NewYour coding agent can read the release notes before it upgrades.Set up the MCP server →
PyPI, #999 most downloaded on PyPI
Open source library for training and deploying models on Amazon SageMaker.
Latest stable
3.29.0
Last release 17 days ago
05 Nov 2026
Ships on a steady schedule
a new release about every 2 weeks
Nearly every release is documented
notes for 60 of the last 60 stable releases
3 versions withdrawn
withdrawn after publishing
9 years old
699 releases · first in 2017
Drag across the ruler to compare, or tap a release to open it.
Added transform with monitoring pipeline step in transformer
image_scope is inference_gravitonUpdate boto3 version to 1.26.20
Your coding agent can read these notes before it upgrades. Set up the MCP server →
### Features * add support for PT1.12.1
support customized timeout for model data download and inference container startup health check for Hosting Endpoints
Add support for TF 2.10 training
Graviton support for XGB and SKLearn frameworks
support torch_distributed distribution for Trainium instances
Update Neo-TF2.x versions to TF2.9(.2)
fix(local-mode): loosen docker requirement to allow 6.0.0
added monitor batch transform step (pipeline)
Edit test file for supporting TF 2.10 training
Support KeepAlivePeriodInSeconds for Training APIs
local pipeline step argument parsing bug
Adding support in HuggingFace estimator for Training Compiler enhanced PyTorch 1.11
support python 3.10, update airflow dependency
Implement Kendra Search in RTD website
Added endpoint_name to clarify.ModelConfig
local mode executor implementation
AutoGluon 0.4.3 and 0.5.2 image_uris
add warnings for xgboost specific rules in debugger rules
added more ml frameworks supported by SageMaker Workflows
Add target_model to support multi-model endpoints
heterogeneous cluster set up in distribution config
edit to clarify how to use inference.py
update prebuilt models documentation
Add helper method to generate pipeline adjacency list
Adding Training Compiler support for TensorFlow estimator starting TF 2.9
AutoGluon 0.4.2 image_uris support
add input parameterization tests for workflow job steps
bump importlib-metadata version upperbound to support TF2.9
turn off Pipeline Parameter inheritance from python primitives
pin protobuf to < 4.0 to fix breaking change
add 'Domain' property to RegisterModel step
Revert Prevent passing PipelineVariable object into image_uris.retrieve
Support Properties for StepCollection
Add ModelStep for SageMaker Model Building Pipeline
repack model locally when local_code local mode
deprecate: Remove deprecated argument s3_data_distribution_type
Automl integ describe job check
Add encryption setting to tar_and_upload_dir method
jumpstart notebook utils -- list model ids, scripts, tasks, frameworks
Add Jumpstart example notebooks
### Bug Fixes and Other Changes * #using uuid to randomize, otherwise system timestamp is used
xgboost, sklearn network isolation for jumpstart
Adds Spark Processing Notebook to Notebook Tests
update lambda code on pipeline create/update/upsert for Lamb…
dependabot integ - move all deps to requirements.txt
Hugging Face Transformers 4.17 for TF 2.6
Revert "fix: Fix Pipeline variables related customer issues (#2959)"
Update Inferentia Image URI Config
pluggable instance fallback mechanism, add CapacityError
Update black-check version, add support for Spark 3.1 Processing
enable EnableInterContainerTrafficEncryption for model monitoring
new from serverlessAutoGluon 0.3.2 and 0.4.0 image_uris
Inferentia Neuron support for HuggingFace
Remove sagemaker_job_name from hyperparameters in TrainingStep
### Bug Fixes and Other Changes * jumpstart model table
override jumpstart content bucket
Add FailStep Support for Sagemaker Pipeline
### Bug Fixes and Other Changes * Add CMH to the non-P3 list
One column per quarter.