Merge branch 'master' into async-backward-compatability

Author: doddaspk-amzn

CHANGELOG.md

@@ -1,5 +1,26 @@
 # Changelog
 
+## v2.156.0 (2023-05-17)
+
+### Features
+
+ * Partition support for DJLModel using SM Training job
+ * Update run-notebook-test to consider skips failures
+
+### Bug Fixes and Other Changes
+
+ * Update apache airflow and update test requirements
+ * Perform integrity checks for remote function execution
+ * Add p2 instances to integ tests
+ * Fix typo in logging message within ir mixin
+ * double Run create on load_run
+ * Update dtype logic for huggingface backend for new containers
+
+### Documentation Changes
+
+ * Update container version for SKLearn
+ * Add description for parameters in TransformInput
+
 ## v2.155.0 (2023-05-15)
 
 ### Features

VERSION

@@ -1 +1 @@
-2.155.1.dev0
+2.156.1.dev0

doc/frameworks/djl/using_djl.rst

@@ -221,6 +221,31 @@ see the `DJL Serving Documentation on Python Mode. <https://docs.djl.ai/docs/ser
 
 For more information about DJL Serving, see the `DJL Serving documentation. <https://docs.djl.ai/docs/serving/index.html>`_
 
+**************************
+Ahead of time partitioning
+**************************
+
+To optimize the deployment of large models that do not fit in a single GPU, the model’s tensor weights are partitioned at
+runtime and each partition is loaded in individual GPU. But runtime partitioning takes significant amount of time and
+memory on model loading. So, DJLModel offers an ahead of time partitioning capability for DeepSpeed and FasterTransformer
+engines, which lets you partition your model weights and save them before deployment. HuggingFace does not support
+tensor parallelism, so ahead of time partitioning cannot be done for it. In our experiment with GPT-J model, loading
+this model with partitioned checkpoints increased the model loading time by 40%.
+
+`partition` method invokes an Amazon SageMaker Training job to partition the model and upload those partitioned
+checkpoints to S3 bucket. You can either provide your desired S3 bucket to upload the partitioned checkpoints or it will be
+uploaded to the default SageMaker S3 bucket. Please note that this S3 bucket will be remembered for deployment. When you
+call `deploy` method after partition, DJLServing downloads the partitioned model checkpoints directly from the uploaded
+s3 url, if available.
+
+.. code::
+
+    # partitions the model using Amazon Sagemaker Training Job.
+    djl_model.partition("ml.g5.12xlarge")
+
+    predictor = deepspeed_model.deploy("ml.g5.12xlarge",
+                                        initial_instance_count=1)
+
 ***********************
 SageMaker DJL Classes
 ***********************

doc/frameworks/sklearn/using_sklearn.rst

@@ -7,7 +7,7 @@ With Scikit-learn Estimators, you can train and host Scikit-learn models on Amaz
 For information about supported versions of Scikit-learn, see the `AWS documentation <https://docs.aws.amazon.com/sagemaker/latest/dg/sklearn.html>`__.
 We recommend that you use the latest supported version because that's where we focus most of our development efforts.
 
-For more information about the framework, see the `Sciket-Learn <https://github.com/scikit-learn/scikit-learn>`_ repository.
+For more information about the framework, see the `Scikit-Learn <https://github.com/scikit-learn/scikit-learn>`_ repository.
 For general information about using the SageMaker Python SDK, see :ref:`overview:Using the SageMaker Python SDK`.
 
 .. contents::
@@ -31,7 +31,7 @@ To train a Scikit-learn model by using the SageMaker Python SDK:
 Prepare a Scikit-learn Training Script
 ======================================
 
-Your Scikit-learn training script must be a Python 3.6 compatible source file.
+Your Scikit-learn training script must be a Python 3.7 compatible source file.
 
 The training script is similar to a training script you might run outside of SageMaker, but you
 can access useful properties about the training environment through various environment variables.
@@ -140,7 +140,7 @@ directories ('train' and 'test').
 
     sklearn_estimator = SKLearn('sklearn-train.py',
                                 instance_type='ml.m4.xlarge',
-                                framework_version='0.20.0',
+                                framework_version='1.0-1',
                                 hyperparameters = {'epochs': 20, 'batch-size': 64, 'learning-rate': 0.1})
     sklearn_estimator.fit({'train': 's3://my-data-bucket/path/to/my/training/data',
                             'test': 's3://my-data-bucket/path/to/my/test/data'})
@@ -204,7 +204,7 @@ operation.
     # Train my estimator
     sklearn_estimator = SKLearn(entry_point='train_and_deploy.py',
                                 instance_type='ml.m4.xlarge',
-                                framework_version='0.20.0')
+                                framework_version='1.0-1')
     sklearn_estimator.fit('s3://my_bucket/my_training_data/')
 
     # Deploy my estimator to a SageMaker Endpoint and get a Predictor
@@ -478,7 +478,7 @@ The following code sample shows how to do this, using the ``SKLearnModel`` class
     sklearn_model = SKLearnModel(model_data="s3://bucket/model.tar.gz",
                                  role="SageMakerRole",
                                  entry_point="transform_script.py",
-                                 framework_version="0.20.0")
+                                 framework_version="1.0-1")
 
     predictor = sklearn_model.deploy(instance_type="ml.c4.xlarge", initial_instance_count=1)
 

requirements/extras/test_requirements.txt

@@ -12,9 +12,9 @@ awslogs==0.14.0
 black==22.3.0
 stopit==1.1.2
 # Update tox.ini to have correct version of airflow constraints file
-apache-airflow==2.5.1
+apache-airflow==2.6.0
 apache-airflow-providers-amazon==7.2.1
-attrs==22.1.0
+attrs>=23.1.0,<24
 fabric==2.6.0
 requests==2.27.1
 sagemaker-experiments==0.1.35
@@ -23,3 +23,7 @@ pyvis==0.2.1
 pandas>=1.3.5,<1.5
 scikit-learn==1.0.2
 cloudpickle==2.2.1
+scipy==1.7.3
+urllib3==1.26.8
+docker>=5.0.2,<7.0.0
+PyYAML==6.0

setup.py

@@ -47,7 +47,7 @@ def read_requirements(filename):
 
 # Declare minimal set for installation
 required_packages = [
-    "attrs>=20.3.0,<23",
+    "attrs>=23.1.0,<24",
     "boto3>=1.26.131,<2.0",
     "cloudpickle==2.2.1",
     "google-pasta",
@@ -60,7 +60,7 @@ def read_requirements(filename):
     "pandas",
     "pathos",
     "schema",
-    "PyYAML==5.4.1",
+    "PyYAML==6.0",
     "jsonschema",
     "platformdirs",
     "tblib==1.7.0",
@@ -75,7 +75,7 @@ def read_requirements(filename):
 # Meta dependency groups
 extras["all"] = [item for group in extras.values() for item in group]
 # Tests specific dependencies (do not need to be included in 'all')
-extras["test"] = (extras["all"] + read_requirements("requirements/extras/test_requirements.txt"),)
+extras["test"] = (read_requirements("requirements/extras/test_requirements.txt"),)
 
 setup(
     name="sagemaker",