def test_warn_if_parameter_server_with_local_multi_gpu(caplog): instance_type = "local_gpu" distribution = {"parameter_server": {"enabled": True}} fw_utils.warn_if_parameter_server_with_multi_gpu( training_instance_type=instance_type, distribution=distribution) assert fw_utils.PARAMETER_SERVER_MULTI_GPU_WARNING in caplog.text
def test_warn_if_parameter_server_with_multi_gpu(caplog): train_instance_type = "ml.p2.8xlarge" distributions = {"parameter_server": {"enabled": True}} fw_utils.warn_if_parameter_server_with_multi_gpu( training_instance_type=train_instance_type, distributions=distributions) assert fw_utils.PARAMETER_SERVER_MULTI_GPU_WARNING in caplog.text
def __init__( self, py_version, entry_point, transformers_version=None, tensorflow_version=None, pytorch_version=None, source_dir=None, hyperparameters=None, image_uri=None, distribution=None, **kwargs ): """This ``Estimator`` executes a HuggingFace script in a managed execution environment. The managed HuggingFace environment is an Amazon-built Docker container that executes functions defined in the supplied ``entry_point`` Python script within a SageMaker Training Job. Training is started by calling :meth:`~sagemaker.amazon.estimator.Framework.fit` on this Estimator. Args: py_version (str): Python version you want to use for executing your model training code. Defaults to ``None``. Required unless ``image_uri`` is provided. List of supported versions: https://github.com/aws/sagemaker-python-sdk#huggingface-sagemaker-estimators entry_point (str): Path (absolute or relative) to the Python source file which should be executed as the entry point to training. If ``source_dir`` is specified, then ``entry_point`` must point to a file located at the root of ``source_dir``. transformers_version (str): Transformers version you want to use for executing your model training code. Defaults to ``None``. Required unless ``image_uri`` is provided. List of supported versions: https://github.com/aws/sagemaker-python-sdk#huggingface-sagemaker-estimators. tensorflow_version (str): TensorFlow version you want to use for executing your model training code. Defaults to ``None``. Required unless ``pytorch_version`` is provided. List of supported versions: https://github.com/aws/sagemaker-python-sdk#huggingface-sagemaker-estimators. pytorch_version (str): PyTorch version you want to use for executing your model training code. Defaults to ``None``. Required unless ``tensorflow_version`` is provided. List of supported versions: https://github.com/aws/sagemaker-python-sdk#huggingface-sagemaker-estimators. source_dir (str): Path (absolute, relative or an S3 URI) to a directory with any other training source code dependencies aside from the entry point file (default: None). If ``source_dir`` is an S3 URI, it must point to a tar.gz file. Structure within this directory are preserved when training on Amazon SageMaker. hyperparameters (dict): Hyperparameters that will be used for training (default: None). The hyperparameters are made accessible as a dict[str, str] to the training code on SageMaker. For convenience, this accepts other types for keys and values, but ``str()`` will be called to convert them before training. image_uri (str): If specified, the estimator will use this image for training and hosting, instead of selecting the appropriate SageMaker official image based on framework_version and py_version. It can be an ECR url or dockerhub image and tag. Examples: * ``123412341234.dkr.ecr.us-west-2.amazonaws.com/my-custom-image:1.0`` * ``custom-image:latest`` If ``framework_version`` or ``py_version`` are ``None``, then ``image_uri`` is required. If also ``None``, then a ``ValueError`` will be raised. distribution (dict): A dictionary with information on how to run distributed training (default: None). Currently, the following are supported: distributed training with parameter servers, SageMaker Distributed (SMD) Data and Model Parallelism, and MPI. SMD Model Parallelism can only be used with MPI. To enable parameter server use the following setup: .. code:: python { "parameter_server": { "enabled": True } } To enable MPI: .. code:: python { "mpi": { "enabled": True } } To enable SMDistributed Data Parallel or Model Parallel: .. code:: python { "smdistributed": { "dataparallel": { "enabled": True }, "modelparallel": { "enabled": True, "parameters": {} } } } **kwargs: Additional kwargs passed to the :class:`~sagemaker.estimator.Framework` constructor. .. tip:: You can find additional parameters for initializing this class at :class:`~sagemaker.estimator.Framework` and :class:`~sagemaker.estimator.EstimatorBase`. """ self.framework_version = transformers_version self.py_version = py_version self.tensorflow_version = tensorflow_version self.pytorch_version = pytorch_version self._validate_args(image_uri=image_uri) if distribution is not None: instance_type = renamed_kwargs( "train_instance_type", "instance_type", kwargs.get("instance_type"), kwargs ) base_framework_name = "tensorflow" if tensorflow_version is not None else "pytorch" base_framework_version = ( tensorflow_version if tensorflow_version is not None else pytorch_version ) validate_smdistributed( instance_type=instance_type, framework_name=base_framework_name, framework_version=base_framework_version, py_version=self.py_version, distribution=distribution, image_uri=image_uri, ) warn_if_parameter_server_with_multi_gpu( training_instance_type=instance_type, distribution=distribution ) if "enable_sagemaker_metrics" not in kwargs: kwargs["enable_sagemaker_metrics"] = True super(HuggingFace, self).__init__( entry_point, source_dir, hyperparameters, image_uri=image_uri, **kwargs ) self.distribution = distribution or {}
def __init__(self, py_version=None, framework_version=None, model_dir=None, image_uri=None, distribution=None, **kwargs): """Initialize a ``TensorFlow`` estimator. Args: py_version (str): Python version you want to use for executing your model training code. Defaults to ``None``. Required unless ``image_uri`` is provided. framework_version (str): TensorFlow version you want to use for executing your model training code. Defaults to ``None``. Required unless ``image_uri`` is provided. List of supported versions: https://github.com/aws/sagemaker-python-sdk#tensorflow-sagemaker-estimators. model_dir (str): S3 location where the checkpoint data and models can be exported to during training (default: None). It will be passed in the training script as one of the command line arguments. If not specified, one is provided based on your training configuration: * *distributed training with SMDistributed or MPI with Horovod* - ``/opt/ml/model`` * *single-machine training or distributed training without MPI* - \ ``s3://{output_path}/model`` * *Local Mode with local sources (file:// instead of s3://)* - \ ``/opt/ml/shared/model`` To disable having ``model_dir`` passed to your training script, set ``model_dir=False``. image_uri (str): If specified, the estimator will use this image for training and hosting, instead of selecting the appropriate SageMaker official image based on framework_version and py_version. It can be an ECR url or dockerhub image and tag. Examples: 123.dkr.ecr.us-west-2.amazonaws.com/my-custom-image:1.0 custom-image:latest. If ``framework_version`` or ``py_version`` are ``None``, then ``image_uri`` is required. If also ``None``, then a ``ValueError`` will be raised. distribution (dict): A dictionary with information on how to run distributed training (default: None). Currently, the following are supported: distributed training with parameter servers, SageMaker Distributed (SMD) Data and Model Parallelism, and MPI. SMD Model Parallelism can only be used with MPI. To enable parameter server use the following setup: .. code:: python { "parameter_server": { "enabled": True } } To enable MPI: .. code:: python { "mpi": { "enabled": True } } To enable SMDistributed Data Parallel or Model Parallel: .. code:: python { "smdistributed": { "dataparallel": { "enabled": True }, "modelparallel": { "enabled": True, "parameters": {} } } } **kwargs: Additional kwargs passed to the Framework constructor. .. tip:: You can find additional parameters for initializing this class at :class:`~sagemaker.estimator.Framework` and :class:`~sagemaker.estimator.EstimatorBase`. """ distribution = renamed_kwargs("distributions", "distribution", distribution, kwargs) instance_type = renamed_kwargs("train_instance_type", "instance_type", kwargs.get("instance_type"), kwargs) fw.validate_version_or_image_args(framework_version, py_version, image_uri) if py_version == "py2": logger.warning( fw.python_deprecation_warning(self._framework_name, defaults.LATEST_PY2_VERSION)) self.framework_version = framework_version self.py_version = py_version self.instance_type = instance_type if distribution is not None: fw.warn_if_parameter_server_with_multi_gpu( training_instance_type=instance_type, distribution=distribution) fw.validate_smdistributed( instance_type=instance_type, framework_name=self._framework_name, framework_version=framework_version, py_version=py_version, distribution=distribution, image_uri=image_uri, ) if "enable_sagemaker_metrics" not in kwargs: # enable sagemaker metrics for TF v1.15 or greater: if framework_version and version.Version( framework_version) >= version.Version("1.15"): kwargs["enable_sagemaker_metrics"] = True super(TensorFlow, self).__init__(image_uri=image_uri, **kwargs) self.model_dir = model_dir self.distribution = distribution or {} self._validate_args(py_version=py_version)
