NewYour coding agent can read the release notes before it upgrades.Set up the MCP server →
PyPI · #1039 most downloaded on PyPI
Open source library for training and deploying models on Amazon SageMaker.
Last release 6 days ago
24 Sep 2026
Ships on a steady schedule
a new release about every 2 weeks
Nearly every release is documented
notes for 59 of the last 60 stable releases
3 versions withdrawn
withdrawn after publishing
9 years old
692 releases · first in 2017
pin pytest version <6.1.0 to avoid pytest-rerunfailures breaking changes
add inferentia pytorch inference container config
One column per quarter.
allow kms encryption upload for processing
update spark image_uri config with eu-north-1 account
removed Kubernetes workflow content
add spark processing support to processing jobs
reshape Parents into experiment analytics dataframe
add model monitor image accounts for af-south-1 and eu-south-1
update PyTorch 1.6.0 inference image uri config
update max_run_wait to max_wait in v2.rst for estimator parameters
Revert "change: update image uri config for pytorch 1.6.0 inference (#1864)"
refactor normalization of args for processing
formatting changes from updates to black
break out methods to get processing arguments
check ast node on later renamers for cli v2 updater
### Bug Fixes and Other Changes * update rulesconfig to 0.1.5
Neo algorithm accounts for af-south-1 and eu-south-1
update 1p estimators class description
new 1P algorithm accounts for af-south-1 and eu-south-1
add DLC account numbers for af-south-1 and eu-south-1
use pathlib.PurePosixPath for S3 URLs and Unix paths
Update migration tool to support breaking changes to create_model
Nothing published for this version
Neo: Add Granular Target Description support for compilation
remove redundant information from the user_agent string.
Add mpi support for mxnet estimator api
convert network_config in processing_config to dict
Nullable fields in processing_config
Add model monitor support for us-gov-west-1
Add ModelClientConfig Fields for Batch Transform
add spot instance support for AlgorithmEstimator
### Documentation Changes * add Step Functions SDK info
add deprecation warnings for estimator.delete_endpoint() and tuner.delete_endpoint()
Apache Airflow integration for SageMaker Processing Jobs
add 3.8 as supported python version
### Testing and Release Infrastructure * add py38 to buildspecs
document that Local Mode + local code doesn't support dependencies arg
remove include_package_data=True from setup.py
support for describing hyperparameter tuning job
include py38 tox env and some dependency upgrades
### Features * add support for SKLearn 0.23
Allow selecting inference response content for automl generated models
Support for multi variant endpoint invocation with target variant param
Use boto3 DEFAULT_SESSION when no boto3 session specified.
[doc] Added Amazon Components for Kubeflow Pipelines
fix undoc directive; removes extra tabs
remove some duplicated documentation from main README
support TensorFlow training 2.2
MXNet elastic inference support
update AutoML default max_candidate value to use the service default
update DatasetFormat key name for sagemakerCaptureJson
specify S3 source_dir needs to point to a tar file
### Bug Fixes and Other Changes * address flake8 error
### Bug Fixes and Other Changes * upgrade boto3 to 1.13.6
support inter container traffic encryption for processing jobs
add tensorflow training 1.15.2 py37 support
update xgboost latest image version
training_config returns MetricDefinitions
document model.tar.gz structure for MXNet and PyTorch
add doc8 check for documentation files
Your coding agent can read these notes before it upgrades. Set up the MCP server →