ray/doc/source/tune-usage.rst

Tune User Guide
===============

Tune Overview
-------------

.. image:: images/tune-api.svg

Tune schedules a number of *trials* in a cluster. Each trial runs a user-defined Python function or class and is parameterized either by a *config* variation from Tune's Variant Generator or a user-specified **search algorithm**. The trials are scheduled and managed by a **trial scheduler**.

More information about Tune's `search algorithms can be found here <tune-searchalg.html>`__.

More information about Tune's `trial schedulers can be found here <tune-schedulers.html>`__.


Start by installing, importing, and initializing Ray.

.. code-block:: python

    import ray
    import ray.tune as tune

    ray.init()

Tune provides a ``run_experiments`` function that generates and runs the trials as described by the `experiment specification <tune-usage.html#experiment-configuration>`__.

.. autofunction:: ray.tune.run_experiments
    :noindex:

This function will report status on the command line until all Trials stop:

::

    == Status ==
    Using FIFO scheduling algorithm.
    Resources used: 4/8 CPUs, 0/0 GPUs
    Result logdir: ~/ray_results/my_experiment
     - train_func_0_lr=0.2,momentum=1:  RUNNING [pid=6778], 209 s, 20604 ts, 7.29 acc
     - train_func_1_lr=0.4,momentum=1:  RUNNING [pid=6780], 208 s, 20522 ts, 53.1 acc
     - train_func_2_lr=0.6,momentum=1:  TERMINATED [pid=6789], 21 s, 2190 ts, 100 acc
     - train_func_3_lr=0.2,momentum=2:  RUNNING [pid=6791], 208 s, 41004 ts, 8.37 acc
     - train_func_4_lr=0.4,momentum=2:  RUNNING [pid=6800], 209 s, 41204 ts, 70.1 acc
     - train_func_5_lr=0.6,momentum=2:  TERMINATED [pid=6809], 10 s, 2164 ts, 100 acc


Experiment Configuration
------------------------

Specifying Experiments
~~~~~~~~~~~~~~~~~~~~~~

There are two ways to specify the configuration for an experiment - one via Python and one via JSON.

**Using Python**: specify a configuration is to create an Experiment object.

.. autoclass:: ray.tune.Experiment
    :noindex:

An example of this can be found in `hyperband_example.py <https://github.com/ray-project/ray/blob/master/python/ray/tune/examples/hyperband_example.py>`__.

**Using JSON/Dict**: This uses the same fields as the ``ray.tune.Experiment``, except the experiment name is the key of the top level
dictionary. Tune will convert the dict into an ``ray.tune.Experiment`` object.

.. code-block:: python

    experiment_spec = {
        "my_experiment_name": {
            "run": my_func,
            "stop": { "mean_accuracy": 100 },
            "config": {
                "alpha": tune.grid_search([0.2, 0.4, 0.6]),
                "beta": tune.grid_search([1, 2]),
            },
            "trial_resources": { "cpu": 1, "gpu": 0 },
            "num_samples": 10,
            "local_dir": "~/ray_results",
            "upload_dir": "s3://your_bucket/path",
            "checkpoint_freq": 10,
            "max_failures": 2
        }
    }
    run_experiments(experiment_spec)


An example of this can be found in `async_hyperband_example.py <https://github.com/ray-project/ray/blob/master/python/ray/tune/examples/async_hyperband_example.py>`__.

Model API
~~~~~~~~~

You can either pass in a Python function or Python class for model training as follows, each requiring a specific signature/interface:

.. code-block:: python
   :emphasize-lines: 3,8

    experiment_spec = {
        "my_experiment_name": {
            "run": my_trainable
        }
    }

    # or with the Experiment API
    experiment_spec = Experiment("my_experiment_name", my_trainable)

    run_experiments(experiments=experiment_spec)


**Python functions** will need to have the following signature:

.. code-block:: python

    def trainable(config, reporter):
        """
        Args:
            config (dict): Parameters provided from the search algorithm
                or variant generation.
            reporter (Reporter): Handle to report intermediate metrics to Tune.
        """

Tune will run this function on a separate thread in a Ray actor process. Note that trainable functions are not checkpointable, since they never return control back to their caller. See `Trial Checkpointing for more details <tune-usage.html#trial-checkpointing>`__.

.. note::
    If you have a lambda function that you want to train, you will need to first register the function: ``tune.register_trainable("lambda_id", lambda x: ...)``. You can then use ``lambda_id`` in place of ``my_trainable``.

