3.14.0

What's Changed

• fix(integ): remove module-level env var from MTRL tests by @lucasjia-aws in #5929
• test: run integ-tests-us-east-1 for all submodules by @lucasjia-aws in #5934
• test: add nova tests by @lucasjia-aws in #5933
• chore: Run only the changed submodule's tests for test-only PRs by @lucasjia-aws in #5936
• test: fix canaries-v3 by @lucasjia-aws in #5940
• chore: deprecate Python 3.9 support by @mollyheamazon in #5941
• chore: Change GPU integ tests and Canaries schedule by @lucasjia-aws in #5938
• chore: Change GPU integ test schedule to 3h interval by @lucasjia-aws in #5945
• chore: Add unit/slow/localmode and unit-test-v3 jobs to CI Health workflow by @lucasjia-aws in #5949
• fix(ci): Use source-version-override so CI Health builds checkout the right branch by @lucasjia-aws in #5950
• fix:unskip and add mock test setup for mtrl integ tests by @rsareddy0329 in #5937
• chore: Register 'serial' pytest marker in mlops tox.ini by @lucasjia-aws in #5951
• test(serve): isolate Bedrock model-import integ tests behind import_model marker by @lucasjia-aws in #5952
• docs(ml_ops): Add Data Catalog Config notebook for Feature Store by @Vishakha263 in #5927
• fix(train): fall back to subscription recipes for models without base recipes by @aviruthen in #5946
• updated the SDK to use latest LMI image for sdk v3.x by @access2rohit in #5954
• Update training resource name in MTRL tests as previous MPG got cleanedup by @rsareddy0329 in #5953
• fix(serve): repair HuggingFace -> JumpStart redirect in ModelBuilder by @ZealSV in #5958
• fix(train): use latest model package in LLM-as-judge base model integ tests by @mollyheamazon in #5959
• Release 3.14.0 by @mollyheamazon in #5961

New Contributors

• @Vishakha263 made their first contribution in #5927

Full Changelog: v3.13.1...v3.14.0

v3.13.1

Features

• feat: add import job polling and provisioned throughput for Bedrock OSS deployments

Bug Fixes

• fix: Address MTRL Eval Hyperparameters issue

v3.13.0

New Features

• feat: Model customization - Add new finetuning Trainer - MultiTurnRLTrainer(Multi-Turn Reinforcement Learning)
• feat: Model customization - Add new evaluator - MultiTurnRLEvaluator
• feat: Deployment - Add MTRL support for BedrockModelBuilder and ModelBuilder.

Documentation

• Add details for MTRL trainer along with other finetuning interfaces under Model customization Section - https://sagemaker.readthedocs.io/en/stable/model_customization/model_customization.html
• Add details for MTRL evaluator along with existing evaluators.

Fixes

• fix: set sagemaker_config=None on mock session in test_from_jumpstart_config_applies_volume_size
• Restore BatchTransformInput.destination attribute in v3

v3.12.0

New Features

• SageMaker Token Generator (#5868): Embed the aws-sagemaker-token-generator library into sagemaker.core so users can generate SageMaker bearer tokens without installing a separate wheel. Usage: from sagemaker.core.aws_sagemaker_token_generator import provide_token
• Feature Processor - Spark 3.5 / Python 3.12 support (#5816): Dynamic Spark image resolution based on installed PySpark and Python versions. Supports Spark 3.1/3.2/3.3/3.5 with Python 3.9 and 3.12. Auto-installs sagemaker-feature-store-pyspark for Spark remote jobs.

Bug Fixes

• Networking vpc_config AttributeError and telemetry region fallback (#5839): Fix AttributeError on vpc_config in networking and telemetry region fallback for classmethods.
• Add CustomAttributes field to DefaultPayloadsModel (#5870): Add missing CustomAttributes field to DefaultPayloadsModel.
• sagemaker-core: Preserve falsy values in serialize() output (#5860): Fix bug where False, 0, and "" were silently dropped by serialize() due to truthy check. This caused issues like optimize_model=False being sent as True.
• serve: Prevent code injection in capture_dependencies path interpolation (#5792): Security fix — use repr() escaping to prevent code injection via crafted directory names in ModelBuilder with dependencies={"auto": True}. (CWE-94, P414309851)
• VolumeSizeInGB missing from v3 deploy for JumpStart models (#5847): Fix VolumeSizeInGB not being passed through when deploying models with inference_volume_size from JumpStart config.

v3.11.0

New Features

• Auto-detect subscription recipe hyperparameters in SFTTrainer for Nova Forge datamix support
• Create asymmetric ECDSA signing key in feature processor step compiler for remote function payload verification

Documentation

• Add Feature Store reference to Implement MLOps page
• Replace internal S3 URIs with user placeholders in SFT notebook

v3.10.1

v3.10.1 (2026-05-07)

• Bug Fixes
  • Fix KMS key propagation in check steps (QualityCheckStep, ClarifyCheckStep)
  • Fix JumpStart network isolation in ModelBuilder
  • Fix base_model_arn construction to use private hub when SAGEMAKER_HUB_NAME is set
  • Fix imports for Model Customization interfaces
  • Fix handling of unrecognized JumpStart container images in ModelBuilder
  • Increase default timeout for training jobs

Release v2.257.3

Bug Fixes

• S3 bucket operations in V2 (#5803)
• Improve subprocess exception handling in git_utils (#5812)
• Fix v2 integ tests (#5822)

Other Changes

• Added ISO accounts for DJI LMI (#5789)

v3.10.0

v3.10.0 (2026-05-01)

New Features

• Make _PipelineExecution a public class
• Add CodeArtifact support for ModelTrainer and FrameworkProcessor requirements.txt installation

Bug Fixes

• Fix S3 bucket operations
• Fix potential S3 path traversal
• Wire FrameworkProcessor code_location into code upload paths
• Improve subprocess exception handling in git_utils

Other

• Update service-2.json with latest public botocore service model

v3.9.0

v3.9.0 (2026-04-23)

New Features

• Add wait_timeout parameter to train()
• Add MLFlow experiment link to eval output
• Allow SAGEMAKER_HUB_NAME env var override for HUB_NAME constant

Bug Fixes

• Fix output data config in HyperparameterTuner
• Improve error messages for waiter timeouts
• Updated DJI LMI ISO accounts
• Skip none hyperparameters
• Add us-west-2 to Nova supported regions
• Fix ModelBuilder overwriting user-provided HF_MODEL_ID for DJL Serving
• Fix ModelBuilder with source_code + DJL LMI: /opt/ml/model becomes read-only
• Fix ModelTrainer and HyperparameterTuner missing environment variables

Release v2.257.2

Enhancements

• Update SDK to use latest LMI v23 image for sdk v2.x (#5710)
• Update SDK to use latest LMIv22 image for sdk v2.x (#5641)
• Update SDK to use latest LMI image for sdk v2.x (#5617)

Bug Fixes

• Security fixes for Triton HMAC key exposure and missing integrity check (v2) (#5656)
• Include jumpstart/region_config.json in MANIFEST.in (#5605)
• Apply default experiment config for pipelines only in regions with SageMaker Experiments (#5570)

Other Changes

• Remove record-related telemetry decorators (#5626)
• Fix integ tests (#5674)