sagemaker-python-sdk

Documentation Changes

• remove labels from issue templates

Bug Fixes and Other Changes

• account for EI and version-based ECR repo naming in serving_image_uri()

Documentation Changes

• correct broken AutoML API documentation link
• fix MXNet version lists

Bug Fixes and Other Changes

• disable Debugger defaults in unsupported regions
• modify session and kms_utils to check for S3 bucket before creation
• update docker-compose and PyYAML dependencies
• enable smdebug for Horovod (MPI) training setup
• create lib dir for dependencies safely (only if it doesn't exist yet).
• create the correct session for MultiDataModel

Documentation Changes

• update links to the local mode notebooks examples.
• Remove outdated badges from README
• update links to TF notebook examples to link to script mode examples.
• clean up headings, verb tenses, names, etc. in MXNet overview
• Update SageMaker operator Helm chart installation guide

Testing and Release Infrastructure

• choose faster notebook for notebook PR build
• properly fail PR build if has-matching-changes fails
• properly fail PR build if has-matching-changes fails

Bug fixes and other changes

• do not use script for TFS when entry_point is not provided
• remove usage of pkg_resources
• update py2 warning message since python 2 is deprecated
• cleanup experiments, trials, and trial components in integ tests

Documentation changes

• add additional information to Transformer class transform function doc string

Bug fixes and other changes

• Append serving to model framework name for PyTorch, MXNet, and TensorFlow

Bug fixes and other changes

• Use serving_image_uri for Airflow

Documentation changes

• revise Processing docstrings for formatting and class links
• Add processing readthedocs

Bug fixes and other changes

• Remove version number from default version comment
• remove remaining instances of python-dateutil pin
• upgrade boto3 and remove python-dateutil pin

Documentation changes

• Add issue templates and configure issue template chooser
• Update error type in delete_endpoint docstring
• add version requirement for using "requirements.txt" when serving an MXNet model
• update container dependency versions for MXNet and PyTorch
• Update supported versions of PyTorch

Bug fixes and other changes

• ignore private Automatic Model Tuning hyperparameter when attaching AlgorithmEstimator

Documentation changes

• add Debugger API docs

Bug fixes and other changes

• add tests to quick canary
• honor 'wait' flag when updating endpoint
• add default framework version warning message in Model classes
• Adding role arn explanation for sagemaker role
• allow predictor to be returned from AutoML.deploy()
• add PR checklist item about unique_name_from_base()
• use unique_name_from_base for multi-algo tuning test
• update copyright year in license header

Documentation changes

• add version requirement for using "requirement.txt" when serving a PyTorch model
• add SageMaker Debugger overview
• clarify requirements.txt usage for Chainer, MXNet, and Scikit-learn
• change "associate" to "create" for OpenID connector
• fix typo and improve clarity on installing packages via "requirements.txt"

Bug fixes and other changes

• fix PyTorchModel deployment crash on Windows
• make PyTorch empty framework_version warning include the latest PyTorch version

Features

• allow disabling debugger_hook_config

Bug fixes and other changes

• relax urllib3 and requests restrictions.
• Add uri as return statement for upload_string_as_file_body
• refactor logic in fw_utils and fill in docstrings
• increase poll from 5 to 30 for DescribeEndpoint lambda.
• fix test_auto_ml tests for regions without ml.c4.xlarge hosts.
• fix test_processing for regions without m4.xlarge instances.
• reduce test's describe frequency to eliminate throttling error.
• Increase number of retries when describing an endpoint since tf-2.0 has larger images and takes longer to start.

Documentation changes

• generalize Model Monitor documentation from SageMaker Studio tutorial

Features

• Add support for TF-2.0.0.
• create ProcessingJob from ARN and from name

Bug fixes and other changes

• Make tf tests tf-1.15 and tf-2.0 compatible.

Documentation changes

• add Model Monitor documentation
• add link to Amazon algorithm estimator parent class to clarify **kwargs

Bug fixes and other changes

• use name_from_base in auto_ml.py but unique_name_from_base in tests.
• make test's custom bucket include region and account name.
• add Keras to the list of Neo-supported frameworks

Documentation changes

• add link to parent classes to clarify **kwargs
• add link to framework-related parent classes to clarify **kwargs

Features

• allow setting the default bucket in Session

Bug fixes and other changes

• set integration test parallelization to 512
• shorten base job name to avoid collision
• multi model integration test to create ECR repo with unique names to allow independent parallel executions

Bug fixes and other changes

• Revert "feature: allow setting the default bucket in Session (#1168)"

Documentation changes

• add AutoML README
• add missing classes to API docs

Features

• support Multi-Model endpoints

Bug fixes and other changes

• update PR template with items about tests, regional endpoints, and API docs

Bug fixes and other changes

• modify schedule cleanup to abide by latest validations
• lower log level when getting execution role from a SageMaker Notebook
• Fix "ValueError: too many values to unpack (expected 2)" is occurred in windows local mode
• allow ModelMonitor and Processor to take IAM role names (in addition to ARNs)

Documentation changes

• mention that the entry_point needs to be named inference.py for tfs

Bug fixes and other changes

• create auto ml job for tests that based on existing job
• fixing py2 support for latest TF version
• fix tags in deploy call for generic estimators
• make multi algo integration test assertion less specific

Features

• add support for TF 1.15.0, PyTorch 1.3.1 and MXNet 1.6rc0.
• add S3Downloader.list(s3_uri) functionality
• introduce SageMaker AutoML
• wrap up Processing feature
• add a few minor features to Model Monitoring
• add enable_sagemaker_metrics flag
• Amazon SageMaker Model Monitoring
• add utils.generate_tensorboard_url function
• Add jobs list to Estimator

Bug fixes and other changes

• remove unnecessary boto model files
• update boto version to >=1.10.32
• correct Debugger tests
• fix bug in monitor.attach() for empty network_config
• Import smdebug_rulesconfig from PyPI
• bump the version to 1.45.0 (publishes 1.46.0) for re:Invent-2019
• correct AutoML imports and expose current_job_name
• correct Model Monitor eu-west-3 image name.
• use DLC prod images
• remove unused env variable for Model Monitoring
• aws model update
• rename get_debugger_artifacts to latest_job_debugger_artifacts
• remove retain flag from update_endpoint
• correct S3Downloader behavior
• consume smdebug_ruleconfig .whl for ITs
• disable DebuggerHook and Rules for TF distributions
• incorporate smdebug_ruleconfigs pkg until availability in PyPI
• remove pre/post scripts per latest validations
• update rules_config .whl
• remove py_version from SKLearnProcessor
• AutoML improvements
• stop overwriting custom rules volume and type
• fix tests due to latest server-side validations
• Minor processing changes
• minor processing changes (instance_count + docs)
• update api to latest
• Eureka master
• Add support for xgboost version 0.90-2
• SageMaker Debugger revision
• Add support for SageMaker Debugger [WIP]
• Fix linear learner crash when num_class is string and predict type is multiclass_classifier
• Additional Processing Jobs integration tests
• Migrate to updated Processing Jobs API
• Processing Jobs revision round 2
• Processing Jobs revision
• remove instance_pools parameter from tuner
• Multi-Algorithm Hyperparameter Tuning Support
• Import Processors in init files
• Remove SparkML Processors and corresponding unit tests
• Processing Jobs Python SDK support