submit_experiment() is an asynchronous call to Azure Machine Learning service to execute a trial on local or remote compute. Depending on the configuration, submit_experiment() will automatically prepare your execution environments, execute your code, and capture your source code and results in the experiment's run history.

To submit an experiment you first need to create a configuration object describing how the experiment is to be run. The configuration depends on the type of trial required. For a script run, provide an Estimator object to the config parameter. For a HyperDrive run for hyperparameter tuning, provide a HyperDriveConfig to config.

submit_experiment(experiment, config, tags = NULL)



The Experiment object.


The Estimator or HyperDriveConfig object.


A named list of tags for the submitted run, e.g. list("tag" = "value").


The ScriptRun or HyperDriveRun object.

See also

estimator(), hyperdrive_config()


# This example submits an Estimator experiment if (FALSE) { ws <- load_workspace_from_config() compute_target <- get_compute(ws, cluster_name = 'mycluster') exp <- experiment(ws, name = 'myexperiment') est <- estimator(source_directory = '.', entry_script = 'train.R', compute_target = compute_target) run <- submit_experiment(exp, est) }