Skip to content

Latest commit

 

History

History
124 lines (73 loc) · 6.04 KB

CHANGELOG.md

File metadata and controls

124 lines (73 loc) · 6.04 KB
Raw

Changelog

v0.23.0 (2022-11-10)

Fixes a bug in v0.22.0 where it didn't update the lifecycle scripts for SageMaker Studio and SageMaker Notebook Instances correctly.

v0.22.0 (2022-10-21)

The GUI plugin is now updated to support JupyterLab 3.x and, therefore, supports the current versions of SageMaker Notebook Instances and SageMaker Studio notebooks.

As of this release, the JupyterLab extension is no longer compatible with JupyterLab 2.x. If you're using JupyterLab 2.x, please use release v0.21.0.

Additional small features:

  • Add a real life cycle script for SageMaker Studio notebooks so you can install the plugin permanently in your domain.
  • Support for China regions (thanks to @yosefbs).
  • Update default container environment to Python 3.10.

v0.21.0 (2022-10-04)

This release is just security patches in the dependencies.

v0.20.0 (2021-12-09)

This release provides support for JupyterLab 2.x.

As of this release, the JupyterLab extension is no longer compatible with JupyterLab 1.x. If you're using JupyterLab 1.x, please use release v0.19.0.

v0.19.0 (2021-12-07)

This release has a number of small improvements and bug fixes.

After upgrading to this release, we recommend that you delete and recreate your infrastructure with the following commands:

  1. aws cloudformation delete-stack --stack-name sagemaker-run-notebook
  2. run-notebook create-infrastructure

This will pick up the new managed policies (see under [Bug Fixes])(#bug-fixes) below) which can not be updated with the --update option.

Bug fixes

  • Fixed an install failure on versions >= 3.10. (Fixes issue #35)
  • Use managed policies so that we can include the policies in other roles (See Change policies to managed policies by @dmoser04).
  • Use a single permission statement on the Lambda function for all EventBridge rules. This prevents us overflowing the number of separate permissions when we have many scheduled notebook runs. This also means that we're not creating permissions at schedule time. (Fixes issue #9)
  • Correctly handle scheduled runs with no supplied parameters. (Fixes issue #25)
  • Update various JS dependencies for security fixes.

v0.18.0 (2020-11-06)

This is a documentation only release:

  • It adds the documentation tarball (docs.tar.gz) to the release.
  • It adds reasonable docstrings to all the exported functions in sagemaker_run_notebook.

v0.17.0 (2020-11-05)

Ignore. Not actually released.

v0.16.0 (2020-10-23)

Users moving to this version should update their Lambda function and permissions by running:

$ run-notebook create-infrastructure --update

Features

  • Added the ability to run notebooks connected to an EMR cluster using SparkMagic. See the EMR examples for more information on using this feature.
  • Add the permissions to the default execution policy and role that allow notebook executions to attach to user VPCs via the --extra option.

Bug fixes

  • Pass environment variables specified in --extra parameters through to the notebook execution.
  • Restrict S3 permissions in the default role to buckets whose name contains the word "SageMaker".
  • Unpin the version of the Python "requests" library because it is (a) no longer necessary and (b) cause a dependency error with the latest version of boto3.

v0.15.0 (2020-09-23)

Bugs

  • Fixed Issue #4, a regression where the invocation of papermill would fail with "No such kernel". (Note that any containers built under v0.14.0, should be rebuilt with `run-notebook )

Features

Two small changes:

  • Added a -v option to run-notebook to display the current installed version of the library.
  • Changed the install scripts for SageMaker notebooks to install directly from the release on GitHub so users don't need to copy to an S3 bucket first.

v0.14.0 (2020-09-14)

Features

  • Run notebooks written in R or other languages. See the newly added R example for information.
  • When using the CLI or Library, you can supply extra arguments to the SageMaker Processing Job used to execute the notebook. For example to allow your notebook to run for up to a full day, invoke your notebook with run-notebook run foo.ipynb --extra '{"StoppingCondition":{"MaxRuntimeInSeconds":86400}}'. Using this mechanism, you can add extra inputs and outputs, connect to a VPC, add environment variables, expand disk space, add the run to a specific experiment, etc. See the documentation for SageMaker Processing Jobs for more.

Running Notebooks

  • Add an option to pass extra parameters to SageMaker Processing jobs from the CLI/Library (not yet supported in the JupyterLab extension).
  • Add the ability to use the CLI to run the notebook using local Docker containers (run-notebook local).

Container building (run-notebook create-container)

  • Support base containers from ECR repos in other AWS accounts (to enable using AWS Deep Learning Containers).
  • Add the -k option to allow alternate Jupyter kernels
  • Add the --script option for running arbitrary shell scripts as part of building the container.
  • Tail the build logs when running the build so you can see any errors that happen inline.
  • Automatically install Python if the base image doesn't already have it.

JupyterLab Extension

  • Fixes for running in Safari

v0.13.0 (2020-06-15)

Initial release.