PyTorchDistribution Classe

Configuration de la distribution PyTorch.

Héritage

azure.ai.ml.entities._job.distribution.DistributionConfiguration

PyTorchDistribution

Constructeur

PyTorchDistribution(*, process_count_per_instance: int | None = None, **kwargs: Any)

Paramètres de mot clé uniquement

Nom	Description
process_count_per_instance	Optional[int] nombre de processus par nœud.

Exemples

Configuration d’un CommandComponent avec pyTorchDistribution.

   from azure.ai.ml import PyTorchDistribution
   from azure.ai.ml.entities import CommandComponent

   component = CommandComponent(
       name="microsoftsamples_torch",
       description="This is the PyTorch command component",
       inputs={
           "component_in_number": {"description": "A number", "type": "number", "default": 10.99},
           "component_in_path": {"description": "A path", "type": "uri_folder"},
       },
       outputs={"component_out_path": {"type": "uri_folder"}},
       command="echo Hello World & echo ${{inputs.component_in_number}} & echo ${{inputs.component_in_path}} "
       "& echo ${{outputs.component_out_path}}",
       environment="AzureML-sklearn-1.0-ubuntu20.04-py38-cpu:33",
       distribution=PyTorchDistribution(
           process_count_per_instance=2,
       ),
       instance_count=2,
   )

Variables

Nom	Description
type	str Spécifie le type de distribution. Définissez automatiquement sur « pytorch » pour cette classe.