2914 release note, and what's new for v0.7

CHANGELOG.md

@@ -5,7 +5,50 @@ The format is based on [Keep a Changelog](http://keepachangelog.com/en/1.0.0/)
 and this project adheres to [Semantic Versioning](http://semver.org/spec/v2.0.0.html).
 
 ## [Unreleased]
-* renamed model's `n_classes` to `num_classes`
+
+## [0.7.0] - 2021-09-24
+### Added
+* Overview of [new features in v0.7](docs/source/whatsnew_0_7.md)
+* Initial phase of major usability improvements in `monai.transforms` to support input and backend in PyTorch and NumPy
+* Performance enhancements, with [profiling and tuning guides](https://github.com/Project-MONAI/tutorials/blob/master/acceleration/fast_model_training_guide.md) for typical use cases
+* Reproducing [training modules and workflows](https://github.com/Project-MONAI/tutorials/tree/master/kaggle/RANZCR/4th_place_solution) of state-of-the-art Kaggle competition solutions
+* 24 new transforms, including
+  * `OneOf` meta transform
+  * DeepEdit guidance signal transforms for interactive segmentation
+  * Transforms for self-supervised pre-training
+  * Integration of [NVIDIA Tools Extension](https://developer.nvidia.com/blog/nvidia-tools-extension-api-nvtx-annotation-tool-for-profiling-code-in-python-and-c-c/) (NVTX)
+  * Integration of [cuCIM](https://github.com/rapidsai/cucim)
+  * Stain normalization and contextual grid for digital pathology
+* `Transchex` network for vision-language transformers for chest X-ray analysis
+* `DatasetSummary` utility in `monai.data`
+* `WarmupCosineSchedule`
+* Deprecation warnings and documentation support for better backwards compatibility
+* Padding with additional `kwargs` and different backend API
+* Additional options such as `dropout` and `norm` in various networks and their submodules
+
+### Changed
+* Base Docker image upgraded to `nvcr.io/nvidia/pytorch:21.08-py3` from `nvcr.io/nvidia/pytorch:21.06-py3`
+* Deprecated input argument `n_classes`, in favor of `num_classes`
+* Deprecated input argument `dimensions` and `ndims`, in favor of `spatial_dims`
+* Updated the Sphinx-based documentation theme for better readability
+* `NdarrayTensor` type is replaced by `NdarrayOrTensor` for simpler annotations
+* Attention-based network blocks now support both 2D and 3D inputs
+
+### Removed
+* The deprecated `TransformInverter`, in favor of `monai.transforms.InvertD`
+* GitHub self-hosted CI/CD pipelines for nightly and post-merge tests
+* `monai.handlers.utils.evenly_divisible_all_gather`
+* `monai.handlers.utils.string_list_all_gather`
+
+### Fixed
+* A Multi-thread cache writing issue in `LMDBDataset`
+* Output shape convention inconsistencies of the image readers
+* Output directory and file name flexibility issue for `NiftiSaver`, `PNGSaver`
+* Requirement of the `label` field in test-time augmentation
+* Input argument flexibility issues for  `ThreadDataLoader`
+* Decoupled `Dice` and `CrossEntropy` intermediate results in `DiceCELoss`
+* Improved documentation, code examples, and warning messages in various modules
+* Various usability issues reported by users
 
