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 9 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3 versions withdrawn
withdrawn after publishing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
Correct mypy type checking through PEP 561
include model channel for gated uncompressed models
One column per quarter.
Fix test_huggingface_tei_uris()
support custom workflow deployment in ModelBuilder using SMD image.
Nothing published for this version
update pr test to deprecate py38 and add py312
### Bug Fixes and Other Changes * tgi image uri unit tests * Fix deepdiff dependencies
Added handler for pipeline variable while creating process job
Skip tests failed due to deprecated instance type
add integ tests for training JumpStart models in private hub
Skip tests with deprecated instance type
Add support for TGI Neuronx 0.0.27 and HF PT 2.3.0 image in PySDK
added ap-southeast-7 and mx-central-1 for Jumpstart
Add warning about not supporting torch.nn.SyncBatchNorm
keep sagemaker_session from being overridden to None
Add support for deepseek recipes
use jumpstart deployment config image as default optimization image
model server might have already done a serialization. honor that by not decoding the request again if it is not already bytes or bytestream
chore: remove support for ecr spec fallbacks for jumpstart models
Support SageMakerTrainingPlan for training jobs
Partner App Auth Provider for SDK support
Nothing published for this version
update image_uri_configs 11-20-2024 06:17:41 PST
Optimize() validations across TRT, VLLM, Neuron container optimizations
optimization technique related validations.
Marketplace model support in HubService
update image_uri_configs 10-29-2024 07:17:56 PST
Pass kwargs to HuggingFaceModel.deploy()
update image_uri_configs 09-17-2024 07:17:54 PST
add deployment config name in modelbuilder telemetry
Add SageMaker Core to the dependency
FastAPI integration for In_Process Mode (2/2)
Support for ModelBuilder In_Process Mode (1/2)
chore: telemetry for deployment configs
added code scanning through CodeQL
Nothing published for this version
InferenceSpec support for MMS and testing
update image_uri_configs 07-04-2024 07:17:24 PST
allow for inf spec and server override to be passed
Improve docstrings for estimator tags
Update DJLModel class for latest container releases
Update README.rst to show conda-forge version of SageMaker SDK
feat(sagemaker-mlflow): New features for SageMaker MLflow
add 'ModelCard' property to Register step
estimator.deploy not respecting instance type
update image_uri_configs 06-06-2024 07:17:31 PST
Convert pytorchddp distribution to smdistributed distribution
onboard tei image config to pysdk
AutoGluon 1.1.0 image_uris update
allow choosing js payload by alias in private method
Fix UserAgent logging in Python SDK
set default allow_pickle param to False
support session tag chaining for training job
add DXB and CGK to Jumpstart regions
Add TF 2.14 Graviton Inference support
JumpStart Gated Model Support in ModelBuilder Local Modes
[Fix] Switch to subprocess in ModelBuilder when capturing dependencies
bump apache-airflow version to 2.8.3
remove failing deprecated tests from suite
remove pytorch test for deprecated function
clarify.py for time series explainability jobsYour coding agent can read these notes before it upgrades. Set up the MCP server →