def __init__(self, training_steps=None, evaluation_steps=None, checkpoint_path=None, py_version=None, framework_version=None, model_dir=None, requirements_file="", image_name=None, script_mode=False, distributions=None, **kwargs): """Initialize a ``TensorFlow`` estimator. Args: training_steps (int): Perform this many steps of training. `None`, the default means train forever. evaluation_steps (int): Perform this many steps of evaluation. `None`, the default means that evaluation runs until input from eval_input_fn is exhausted (or another exception is raised). checkpoint_path (str): Identifies S3 location where checkpoint data during model training can be saved (default: None). For distributed model training, this parameter is required. py_version (str): Python version you want to use for executing your model training code (default: 'py2'). framework_version (str): TensorFlow version you want to use for executing your model training code. If not specified, this will default to 1.11. model_dir (str): S3 location where the checkpoint data and models can be exported to during training (default: None). It will be passed in the training script as one of the command line arguments. If not specified, one is provided based on your training configuration: * *distributed training with MPI* - ``/opt/ml/model`` * *single-machine training or distributed training without MPI* - \ ``s3://{output_path}/model`` * *Local Mode with local sources (file:// instead of s3://)* - \ ``/opt/ml/shared/model`` requirements_file (str): Path to a ``requirements.txt`` file (default: ''). The path should be within and relative to ``source_dir``. Details on the format can be found in the Pip User Guide: <https://pip.pypa.io/en/stable/reference/pip_install/#requirements-file-format> image_name (str): If specified, the estimator will use this image for training and hosting, instead of selecting the appropriate SageMaker official image based on framework_version and py_version. It can be an ECR url or dockerhub image and tag. Examples: 123.dkr.ecr.us-west-2.amazonaws.com/my-custom-image:1.0 custom-image:latest. script_mode (bool): If set to True will the estimator will use the Script Mode containers (default: False). This will be ignored if py_version is set to 'py3'. distributions (dict): A dictionary with information on how to run distributed training (default: None). Currently we support distributed training with parameter servers and MPI. To enable parameter server use the following setup: .. code:: python { 'parameter_server': { 'enabled': True } } To enable MPI: .. code:: python { 'mpi': { 'enabled': True } } **kwargs: Additional kwargs passed to the Framework constructor. .. tip:: You can find additional parameters for initializing this class at :class:`~sagemaker.estimator.Framework` and :class:`~sagemaker.estimator.EstimatorBase`. """ if framework_version is None: logger.warning( fw.empty_framework_version_warning(defaults.TF_VERSION, self.LATEST_VERSION)) self.framework_version = framework_version or defaults.TF_VERSION if not py_version: py_version = "py3" if self._only_python_3_supported() else "py2" if py_version == "py2": logger.warning( fw.python_deprecation_warning(self.__framework_name__, defaults.LATEST_PY2_VERSION)) if distributions is not None: logger.warning( fw.parameter_v2_rename_warning("distribution", distributions)) train_instance_type = kwargs.get("train_instance_type") fw.warn_if_parameter_server_with_multi_gpu( training_instance_type=train_instance_type, distributions=distributions) if "enable_sagemaker_metrics" not in kwargs: # enable sagemaker metrics for TF v1.15 or greater: if fw.is_version_equal_or_higher([1, 15], self.framework_version): kwargs["enable_sagemaker_metrics"] = True super(TensorFlow, self).__init__(image_name=image_name, **kwargs) self.checkpoint_path = checkpoint_path self.py_version = py_version self.training_steps = training_steps self.evaluation_steps = evaluation_steps self.model_dir = model_dir self.script_mode = script_mode self.distributions = distributions or {} self._validate_args( py_version=py_version, script_mode=script_mode, framework_version=self.framework_version, training_steps=training_steps, evaluation_steps=evaluation_steps, requirements_file=requirements_file, checkpoint_path=checkpoint_path, ) self._validate_requirements_file(requirements_file) self.requirements_file = requirements_file
