space. Parameters will be sampled from diff --git a/python/ray/tune/search/optuna/optuna_search.py b/python/ray/tune/search/optuna/optuna_search.py index 2462ac1af482..0c20fbb51abc 100644 --- a/python/ray/tune/search/optuna/optuna_search.py +++ b/python/ray/tune/search/optuna/optuna_search.py @@ -120,6 +120,8 @@ class OptunaSearch(Searcher): draw hyperparameter configurations. Defaults to ``MOTPESampler`` for multi-objective optimization with Optuna<2.9.0, and ``TPESampler`` in every other case. + See https://optuna.readthedocs.io/en/stable/reference/samplers/index.html + for available Optuna samplers. .. warning:: Please note that with Optuna 2.10.0 and earlier diff --git a/python/ray/tune/search/skopt/skopt_search.py b/python/ray/tune/search/skopt/skopt_search.py index a207f938f1df..c3d8acf752aa 100644 --- a/python/ray/tune/search/skopt/skopt_search.py +++ b/python/ray/tune/search/skopt/skopt_search.py @@ -43,10 +43,10 @@ class SkOptSearch(Searcher): pip install scikit-optimize - This Search Algorithm requires you to pass in a `skopt Optimizer object`_. + This Search Algorithm requires you to pass in a `skopt Optimizer object + `_. - This searcher will automatically filter out any NaN, inf or -inf - results. + This searcher will automatically filter out any NaN, inf or -inf results. Parameters: optimizer: Optimizer provided