Restore SKLearn FrameworkProcessor via _normalize_args

doc/amazon_sagemaker_processing.rst

@@ -62,6 +62,22 @@ Then you can run a scikit-learn script ``preprocessing.py`` in a processing job.
 
     preprocessing_job_description = sklearn_processor.jobs[-1].describe()
 
+Instead of a single script file, you can submit an entire folder of code for a processing job, and optionally specify additional dependencies to be installed at job start-up by including a `requirements.txt` file.
+
+To do this, also specify the ``source_dir`` parameter in ``.run()`` calls for :class:`SKLearnProcessor` or any other :class:`FrameworkProcessor`-based processor:
+
+.. code:: python
+
+    sklearn_processor.run(
+        code="run.py",  # 'processing/run.py' is the main script to run
+        source_dir="processing",  # Upload the whole contents of 'processing/'
+
+        # If 'processing/requirements.txt' exists, the dependencies it specifies
+        # will be automatically installed before 'run.py' is started.
+        inputs=[...],
+        outputs=[...],
+    )
+
 For an in-depth look, please see the `Scikit-learn Data Processing and Model Evaluation`_ example notebook.
 
 .. _Scikit-learn Data Processing and Model Evaluation: https://github.com/awslabs/amazon-sagemaker-examples/blob/master/sagemaker_processing/scikit_learn_data_processing_and_model_evaluation/scikit_learn_data_processing_and_model_evaluation.ipynb
@@ -220,6 +236,13 @@ For an in-depth look, please see the `Feature Transformation with Spark`_ exampl
 
 .. _Feature Transformation with Spark: https://github.com/awslabs/amazon-sagemaker-examples/blob/master/sagemaker_processing/feature_transformation_with_sagemaker_processing/feature_transformation_with_sagemaker_processing.ipynb
 
+Data Processing with Other Frameworks
+=====================================
+
+:class:`FrameworkProcessor`-based classes are also provided for a range of other ML frameworks: For example PyTorch, TensorFlow, and MXNet.
+
+You can use these to run data processing jobs in pre-built container environments, similarly to model training with :class:`Framework`-based Estimators.
+
 
 Learn More
 ==========
@@ -229,12 +252,18 @@ Processing class documentation
 
 - :class:`sagemaker.processing.Processor`
 - :class:`sagemaker.processing.ScriptProcessor`
+- :class:`sagemaker.processing.FrameworkProcessor`
 - :class:`sagemaker.sklearn.processing.SKLearnProcessor`
 - :class:`sagemaker.spark.processing.PySparkProcessor`
 - :class:`sagemaker.spark.processing.SparkJarProcessor`
 - :class:`sagemaker.processing.ProcessingInput`
 - :class:`sagemaker.processing.ProcessingOutput`
 - :class:`sagemaker.processing.ProcessingJob`
+- :class:`sagemaker.huggingface.processing.HuggingFaceProcessor`
+- :class:`sagemaker.mxnet.processing.MXNetProcessor`
+- :class:`sagemaker.pytorch.processing.PyTorchProcessor`
+- :class:`sagemaker.tensorflow.processing.TensorFlowProcessor`
+- :class:`sagemaker.xgboost.processing.XGBoostProcessor`
 
 
 Further documentation

doc/frameworks/huggingface/sagemaker.huggingface.rst

@@ -17,10 +17,18 @@ Hugging Face Model
     :undoc-members:
     :show-inheritance:
 
-HuggingFace Predictor
----------------------
+Hugging Face Predictor
+----------------------
 
 .. autoclass:: sagemaker.huggingface.model.HuggingFacePredictor
     :members:
     :undoc-members:
     :show-inheritance:
+
+Hugging Face Processor
+----------------------
+
+.. autoclass:: sagemaker.huggingface.processing.HuggingFaceProcessor
+    :members:
+    :undoc-members:
+    :show-inheritance:

doc/frameworks/mxnet/sagemaker.mxnet.rst

@@ -25,3 +25,11 @@ MXNet Predictor
     :members:
     :undoc-members:
     :show-inheritance:
+
+MXNet Processor
+---------------------------
+
+.. autoclass:: sagemaker.mxnet.processing.MXNetProcessor
+    :members:
+    :undoc-members:
+    :show-inheritance:

doc/frameworks/pytorch/sagemaker.pytorch.rst

@@ -24,3 +24,11 @@ PyTorch Predictor
     :members:
     :undoc-members:
     :show-inheritance:
+
+PyTorch Processor
+-----------------
+
+.. autoclass:: sagemaker.pytorch.processing.PyTorchProcessor
+    :members:
+    :undoc-members:
+    :show-inheritance:

doc/frameworks/tensorflow/sagemaker.tensorflow.rst

@@ -25,3 +25,11 @@ TensorFlow Serving Predictor
     :members:
     :undoc-members:
     :show-inheritance:
+
+TensorFlow Processor
+--------------------
+
+.. autoclass:: sagemaker.tensorflow.processing.TensorFlowProcessor
+    :members:
+    :undoc-members:
+    :show-inheritance:

doc/frameworks/xgboost/xgboost.rst

@@ -16,3 +16,8 @@ The Amazon SageMaker XGBoost open source framework algorithm.
     :members:
     :undoc-members:
     :show-inheritance:
+
+.. autoclass:: sagemaker.xgboost.processing.XGBoostProcessor
+    :members:
+    :undoc-members:
+    :show-inheritance:

src/sagemaker/huggingface/processing.py

@@ -105,6 +105,8 @@ def _create_estimator(
         source_dir=None,
         dependencies=None,
         git_config=None,
+        output_kms_key=None,
+        volume_kms_key=None,
     ):
         """Override default estimator factory function for HuggingFace's different parameters
 
@@ -121,6 +123,8 @@ def _create_estimator(
             dependencies=dependencies,
             git_config=git_config,
             code_location=self.code_location,
+            output_kms_key=output_kms_key,
+            volume_kms_key=volume_kms_key,
             enable_network_isolation=False,
             image_uri=self.image_uri,
             role=self.role,

src/sagemaker/mxnet/processing.py

@@ -28,17 +28,17 @@ class MXNetProcessor(FrameworkProcessor):
 
     def __init__(
         self,
-        framework_version,  # New arg
+        framework_version,
         role,
         instance_count,
         instance_type,
-        py_version="py3",  # New kwarg
+        py_version="py3",
         image_uri=None,
         command=None,
         volume_size_in_gb=30,
         volume_kms_key=None,
         output_kms_key=None,
-        code_location=None,  # New arg
+        code_location=None,
         max_runtime_in_seconds=None,
         base_job_name=None,
         sagemaker_session=None,