def __init__(self, entry_point, framework_version=None, py_version=None, source_dir=None, hyperparameters=None, image_uri=None, distribution=None, **kwargs): """This ``Estimator`` executes an PyTorch script in a managed PyTorch execution environment, within a SageMaker Training Job. The managed PyTorch environment is an Amazon-built Docker container that executes functions defined in the supplied ``entry_point`` Python script. Training is started by calling :meth:`~sagemaker.amazon.estimator.Framework.fit` on this Estimator. After training is complete, calling :meth:`~sagemaker.amazon.estimator.Framework.deploy` creates a hosted SageMaker endpoint and returns an :class:`~sagemaker.amazon.pytorch.model.PyTorchPredictor` instance that can be used to perform inference against the hosted model. Technical documentation on preparing PyTorch scripts for SageMaker training and using the PyTorch Estimator is available on the project home-page: https://github.com/aws/sagemaker-python-sdk Args: entry_point (str): Path (absolute or relative) to the Python source file which should be executed as the entry point to training. If ``source_dir`` is specified, then ``entry_point`` must point to a file located at the root of ``source_dir``. framework_version (str): PyTorch version you want to use for executing your model training code. Defaults to ``None``. Required unless ``image_uri`` is provided. List of supported versions: https://github.com/aws/sagemaker-python-sdk#pytorch-sagemaker-estimators. py_version (str): Python version you want to use for executing your model training code. One of 'py2' or 'py3'. Defaults to ``None``. Required unless ``image_uri`` is provided. source_dir (str): Path (absolute, relative or an S3 URI) to a directory with any other training source code dependencies aside from the entry point file (default: None). If ``source_dir`` is an S3 URI, it must point to a tar.gz file. Structure within this directory are preserved when training on Amazon SageMaker. hyperparameters (dict): Hyperparameters that will be used for training (default: None). The hyperparameters are made accessible as a dict[str, str] to the training code on SageMaker. For convenience, this accepts other types for keys and values, but ``str()`` will be called to convert them before training. image_uri (str): If specified, the estimator will use this image for training and hosting, instead of selecting the appropriate SageMaker official image based on framework_version and py_version. It can be an ECR url or dockerhub image and tag. Examples: * ``123412341234.dkr.ecr.us-west-2.amazonaws.com/my-custom-image:1.0`` * ``custom-image:latest`` If ``framework_version`` or ``py_version`` are ``None``, then ``image_uri`` is required. If also ``None``, then a ``ValueError`` will be raised. distribution (dict): A dictionary with information on how to run distributed training (default: None). Currently, the following are supported: distributed training with parameter servers, SageMaker Distributed (SMD) Data and Model Parallelism, and MPI. SMD Model Parallelism can only be used with MPI. To enable parameter server use the following setup: .. code:: python { "parameter_server": { "enabled": True } } To enable MPI: .. code:: python { "mpi": { "enabled": True } } To enable SMDistributed Data Parallel or Model Parallel: .. code:: python { "smdistributed": { "dataparallel": { "enabled": True }, "modelparallel": { "enabled": True, "parameters": {} } } } **kwargs: Additional kwargs passed to the :class:`~sagemaker.estimator.Framework` constructor. .. tip:: You can find additional parameters for initializing this class at :class:`~sagemaker.estimator.Framework` and :class:`~sagemaker.estimator.EstimatorBase`. """ validate_version_or_image_args(framework_version, py_version, image_uri) if py_version == "py2": logger.warning( python_deprecation_warning(self._framework_name, defaults.LATEST_PY2_VERSION)) self.framework_version = framework_version self.py_version = py_version if distribution is not None: instance_type = renamed_kwargs("train_instance_type", "instance_type", kwargs.get("instance_type"), kwargs) validate_smdistributed( instance_type=instance_type, framework_name=self._framework_name, framework_version=framework_version, py_version=py_version, distribution=distribution, image_uri=image_uri, ) warn_if_parameter_server_with_multi_gpu( training_instance_type=instance_type, distribution=distribution) if "enable_sagemaker_metrics" not in kwargs: # enable sagemaker metrics for PT v1.3 or greater: if self.framework_version and Version( self.framework_version) >= Version("1.3"): kwargs["enable_sagemaker_metrics"] = True super(PyTorch, self).__init__(entry_point, source_dir, hyperparameters, image_uri=image_uri, **kwargs) self.distribution = distribution or {}
