v0.2.11

What's Changed

• FIX: Replace special characters in experiment name with underscores by @mebristo in #688

Full Changelog: v0.2.10...v0.2.11

Incorporate recent fixes

What's Changed

• ENH: Enable flexible finetuning of deepmil submodules by @kenza-bouzid in #549
• ENH: Enable flexible weights initialization from src_checkpoint by @kenza-bouzid in #554
• ENH: Update metrics for binary classification by @harshita-s in #557
• ENH: Updating transfer_main to main by @vale-salvatelli in #558
• ENH: Add command line arg for custom SSL checkpoint by @mebristo in #560
• ENH: Add dataframe kwargs for flexible csv read by @kenza-bouzid in #564
• DOC: Clean up logs (#565) by @mebristo in #566
• BUG: Set dataframe index column only when necessary by @kenza-bouzid in #569
• ENH: Add pl_sync_batchnorm flag by @kenza-bouzid in #571
• ENH: Add smoke test for flexible finetuning by @kenza-bouzid in #563
• ENH: Add debug_ddp flag by @kenza-bouzid in #573
• ENH: Parameterize tiles intensity scaling by @kenza-bouzid in #572
• ENH: Add specific ImageNet encoders for transferability study by @kenza-bouzid in #574
• BUG: Fix SettingWithCopyWarning and parametrize metadata validation by @kenza-bouzid in #577
• ENH: Add/reorganise metrics and correctly normalise confusion matrix by @harshita-s in #578
• BUG: Fix probability scores displayed on histo plots by @kenza-bouzid in #581
• BUG: Cast binary labels into integers. by @kenza-bouzid in #583
• ENH: Log epoch of best checkpoint by @kenza-bouzid in #584
• ENH: Adding PR curve plot for val and test for single fold by @vale-salvatelli in #585
• ENH: Enable stain normalization for multiple images by @harshita-s in #587
• ENH: Dropout in transformer layers in DeepMIL by @harshita-s in #590
• ENH: Adding gradient accumulation flag to trainer by @vale-salvatelli in #592
• ENH: Add essential changes from main by @ant0nsc in #602
• BUG: Fix top bottom tiles selection mismatch. by @kenza-bouzid in #599
• ENH: Enable adding tags to aml runs from LightningContainer by @kenza-bouzid in #603
• ENH: Make crossval html report generic to all hyperdrive runs by @kenza-bouzid in #605
• ENH: Refactor extra validation epoch logic by @kenza-bouzid in #608
• ENH: Improve Amulet support and documentation by @ant0nsc in #615
• ENH: Set additional tags to Amulet Jobs by @kenza-bouzid in #620
• ENH: Add Loss Outliers Analysis callback by @kenza-bouzid in #618
• BUG: Fix double root folder for heatmaps by @kenza-bouzid in #622
• MNT: Fix YAML syntax for Amulet jobs (#624) by @kenza-bouzid in #626
• ENH: Track validation loss and add entropy value by @kenza-bouzid in #629
• BUG: Checkpoint path not picked up correctly by @harshita-s in #630
• ENH: Enable running cross validation via amulet by @kenza-bouzid in #631
• ENH: Validate model weights before inference by @kenza-bouzid in #635
• ENH: Enable heatmaps and thumbnails plots for WSI without masks by @kenza-bouzid in #637
• BUG: Set instance shuffling to True by @kenza-bouzid in #638
• ENH: Run extra validation epoch on a single gpu by @kenza-bouzid in #639
• ENH: Separate attention heatmap and slide thumbnail plot options by @kenza-bouzid in #640
• BUG: Retrieve tiles_ids from outputs dictionary for loss analysis by @kenza-bouzid in #641
• ENH: Extend StainNormalisation Transform to work with Slides Pipeline by @kenza-bouzid in #644
• ENH: HEDJitter update by @maxilse in #643
• ENH: Add log every nsteps parameter by @kenza-bouzid in #645
• ENH: Download Checkpoints across AML workspaces by @kenza-bouzid in #642
• BUG: Fix HTML report by @harshita-s in #647
• ENH: Run tests for histopathology folder in Conda environment by @ant0nsc in #487
• ENH: Add openslide to cpath conda environment by @ant0nsc in #646
• FIX: Fix bug in saving validation outputs by @mebristo in #649
• ENH: Merge Main into Transfer Main by @kenza-bouzid in #653
• ENH: Add pl_replace_sampler_ddp flag for small validation sets by @kenza-bouzid in #648
• ENH: Add display name, tags and shm size to SDK v2 job by @mebristo in #654
• ENH: Merge Transfer Main by @kenza-bouzid in #658
• BUG: Raise ClientAuthenticationError in default get credential by @kenza-bouzid in #659
• ENH: Parameterize CuCim backend by @kenza-bouzid in #661
• BUG: Fix mlflow authentication error by @kenza-bouzid in #662
• BUG: Move MlClient call to v2 scope only by @kenza-bouzid in #663
• ENH: Upgrade azureml packages by @kenza-bouzid in #664
• BUG: Remove all calls to get_ml_client outside v2 by @kenza-bouzid in #665
• BUG: Fix hanging smoke tests by @kenza-bouzid in #666
• BUG: Use permanent permalinks in docs to fix precommit check by @kenza-bouzid in #667
• BUG: Use micro average for MultiClassAccuracy by @kenza-bouzid in #671
• ENH: Upgrade monai to 1.0.1 by @kenza-bouzid in #668
• ENH: Fix security Alerts by @kenza-bouzid in #675
• ENH: Enable LoadROId with Openslide by @kenza-bouzid in #672
• BUG: Add symbolic link that had gotten deleted by accident by @ant0nsc in #676
• ENH: Extract coordinates from tensors metada by @kenza-bouzid in #677
• ENH: Enable custom workspace config path by @kenza-bouzid in #678
• ENH: Enable custom datastore name when running outside AML by @kenza-bouzid in #679
• BUG: Cast true labels to int in results dictionary by @kenza-bouzid in #682
• ENH: Group AML experiments by PR number by @ant0nsc in #681

