- support mxnet 1.7.0
- drop py27, add py38 support
- remove the openssh addition
- update test readme
- add issue templates
- Add MPI support for distributed training in MXNet
- bump version of sagemaker-training for script entry point fix
- remove confusing information from the Readme.
- Rename buildspec files.
- Make docker folder read only, remove unused tests, rename test-toolkit/ -> test/.
- Bump version of sagemaker-training for typing fix
- remove unused scripts.
- add Python 3.7 support
- use tox in buildspecs
- Replace sagemaker-containers with sagemaker-training
- update smdebug version
- add requirements.txt integ test
- fix for too many gpus
- upgrade pillow etc. to fix safety issues
- Add gluoncv
- use sagemaker_mxnet_training<4 in Dockerfiles instead of pinning the version
- upgrade sagemaker-containers to 2.8.2
- do not run DLC tests when toolkit has changes.
- update smdebug
- update container build instructions with correct tar file name
- Switch to pypi gluonnlp package
- refactor toolkit tests.
- install sm experiments on any python version 3.5 or above
- Update smdebug to 0.7.0
- move experiments import statements after python version check
- install experiments sdk only for python 3.6
- install SageMaker Python SDK into Python 3 images
- Upgrade to latest version of sagemaker-experiments
- copy all tests to test-toolkit folder.
- Fix issue in installation of sagemaker-containers
- update: Update license URL
- update: Horovod Support with MXNet backend on DLC for MXNet 1.6.0
- Use GluonNLP stable release tag
- Use sagemaker-mxnet-training package from pypi.
- Rename package to be 'sagemaker-mxnet-training'. Add automated release to PyPI.
- Add GluonNLP
- Update AWS-MXNet version to 1.6.0 - official release of 1.6
- Update build artifacts
- Misspelling of sagemaker_mxnet_container_*.tar.gz
- Downgrade pip version and unpin awscli
- Revert "Add GluonNLP 0.9 pre-release (#116)"
- Add GluonNLP 0.9 pre-release
- update smdebug version to 0.5.0.post0
- update: Constrain package versions in 1.6.0 dockerfiles
- Build context changes based on new requirements
- Update README for 1.6.0 release
- update copyright year in license header
- Change build context for building MX Training Dockerfiles
- properly fail build if has-matching-changes fails
- properly fail build if has-matching-changes fails
- Rename package to be 'sagemaker-mxnet-training'. Add automated release to PyPI.
- Add GluonNLP
- Update AWS-MXNet version to 1.6.0 - official release of 1.6
- Update build artifacts
- Misspelling of sagemaker_mxnet_container_*.tar.gz
- Downgrade pip version and unpin awscli
- Revert "Add GluonNLP 0.9 pre-release (#116)"
- Add GluonNLP 0.9 pre-release
- update smdebug version to 0.5.0.post0
- update: Constrain package versions in 1.6.0 dockerfiles
- Build context changes based on new requirements
- Update README for 1.6.0 release
- update copyright year in license header
- Change build context for building MX Training Dockerfiles
- Release 1.6.0 Dockerfiles
- properly fail build if has-matching-changes fails
- properly fail build if has-matching-changes fails
- use SageMaker Containers' ProcessRunner for executing the entry point
- use regional endpoint for STS in builds
- update instance type region availability
- split cpu and gpu test commands.
- add flake8 in buildspec.yml and fix flake8.
- fix placeholder name cpu-instance-type in buildspec-release.yml
- Add eu-west-3, eu-north-1, sa-east-1 and ap-east-1 to the no-p2 regions.
- add MXNet 1.4.1 Dockerfiles
- fix build directory in scripts/build_all.py
- fix copy command in buildspec-release.yml
- remove JSON syntax error from buildspec-release.yml
- Updating SageMaker Containers to the latest version
- pin sagemaker in test dependencies
- fix local GPU test command in buildspec-release.yml
- explicitly set lower bound for botocore version
- parametrize Python version and processor type in integ tests
- add hyperparameter tuning integ test
- flesh out SageMaker training integ tests
- skip GPU test in regions with limited p2s
- use correct variable name in release buildspec
- adjust parallelism in buildspec-release.yml
- Add release buildspec
- Remove outdated information from README
Compatible with MXNet version 1.4.0.
This version of the SageMaker MXNet container only accounts for training.
Serving support is now split and can be found here: https://github.com/aws/sagemaker-mxnet-serving-container
Compatible with MXNet version 1.3.0
This version of MXNet introduced an updated training script format. See https://sagemaker.readthedocs.io/en/v1.22.0/using_mxnet.html#preparing-the-mxnet-training-script for more information.
Compatible with MXNet versions 0.12.1-1.2.1.