def __init__(self, entry_point, source_dir=None, hyperparameters=None, py_version="py2", framework_version=None, image_name=None, distributions=None, **kwargs): """This ``Estimator`` executes an MXNet script in a managed MXNet execution environment, within a SageMaker Training Job. The managed MXNet environment is an Amazon-built Docker container that executes functions defined in the supplied ``entry_point`` Python script. Training is started by calling :meth:`~sagemaker.amazon.estimator.Framework.fit` on this Estimator. After training is complete, calling :meth:`~sagemaker.amazon.estimator.Framework.deploy` creates a hosted SageMaker endpoint and returns an :class:`~sagemaker.amazon.mxnet.model.MXNetPredictor` instance that can be used to perform inference against the hosted model. Technical documentation on preparing MXNet scripts for SageMaker training and using the MXNet Estimator is available on the project home-page: https://github.com/aws/sagemaker-python-sdk Args: entry_point (str): Path (absolute or relative) to the Python source file which should be executed as the entry point to training. This should be compatible with either Python 2.7 or Python 3.5. source_dir (str): Path (absolute or relative) to a directory with any other training source code dependencies aside from the entry point file (default: None). Structure within this directory are preserved when training on Amazon SageMaker. hyperparameters (dict): Hyperparameters that will be used for training (default: None). The hyperparameters are made accessible as a dict[str, str] to the training code on SageMaker. For convenience, this accepts other types for keys and values, but ``str()`` will be called to convert them before training. py_version (str): Python version you want to use for executing your model training code (default: 'py2'). One of 'py2' or 'py3'. framework_version (str): MXNet version you want to use for executing your model training code. List of supported versions https://github.com/aws/sagemaker-python-sdk#mxnet-sagemaker-estimators. If not specified, this will default to 1.2.1. image_name (str): If specified, the estimator will use this image for training and hosting, instead of selecting the appropriate SageMaker official image based on framework_version and py_version. It can be an ECR url or dockerhub image and tag. Examples: * ``123412341234.dkr.ecr.us-west-2.amazonaws.com/my-custom-image:1.0`` * ``custom-image:latest`` distributions (dict): A dictionary with information on how to run distributed training (default: None). To have parameter servers launched for training, set this value to be ``{'parameter_server': {'enabled': True}}``. **kwargs: Additional kwargs passed to the :class:`~sagemaker.estimator.Framework` constructor. .. tip:: You can find additional parameters for initializing this class at :class:`~sagemaker.estimator.Framework` and :class:`~sagemaker.estimator.EstimatorBase`. """ if framework_version is None: logger.warning( empty_framework_version_warning(defaults.MXNET_VERSION, self.LATEST_VERSION)) self.framework_version = framework_version or defaults.MXNET_VERSION if "enable_sagemaker_metrics" not in kwargs: # enable sagemaker metrics for MXNet v1.6 or greater: if is_version_equal_or_higher([1, 6], self.framework_version): kwargs["enable_sagemaker_metrics"] = True super(MXNet, self).__init__(entry_point, source_dir, hyperparameters, image_name=image_name, **kwargs) if py_version == "py2": logger.warning( python_deprecation_warning(self.__framework_name__, defaults.LATEST_PY2_VERSION)) if distributions is not None: train_instance_type = kwargs.get("train_instance_type") warn_if_parameter_server_with_multi_gpu( training_instance_type=train_instance_type, distributions=distributions) self.py_version = py_version self._configure_distribution(distributions)
def __init__( self, py_version, entry_point, transformers_version=None, tensorflow_version=None, pytorch_version=None, source_dir=None, hyperparameters=None, image_uri=None, distribution=None, compiler_config=None, **kwargs, ): """This estimator runs a Hugging Face training script in a SageMaker training environment. The estimator initiates the SageMaker-managed Hugging Face environment by using the pre-built Hugging Face Docker container and runs the Hugging Face training script that user provides through the ``entry_point`` argument. After configuring the estimator class, use the class method :meth:`~sagemaker.amazon.estimator.Framework.fit()` to start a training job. Args: py_version (str): Python version you want to use for executing your model training code. Defaults to ``None``. Required unless ``image_uri`` is provided. If using PyTorch, the current supported version is ``py36``. If