flytekit.extend.TaskPlugins#
- class flytekit.extend.TaskPlugins[source]#
This is the TaskPlugins factory for task types that are derivative of PythonFunctionTask. Every task that the user wishes to use should be available in this factory. Usage
TaskPlugins.register_pythontask_plugin(config_object_type, plugin_object_type) # config_object_type is any class that will be passed to the plugin_object as task_config # Plugin_object_type is a derivative of ``PythonFunctionTask``
Examples of available task plugins include different query-based plugins such as
flytekitplugins.athena.task.AthenaTask
andflytekitplugins.hive.task.HiveTask
, ML tools likeplugins.awssagemaker.flytekitplugins.awssagemaker.training.SagemakerBuiltinAlgorithmsTask
, kubeflow operators likeplugins.kfpytorch.flytekitplugins.kfpytorch.task.PyTorchFunctionTask
andplugins.kftensorflow.flytekitplugins.kftensorflow.task.TensorflowFunctionTask
, and generic plugins likeflytekitplugins.pod.task.PodFunctionTask
which doesn’t integrate with third party tools or services.The task_config is different for every task plugin type. This is filled out by users when they define a task to specify plugin-specific behavior and features. For example, with a query type task plugin, the config might store information related to which database to query.
The plugin_object_type can be used to customize execution behavior and task serialization properties in tandem with the task_config.
- __init__()#
Methods
__init__
()find_pythontask_plugin
(plugin_config_type)Returns a PluginObjectType if found or returns the base PythonFunctionTask
register_pythontask_plugin
(...)Use this method to register a new plugin into Flytekit. Usage ::.