Full Changelog: v0.2.9...v0.2.10

Incorporate AML SDK v2

Incorporate the use of Azure ML SDK v2 into the package.

What's Changed

• ENH: Incorporate AML SDK v2 by @mebristo in #627
• ENH: Enable hyperparameter search with AML SDK v2 by @mebristo in #650

Full Changelog: v0.2.8...v0.2.9

Use new AzureML runtime by default

This is a bug fix release. Most significant update is that hi-ml-azure submits jobs targeting the new AzureML runtime by default now.

What's Changed

• MNT: Fix YAML syntax for Amulet jobs by @fepegar in #624
• MNT: Replace print with logging.info by @fepegar in #623
• FIX: Fix test_get_dataset by @mebristo in #636
• ENH: Switch to new AzureML runtime by @ant0nsc in #576

Full Changelog: v0.2.7...v0.2.8

Relax package constraints

This is a small maintenance release to relax the package version constraints. AzureML packages can now be >= 1.42 rather than pinned.

What's Changed

• MNT: Relax package constraints for AzureML libraries by @ant0nsc in #619
• MNT: Update test helpers to work with latest AML package by @ant0nsc in #621

Full Changelog: v0.2.6...v0.2.7

v0.2.6

This is largely a bugfix release to improve jobs that run on Singularity (#612)

What's Changed

• ENH: Enable Global Similarity Comparison Between Text and Image Inputs by @ozan-oktay in #606
• ENH: Return figure when plotting similarity map by @fepegar in #600
• MNT: Bump hi-ml-multimodal version: 0.1.0 → 0.1.1 by @fepegar in #610
• ENH: Improve Amulet support and documentation by @ant0nsc in #612