using TensorFlow, the current supported version is ``py37``. entry_point (str): Path (absolute or relative) to the Python source file which should be executed as the entry point to training. If ``source_dir`` is specified, then ``entry_point`` must point to a file located at the root of ``source_dir``. transformers_version (str): Transformers version you want to use for executing your model training code. Defaults to ``None``. Required unless ``image_uri`` is provided. The current supported version is ``4.6.1``. tensorflow_version (str): TensorFlow version you want to use for executing your model training code. Defaults to ``None``. Required unless ``pytorch_version`` is provided. The current supported version is ``2.4.1``. pytorch_version (str): PyTorch version you want to use for executing your model training code. Defaults to ``None``. Required unless ``tensorflow_version`` is provided. The current supported versions are ``1.7.1`` and ``1.6.0``. source_dir (str): Path (absolute, relative or an S3 URI) to a directory with any other training source code dependencies aside from the entry point file (default: None). If ``source_dir`` is an S3 URI, it must point to a tar.gz file. Structure within this directory are preserved when training on Amazon SageMaker. hyperparameters (dict): Hyperparameters that will be used for training (default: None). The hyperparameters are made accessible as a dict[str, str] to the training code on SageMaker. For convenience, this accepts other types for keys and values, but ``str()`` will be called to convert them before training. image_uri (str): If specified, the estimator will use this image for training and hosting, instead of selecting the appropriate SageMaker official image based on framework_version and py_version. It can be an ECR url or dockerhub image and tag. Examples: * ``123412341234.dkr.ecr.us-west-2.amazonaws.com/my-custom-image:1.0`` * ``custom-image:latest`` If ``framework_version`` or ``py_version`` are ``None``, then ``image_uri`` is required. If also ``None``, then a ``ValueError`` will be raised. distribution (dict): A dictionary with information on how to run distributed training (default: None). Currently, the following are supported: distributed training with parameter servers, SageMaker Distributed (SMD) Data and Model Parallelism, and MPI. SMD Model Parallelism can only be used with MPI. To enable parameter server use the following setup: .. code:: python { "parameter_server": { "enabled": True } } To enable MPI: .. code:: python { "mpi": { "enabled": True } } To enable SMDistributed Data Parallel or Model Parallel: .. code:: python { "smdistributed": { "dataparallel": { "enabled": True }, "modelparallel": { "enabled": True, "parameters": {} } } } compiler_config (:class:`~sagemaker.huggingface.TrainingCompilerConfig`): Configures SageMaker Training Compiler to accelerate training. **kwargs: Additional kwargs passed to the :class:`~sagemaker.estimator.Framework` constructor. .. tip:: You can find additional parameters for initializing this class at :class:`~sagemaker.estimator.Framework` and :class:`~sagemaker.estimator.EstimatorBase`. """ self.framework_version = transformers_version self.py_version = py_version self.tensorflow_version = tensorflow_version self.pytorch_version = pytorch_version self._validate_args(image_uri=image_uri) instance_type = renamed_kwargs( "train_instance_type", "instance_type", kwargs.get("instance_type"), kwargs ) base_framework_name = "tensorflow" if tensorflow_version is not None else "pytorch" base_framework_version = ( tensorflow_version if tensorflow_version is not None else pytorch_version ) if distribution is not None: validate_smdistributed( instance_type=instance_type, framework_name=base_framework_name, framework_version=base_framework_version, py_version=self.py_version, distribution=distribution, image_uri=image_uri, ) warn_if_parameter_server_with_multi_gpu( training_instance_type=instance_type, distribution=distribution ) if "enable_sagemaker_metrics" not in kwargs: kwargs["enable_sagemaker_metrics"] = True kwargs["py_version"] = self.py_version super(HuggingFace, self).__init__( entry_point, source_dir, hyperparameters, image_uri=image_uri, **kwargs ) if compiler_config is not None: if not isinstance(compiler_config, TrainingCompilerConfig): error_string = ( f"Expected instance of type {TrainingCompilerConfig}" f"for argument compiler_config. " f"Instead got {type(compiler_config)}" ) raise ValueError(error_string) if compiler_config: compiler_config.validate( image_uri=image_uri, instance_type=instance_type, distribution=distribution, ) self.distribution = distribution or {} self.compiler_config = compiler_config