Sagemaker

Features

* Add SageMaker Core to the dependency

Bug Fixes and Other Changes

* Disable test_mnist_async
* SMP v2.5
* update image_uri_configs 08-29-2024 07:17:59 PST

Features

* FastAPI integration for In_Process Mode (2/2)

Bug Fixes and Other Changes

* chore: add HF LLM neuronx 0.0.24 image
* TF-2.16 test modification and handling
* fix test fail
* Add troubleshooting links to exceptions
* cross account private hub model fine-tuning
* chore: cleanup jumpstart factory
* disable failing integration tests

Features

* Support for ModelBuilder In_Process Mode (1/2)
* Pulling in dependencies (in_process mode) using conda environment
* Add optional CodeArtifact login to FrameworkProcessing job script
* implemented security-monitoring to send metrics to CW 1510

Bug Fixes and Other Changes

* alt configs model deployment and training issues
* fix keras extension in integ test
* update image_uri_configs 08-13-2024 07:17:54 PST
* trn1 instance family does not support volume size
* Update model.py
* removed log statement
* update image_uri_configs 08-09-2024 07:18:00 PST
* Added torchrun compatibility for distributet training across multiple GPUs in a single node (single instance)
* BiasConfig type hint
* add model monitor image accounts for ap-southeast-5 and eu-central-2
* aligned UTC times with PST
* ensure hpt jobs inherit tags from config
* add JumpStart PDT and OSU regions
* chore(deps): bump certifi in /src/sagemaker/serve/utils
* Updates for DJL 0.29.0 release
* chore(deps): bump apache-airflow from 2.9.2 to 2.9.3 in /requirements/extras
* chore(deps): bump torch from 2.0.1 to 2.2.0 in /tests/data/serve_resources/mlflow/pytorch
* avoided printing stack trace and escaped input
* removing kwargs as this is breaking predictor_cls param for mode…

Features

* triton v24.05

Bug Fixes and Other Changes

* chore: telemetry for deployment configs
* censoring sensitive values from being logged
* update image_uri_configs 08-05-2024 07:17:38 PST
* enable uncompressed model artifacts upload to S3 for SAGEMAKER_ENDPOINT overwrite for TGI, TEI, MMS model servers
* ModelReference deployment for Alt Configs models
* Add optional typecheck for nullable parameters
* Update package metadata

Features

* added code scanning through CodeQL

Bug Fixes and Other Changes

* Fixed cpu isntance type for the estimator register test
* update image_uri_configs 07-29-2024 11:28:28 PST
* avoid AccessDenied error for a while on SageMaker Studio wtih do…
* SMP PT 2.3 Fix
* chore: pin framework version in serverless inference tests
* image uri in TGI 2.2.0 image
* explicitly access enum member values to avoid Python version related regression
* chore: add huggingface TGI 2.2.0 config
* update image_uri_configs 07-22-2024 11:53:54 PST
* update image_uri_configs 07-17-2024 07:17:38 PST
* update image_uri_configs 07-16-2024 07:17:45 PST
* add support for new regions