Improved Azure authentication

What's Changed

This release enables the use of password-free federated credentials for GitHub workflows.

Note that this release moves some helper functions for authentication to a new namespace, and might break existing code.

hi-ml-azure

• ENH: Add AzureCLI authentication for SDK v1 codepath by @ant0nsc in #922

Full Changelog: v0.4.3...v0.4.4

v0.4.3

What's Changed

General

• MNT: Delete workflows and switch to built-in tooling by @ant0nsc in #920

hi-ml-azure

• ENH: Make logging consistent by @ant0nsc in #921

Full Changelog: v0.4.2...v0.4.3

Hotfix to enable latest version of azureml-dataprep

What's Changed

AzureML has deployed code that is not compatible with versions of azureml-dataprep before 5.1.6. However, that version 5.1.6 does not have the MountContext class that hi-ml-azure is using.

hi-ml-azure

• ENH: Remove mount context imports by @ant0nsc in #919

Full Changelog: v0.4.1...v0.4.2

More flexible MPI job handling

What's Changed

This release adds a small improvement for AML SDK v1, which can now also be run as MPI jobs on a single node.

hi-ml-azure

• ENH: Enable single-node MPI jobs with SDK v1 by @ant0nsc in #914

Full Changelog: v0.4.0...v0.4.1

Switch to Python 3.9

What's Changed

This version of hi-ml now requires Python 3.9.

hi-ml-azure

• BUG: Allow relative path for entry script for SDK v2 submission by @ant0nsc in #904
• ENH: Apply security fixes by @ant0nsc in #912
• ENH: Allow command in script run config by @ant0nsc in #909
• ENH: Upgrade Python and other packages by @ant0nsc in #910

Full Changelog: v0.3.4...v0.4.0

hi-ml-azure improvements, take 2

What's Changed

This is an second attempt to release what was meant to be v0.3.3. Publishing of that version was blocked because pyright consistently ran out of diskspace. This has now been disabled in #906.

Full Changelog: v0.3.3...v0.3.4

hi-ml-azure improvements

What's Changed

hi-ml-azure

• ENH: Single node runs for SDK v2 can be non-distributed by @ant0nsc in #905

Full Changelog: v0.3.2...v0.3.3

Bug fixes in hi-ml-azure, cpath improvements

What's Changed

hi-ml-azure

• BUG: Fix for duplicate authentication by @ant0nsc in #878
• BUG: AmlIgnore file should not lose its last linebreak by @ant0nsc in #903
• BUG: Enable submitting jobs to compute instances by @ant0nsc in #902

hi-ml-cpath

• BUG: TIFF conversion enforces resolution unit to be centimeter by @ant0nsc in #887
• BUG: DeepMIL model uses training data even when running in eval mode by @ant0nsc in #889
• BUG: Fix KeyError when converting svs to tiff by @kenza-bouzid in #891
• ENH: Add DenseNet encoder by @kenza-bouzid in #892
• ENH: Add Macenko stain normalization by @harshita-s in #896

Full Changelog: v0.3.1...v0.3.2

v0.3.1

What's Changed

hi-ml-azure

• FIX: Stop replacing dashes with underscores in experiment name by @fepegar in #884
• FIX: Fix env variables not passed to AML SDK v2 jobs by @fepegar in #885

Full Changelog: v0.3.0...v0.3.1

v0.3.0

What's Changed

hi-ml-azure

• ENH: Expose v1 dataset overwrite by @peterhessey in #853
• ENH: Better commandline parsing of Enums by @ant0nsc in #861
• BUG: Prevent creating v2 data asset without specifying datastore by @peterhessey in #854
• FIX: Fix empty list of script params not used by @fepegar in #881

hi-ml

• ENH: Enable flexible resolution unit for mock data generator by @kenza-bouzid in #827
• MNT: Upgrade pytorch to 1.11 by @kenza-bouzid in #833
• ENH: Improve handling of metric comparison for regression tests by @ant0nsc in #835
• BUG: Misleading warnings when comparing metrics by @ant0nsc in #838
• ENH: Allow variants of model parameters in the same class by @ant0nsc in #840
• ENH: Add log files to AzureML runs to identify preemption by @ant0nsc in #841
• MNT: Fix incorrect type annotations by @ant0nsc in #846
• ENH: Ability to evaluate model on a new dataset by @ant0nsc in #859
• FIX: Fix typo in makefile for make targets by @jubinmathew1995 in #865
• ENH: Add tests for use of checkpoints in inference_only mode by @ant0nsc in #876

hi-ml-multimodal

• FIX: Fix notebook CI when trying to merge from fork by @fepegar in #829
• ENH: Add BioViL-T image and text encoders by @ozan-oktay in #830
• ENH: Consume BioViL-T Pretrained Models in Inference Engines by @ozan-oktay in #845
• FIX: Fix notebook not working on Binder by @fepegar in #852
• MNT: Bump hi-ml-multimodal version by @fepegar in #849
• ENH: Improve support for choosing image model weights by @markpinnock in #807
• MNT: Bump hi-ml-multimodal version: 0.2.0 → 0.2.1 by @fepegar in #869

hi-ml-cpath

• ENH: Add LoadMaskSubROId transform by @kenza-bouzid in #828
• ENH: Use Latest Monai GridPatch to filter both threshold and count by @kenza-bouzid in #834
• BUG: Convert to tifffile if file doesn't exist or if it's empty by @kenza-bouzid in #836
• ENH: Add non normalized confusion matrix to plot options by @kenza-bouzid in #791
• ENH: Enable activation checkpointing for Resnet and SwinTransformer by @kenza-bouzid in #842
• ENH: Refactor Swin Transformer checkpointing by @kenza-bouzid in #843
• ENH: Add display name to montage script by @kenza-bouzid in #847
• ENH: Add diagnosis for Tiff Conversion by @kenza-bouzid in #862
• ENH: Add normalized confusion matrix key to metrics keys by @maxilse in #868
• ENH: Add a class_weights argument to the data module by @ant0nsc in #871
• BUG: Fix extra validation epoch hook call order by @kenza-bouzid in #873
• ENH: Add account key argument to generate checkpoint URL script by @kenza-bouzid in #874
• ENH: Clean up design of slides dataset by @ant0nsc in #875

Other

• ENH: Autoupdate hooks [pre-commit.ci] by @pre-commit-ci in #831
• Bump pypa/gh-action-pypi-publish from 1.5.0 to 1.8.1 by @dependabot in #844
• STY: Add Black pre-commit hook by @fepegar in #782
• Bump pypa/gh-action-pypi-publish from 1.8.1 to 1.8.3 by @dependabot in #856
• Bump pypa/gh-action-pypi-publish from 1.8.3 to 1.8.5 by @dependabot in #863
• ENH: Autoupdate hooks [pre-commit.ci] by @pre-commit-ci in #864

New Contributors

• @jubinmathew1995 made their first contribution in #865

Full Changelog: v0.2.19...v0.3.0