Full Changelog: v0.2.5...v0.2.6

v0.2.5

What's Changed

• ENH: Show run ID and prefix if no files are found by @fepegar in #528
• BUG: Clear the top and bottom slides heaps after every validation epoch by @mebristo in #531
• ENH: Update tcga_prad paths and remove 'slide_' by @maxilse in #533
• ENH: Enable running inference-only with the hi-ml runner by @kenza-bouzid in #518
• FIX: Fixing typo in panda_model.md by @vale-salvatelli in #535
• DOC: Adding tcga DeepMIL results by @vale-salvatelli in #539
• ENH: Allow skipping of top/ bottom tile plotting by @mebristo in #536
• FIX: Flake8 problem that got missed due to auto-merge by @mebristo in #541
• BUG: Fixed PIL import by @maxilse in #542
• ENH: Mark completed regression_test_utils tests as complete by @mebristo in #537
• BUG: Fix column parsing by @maxilse in #544
• TEST: Incorporate additional smoke tests by @mebristo in #532
• ENH: Make label_column an instance attribute by @kenza-bouzid in #538
• BUG: Fix incompatibility between PIP and Conda CUDA versions by @fepegar in #546
• MNT: Prepare hi-ml-multimodal for PyPI by @fepegar in #545
• ENH: Add hubconf to load models without installing by @fepegar in #543
• FIX: Hotfix HTTP error downloading model with hub by @fepegar in #552
• DOC: Fix broken links in cpath documentation by @vale-salvatelli in #547
• BUG: Fix output saving for distributed models by @mebristo in #555
• ENH: Enable custom SSL encoder checkpoint by @mebristo in #562
• ENH: Clean up logs by @mebristo in #565
• ENH: Softmax requires dimension choice by @mebristo in #567
• FIX: Fix best checkpoint name by @mebristo in #568
• MNT: Install PyPI package in demo notebook by @fepegar in #570
• DOC: Use a consistent team name across all packages by @ant0nsc in #575
• ENH: Autoupdate hooks [pre-commit.ci] by @pre-commit-ci in #548
• MNT: Upgrade linting packages by @fepegar in #586
• ENH: Enable logging to AzureML when running training outside AzureML by @ant0nsc in #580
• FIX: Update env var settings for multi node by @mebristo in #588
• ENH: Upgrade hi-ml-cpath to Python 3.9 by @ant0nsc in #589
• ENH: Add a way of quickly starting runs with different seeds by @ant0nsc in #597
• ENH: Submit jobs to Singularity via Amulet by @mebristo in #596
• ENH: Use a locked environment hi-ml folder by @ant0nsc in #598

Full Changelog: v0.2.4...v0.2.5

v0.2.4: First version of pathology package, speed up pathology experiments,

What's Changed

• BUG: Restrict CredScan action to credential scanning only by @kenza-bouzid in #490
• Bump actions/setup-python from 2 to 4 by @dependabot in #469
• Bump codecov/codecov-action from 2 to 3 by @dependabot in #468
• Bump actions/setup-node from 2 to 3 by @dependabot in #470
• Bump actions/checkout from 2 to 3 by @dependabot in #471
• ENH: Reduce hyperparameters serialization time by @kenza-bouzid in #486
• ENH: Autoupdate hooks [pre-commit.ci] by @pre-commit-ci in #488
• BUG: Fix demo notebook failing on PRs by @fepegar in #492
• ENH: Add commit ID when downloading CXR-BERT model by @ozan-oktay in #482
• ENH: Exclude cache files from file watchers to speed up VSCode by @ant0nsc in #495
• ENH: Remove local dataset definitions from histopathology models by @ant0nsc in #494
• DOC: Update documentation for hi-ml-multimodal by @fepegar in #493
• ENH: Turn on mixed precision training by default by @ant0nsc in #496
• ENH: Prepare codebase for image model release by @fepegar in #497
• DOC: More details about creating Azure resources by @ant0nsc in #498
• DEL: Remove large files from RegressionTestResults by @mebristo in #500
• ENH: Move all training related functions to MLRunner class. by @kenza-bouzid in #499
• BUG: Mounted dataset disappears when mount script terminates by @ant0nsc in #504
• ENH: Simplify publishing workflow by using a single PyPi token by @ant0nsc in #510
• DOC: Improve Azure setup instructions by @ant0nsc in #507
• BUG: Plot conf matrix when not all labels present by @mebristo in #509
• ENH: Upgrade Github runner OS from Ubuntu 18 to Ubuntu 20 by @ant0nsc in #511
• ENH: Rename histopathology folder to match overall naming scheme by @ant0nsc in #512
• ENH: Update PR titles to include more tags by @SangamSwadiK in #505
• DOC: Use imperative for PR title in guidelines by @fepegar in #517
• ENH: Only upload codecov reports once by @ant0nsc in #521
• ENH: Add github workflow for smoke testing DeepSMILEPandaSlidesBenchmark by @mebristo in #506
• ENH: Split run requirements for cpath folder into file by @ant0nsc in #519
• ENH: Update tiling script by @maxilse in #516
• ENH: Publishing the hi-ml-cpath package by @ant0nsc in #524

New Contributors

• @pre-commit-ci made their first contribution in #488
• @SangamSwadiK made their first contribution in #505