**Python classes** passed into Tune will need to subclass ``ray.tune.Trainable``.

.. autoclass:: ray.tune.Trainable
    :members: __init__, _save, _restore, _train, _setup, _stop
    :noindex:


Tune Search Space (Default)
~~~~~~~~~~~~~~~~~~~~~~~~~~~


You can use ``tune.grid_search`` to specify an axis of a grid search. By default, Tune also supports sampling parameters from user-specified lambda functions, which can be used independently or in combination with grid search.

.. note::
    If you specify an explicit Search Algorithm such as any SuggestionAlgorithm, you may not be able to specify lambdas or grid search with this interface, as the search algorithm may require a different search space declaration.

The following shows grid search over two nested parameters combined with random sampling from two lambda functions, generating 9 different trials. Note that the value of ``beta`` depends on the value of ``alpha``, which is represented by referencing ``spec.config.alpha`` in the lambda function. This lets you specify conditional parameter distributions.

.. code-block:: python
   :emphasize-lines: 4-11

    run_experiments({
        "my_experiment_name": {
            "run": my_trainable,
            "config": {
                "alpha": lambda spec: np.random.uniform(100),
                "beta": lambda spec: spec.config.alpha * np.random.normal(),
                "nn_layers": [
                    tune.grid_search([16, 64, 256]),
                    tune.grid_search([16, 64, 256]),
                ],
            }
        }
    })


.. note::
    Lambda functions will be evaluated during trial variant generation. If you need to pass a literal function in your config, use ``tune.function(...)`` to escape it.

For more information on variant generation, see `basic_variant.py <https://github.com/ray-project/ray/blob/master/python/ray/tune/suggest/basic_variant.py>`__.

Sampling Multiple Times
~~~~~~~~~~~~~~~~~~~~~~~

By default, each random variable and grid search point is sampled once. To take multiple random samples, add ``num_samples: N`` to the experiment config. If `grid_search` is provided as an argument, the grid will be repeated `num_samples` of times.

.. code-block:: python
   :emphasize-lines: 12

    run_experiments({
        "my_experiment_name": {
            "run": my_trainable,
            "config": {
                "alpha": lambda spec: np.random.uniform(100),
                "beta": lambda spec: spec.config.alpha * np.random.normal(),
                "nn_layers": [
                    tune.grid_search([16, 64, 256]),
                    tune.grid_search([16, 64, 256]),
                ],
            },
            "num_samples": 10
        }
    })

E.g. in the above, ``"num_samples": 10`` repeats the 3x3 grid search 10 times, for a total of 90 trials, each with randomly sampled values of ``alpha`` and ``beta``.


Using GPUs (Resource Allocation)
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Tune will allocate the specified GPU and CPU ``trial_resources`` to each individual trial (defaulting to 1 CPU per trial). Under the hood, Tune runs each trial as a Ray actor, using Ray's resource handling to allocate resources and place actors. A trial will not be scheduled unless at least that amount of resources is available in the cluster, preventing the cluster from being overloaded.

If GPU resources are not requested, the ``CUDA_VISIBLE_DEVICES`` environment variable will be set as empty, disallowing GPU access.
Otherwise, it will be set to the GPUs in the list (this is managed by Ray).

If your trainable function / class creates further Ray actors or tasks that also consume CPU / GPU resources, you will also want to set ``extra_cpu`` or ``extra_gpu`` to reserve extra resource slots for the actors you will create. For example, if a trainable class requires 1 GPU itself, but will launch 4 actors each using another GPU, then it should set ``"gpu": 1, "extra_gpu": 4``.

.. code-block:: python
   :emphasize-lines: 4-8

    run_experiments({
        "my_experiment_name": {
            "run": my_trainable,
            "trial_resources": {
                "cpu": 1,
                "gpu": 1,
                "extra_gpu": 4
            }
        }
    })