Full Changelog: v0.2.3...v0.2.4

Dependency upgrades + bug fixes

What's Changed

• ENH: Add DeepSMILESlidesPandaBenchmark config by @mebristo in #385
• ENH: HTML reports for multiclass case by @harshita-s in #376
• ENH: Add utils to upload annotations to the Digital Slide Archive by @fepegar in #373
• ENH: Upload Codecov results also when tests fail by @ant0nsc in #383
• BUG: Remove padding when loading slides by @fepegar in #387
• ENH: Add visualization parameters to DSA annotations script by @fepegar in #390
• BUG: TilesPandaImageNetMIL runs hang on Azure ML by @harshita-s in #400
• STYLE: Disable pre-commit CI autofixes by @fepegar in #402
• ENH: Add Regression Tests by @mebristo in #359
• STYLE: Fix max line length in autopep8 hook by @fepegar in #403
• ENH: Select top and bottom tiles on the fly by @kenza-bouzid in #388
• BUG: Update contributing link in Readme by @kenza-bouzid in #408
• STYLE: Fix file without end line by @fepegar in #409
• ENH: PANDA SSL with tile selection enabled (random or tile occupancy) by @harshita-s in #398
• ENH: Add gpu tests coverage to codecov reports by @kenza-bouzid in #399
• DEL: Delete deprecated code from outputs utils by @kenza-bouzid in #407
• STYLE: Ensure new line at the end of JSON files by @fepegar in #410
• STYLE: Rename pl_limit_batches to pl_fast_dev_run by @kenza-bouzid in #415
• DEL: Remove TilesPANDA regression test by @mebristo in #418
• BUG: Correctly setting up encoder for SSLMIL by @harshita-s in #404
• ENH: Pin msrest to avoid missing import by @mebristo in #424
• DOC: Change outdated model name by @ant0nsc in #419
• STYLE: Set default PR title for pre-commit CI bot by @fepegar in #413
• DOC: Adding a high-level description of our software design process by @ant0nsc in #380
• ENH: Pin cloudpickle to fix warnings in Azure ML by @mebristo in #426
• BUG: Add path resolution in dataset mounting script by @kenza-bouzid in #416
• DEL: Retire the --azureml flag of the runner by @kenza-bouzid in #420
• DOC: Fix out of sync docs and typo by @kenza-bouzid in #428
• DOC: Document Histopathology Models by @ant0nsc in #429
• ENH: Simplify how we handle conda environment files in AML submissions by @ant0nsc in #414
• ENH: Add Credential Scanning Workflow by @ant0nsc in #370
• ENH: Reduce console logging noise by @ant0nsc in #430
• ENH: Add Cohen Kappa module metric to DeepMIL by @harshita-s in #422
• BUG: Stop looking for environment file if already running in AML by @mebristo in #432
• DOC: Remove retired argument names by @ant0nsc in #435
• ENH: Add unit tests for disabling top and bottom tiles by @kenza-bouzid in #439
• BUG: Default parameters in DeepSMILECRCk container by @harshita-s in #440
• ENH: Add makefile commands for smoke tests by @mebristo in #442
• ENH: Enable setting plot options in the containers config by @kenza-bouzid in #441
• BUG: Workflows time out because of using a too large test dataset by @ant0nsc in #447
• ENH: Improve mypy arguments by @fepegar in #448
• ENH: Redirect profiler traces to outputs folder and add debugging doc by @kenza-bouzid in #444
• ENH: Make building Monai-based models reproducible by @ant0nsc in #443
• ENH: Switch from azureml-sdk to azureml-core by @ant0nsc in #455
• BUG: Fix confusion matrix plots by @kenza-bouzid in #450
• ENH: Add text and image inference code for CXR multimodal by @ozan-oktay in #438
• ENH: Upgrade PyTorch Lightning to 1.6.4 by @kenza-bouzid in #453
• ENH: Remove the dependency on Lightning and Lightning Bolts by @ozan-oktay in #466
• ENH: Add Dependabot config file by @fepegar in #462
• ENH: Rename "multimodal" to "hi-ml-multimodal" by @fepegar in #467
• ENH: Enable pre-commit CI autofixes by @fepegar in #474
• ENH: Run an additional validation epoch at the end of training. by @kenza-bouzid in #465
• ENH: Add flake8 requirement to histology Conda environment by @ant0nsc in #454
• ENH: Update azureml packages versions. by @kenza-bouzid in #477

New Contributors

• @ozan-oktay made their first contribution in #438

Full Changelog: v0.2.2...v0.2.3

Hotfix for Protobuf version problem

What's Changed

• BUG: Fix protobuf version by @ant0nsc in #382
• ENH: Relax constraints on torchvision by @ant0nsc in #379

Full Changelog: v0.2.1...v0.2.2