Trial Checkpointing
~~~~~~~~~~~~~~~~~~~

To enable checkpointing, you must implement a `Trainable class <tune-usage.html#model-api>`__ (Trainable functions are not checkpointable, since they never return control back to their caller). The easiest way to do this is to subclass the pre-defined ``Trainable`` class and implement its ``_train``, ``_save``, and ``_restore`` abstract methods `(example) <https://github.com/ray-project/ray/blob/master/python/ray/tune/examples/hyperband_example.py>`__. Implementing this interface is required to support resource multiplexing in  Trial Schedulers such as HyperBand and PBT.

For TensorFlow model training, this would look something like this `(full tensorflow example) <https://github.com/ray-project/ray/blob/master/python/ray/tune/examples/tune_mnist_ray_hyperband.py>`__:

.. code-block:: python

    class MyClass(Trainable):
        def _setup(self):
            self.saver = tf.train.Saver()
            self.sess = ...
            self.iteration = 0

        def _train(self):
            self.sess.run(...)
            self.iteration += 1

        def _save(self, checkpoint_dir):
            return self.saver.save(
                self.sess, checkpoint_dir + "/save",
                global_step=self.iteration)

        def _restore(self, path):
            return self.saver.restore(self.sess, path)


Additionally, checkpointing can be used to provide fault-tolerance for experiments. This can be enabled by setting ``checkpoint_freq: N`` and ``max_failures: M`` to checkpoint trials every *N* iterations and recover from up to *M* crashes per trial, e.g.:

.. code-block:: python
   :emphasize-lines: 4,5

    run_experiments({
        "my_experiment_name": {
            "run": my_trainable
            "checkpoint_freq": 10,
            "max_failures": 5,
        },
    })

The checkpoint_freq may not coincide with the exact end of an experiment. If you want a checkpoint to be created at the end
of a trial, you can additionally set the checkpoint_at_end to True. An example is shown below:

.. code-block:: python
   :emphasize-lines: 5

    run_experiments({
        "my_experiment_name": {
            "run": my_trainable
            "checkpoint_freq": 10,
            "checkpoint_at_end": True,
            "max_failures": 5,
        },
    })

Handling Large Datasets
-----------------------

You often will want to compute a large object (e.g., training data, model weights) on the driver and use that object within each trial. Tune provides a ``pin_in_object_store`` utility function that can be used to broadcast such large objects. Objects pinned in this way will never be evicted from the Ray object store while the driver process is running, and can be efficiently retrieved from any task via ``get_pinned_object``.

.. code-block:: python

    import ray
    from ray.tune import run_experiments
    from ray.tune.util import pin_in_object_store, get_pinned_object

    import numpy as np

    ray.init()

    # X_id can be referenced in closures
    X_id = pin_in_object_store(np.random.random(size=100000000))

    def f(config, reporter):
        X = get_pinned_object(X_id)
        # use X

    run_experiments({
        "my_experiment_name": {
            "run": f
        }
    })


Logging and Visualizing Results
-------------------------------

All results reported by the trainable will be logged locally to a unique directory per experiment, e.g. ``~/ray_results/my_experiment`` in the above example. On a cluster, incremental results will be synced to local disk on the head node. The log records are compatible with a number of visualization tools:

To visualize learning in tensorboard, install TensorFlow:

.. code-block:: bash

    $ pip install tensorflow

Then, after you run a experiment, you can visualize your experiment with TensorBoard by specifying the output directory of your results. Note that if you running Ray on a remote cluster, you can forward the tensorboard port to your local machine through SSH using ``ssh -L 6006:localhost:6006 <address>``:

.. code-block:: bash

    $ tensorboard --logdir=~/ray_results/my_experiment

.. image:: ray-tune-tensorboard.png

To use rllab's VisKit (you may have to install some dependencies), run:

.. code-block:: bash

    $ git clone https://github.com/rll/rllab.git
    $ python rllab/rllab/viskit/frontend.py ~/ray_results/my_experiment

.. image:: ray-tune-viskit.png

Finally, to view the results with a `parallel coordinates visualization <https://en.wikipedia.org/wiki/Parallel_coordinates>`__, open `ParallelCoordinatesVisualization.ipynb <https://github.com/ray-project/ray/blob/master/python/ray/tune/ParallelCoordinatesVisualization.ipynb>`__ as follows and run its cells:

.. code-block:: bash

    $ cd $RAY_HOME/python/ray/tune
    $ jupyter-notebook ParallelCoordinatesVisualization.ipynb

.. image:: ray-tune-parcoords.png


Client API
----------

You can modify an ongoing experiment by adding or deleting trials using the Tune Client API. To do this, verify that you have the ``requests`` library installed:

.. code-block:: bash

    $ pip install requests

To use the Client API, you can start your experiment with ``with_server=True``:

.. code-block:: python

    run_experiments({...}, with_server=True, server_port=4321)

Then, on the client side, you can use the following class. The server address defaults to ``localhost:4321``. If on a cluster, you may want to forward this port (e.g. ``ssh -L <local_port>:localhost:<remote_port> <address>``) so that you can use the Client on your local machine.

.. autoclass:: ray.tune.web_server.TuneClient
    :members:


For an example notebook for using the Client API, see the `Client API Example <https://github.com/ray-project/ray/tree/master/python/ray/tune/TuneClient.ipynb>`__.


Examples
--------

You can find a comprehensive of examples `using Tune and its various features here <https://github.com/ray-project/ray/tree/master/python/ray/tune/examples>`__, including examples using Keras, TensorFlow, and Population-Based Training.


Further Questions or Issues?
----------------------------

You can post questions or issues or feedback through the following channels:

1. `Our Mailing List`_: For discussions about development, questions about
   usage, or any general questions and feedback.
2. `GitHub Issues`_: For bug reports and feature requests.

.. _`Our Mailing List`: https://groups.google.com/forum/#!forum/ray-dev
.. _`GitHub Issues`: https://github.com/ray-project/ray/issues