 ## [0.6.0] - 2021-07-08
 ### Added
@@ -25,6 +68,7 @@ and this project adheres to [Semantic Versioning](http://semver.org/spec/v2.0.0.
 * Fully compatible with PyTorch 1.9
 * `--disttests` and `--min` options for `runtests.sh`
 * Initial support of pre-merge tests with Nvidia Blossom system
+
 ### Changed
 * Base Docker image upgraded to `nvcr.io/nvidia/pytorch:21.06-py3` from
   `nvcr.io/nvidia/pytorch:21.04-py3`
@@ -34,11 +78,13 @@ and this project adheres to [Semantic Versioning](http://semver.org/spec/v2.0.0.
 * Unified the terms: `post_transform` is renamed to `postprocessing`, `pre_transform` is renamed to `preprocessing`
 * Unified the postprocessing transforms and event handlers to accept the "channel-first" data format
 * `evenly_divisible_all_gather` and `string_list_all_gather` moved to `monai.utils.dist`
+
 ### Removed
 * Support of 'batched' input for postprocessing transforms and event handlers
 * `TorchVisionFullyConvModel`
 * `set_visible_devices` utility function
 * `SegmentationSaver` and `TransformsInverter` handlers
+
 ### Fixed
 * Issue of handling big-endian image headers
 * Multi-thread issue for non-random transforms in the cache-based datasets
@@ -269,9 +315,11 @@ the postprocessing steps should be used before calling the metrics methods
 * Optionally depend on PyTorch-Ignite v0.4.2 instead of v0.3.0
 * Optionally depend on torchvision, ITK
 * Enhanced CI tests with 8 new testing environments
+
 ### Removed
 * `MONAI/examples` folder (relocated into [`Project-MONAI/tutorials`](https://github.com/Project-MONAI/tutorials))
 * `MONAI/research` folder (relocated to [`Project-MONAI/research-contributions`](https://github.com/Project-MONAI/research-contributions))
+
 ### Fixed
 * `dense_patch_slices` incorrect indexing
 * Data type issue in `GeneralizedWassersteinDiceLoss`
@@ -302,6 +350,7 @@ the postprocessing steps should be used before calling the metrics methods
 * Cross-platform CI tests supporting multiple Python versions
 * Optional import mechanism
 * Experimental features for third-party transforms integration
+
 ### Changed
 > For more details please visit [the project wiki](https://github.com/Project-MONAI/MONAI/wiki/Notable-changes-between-0.1.0-and-0.2.0)
 * Core modules now require numpy >= 1.17
@@ -311,9 +360,11 @@ the postprocessing steps should be used before calling the metrics methods
 * Base Docker image upgraded to `nvcr.io/nvidia/pytorch:20.03-py3` from `nvcr.io/nvidia/pytorch:19.10-py3`
 * Enhanced local testing tools
 * Documentation website domain changed to https://docs.monai.io
+
 ### Removed
 * Support of Python < 3.6
 * Automatic installation of optional dependencies including pytorch-ignite, nibabel, tensorboard, pillow, scipy, scikit-image
+
 ### Fixed
 * Various issues in type and argument names consistency
 * Various issues in docstring and documentation site
@@ -336,7 +387,8 @@ the postprocessing steps should be used before calling the metrics methods
 
 [highlights]: https://github.com/Project-MONAI/MONAI/blob/master/docs/source/highlights.md
 
-[Unreleased]: https://github.com/Project-MONAI/MONAI/compare/0.6.0...HEAD
+[Unreleased]: https://github.com/Project-MONAI/MONAI/compare/0.7.0...HEAD
+[0.7.0]: https://github.com/Project-MONAI/MONAI/compare/0.6.0...0.7.0
 [0.6.0]: https://github.com/Project-MONAI/MONAI/compare/0.5.3...0.6.0
 [0.5.3]: https://github.com/Project-MONAI/MONAI/compare/0.5.0...0.5.3
 [0.5.0]: https://github.com/Project-MONAI/MONAI/compare/0.4.0...0.5.0

docs/source/whatsnew.rst

@@ -6,5 +6,6 @@ What's New
 .. toctree::
    :maxdepth: 1
 
+   whatsnew_0_7.md
    whatsnew_0_6.md
    whatsnew_0_5.md

docs/source/whatsnew_0_6.md

@@ -1,4 +1,4 @@
-# What's new in 0.6 🎉🎉
+# What's new in 0.6
 
 - Decollating mini-batches as an essential post-processing step
 - Pythonic APIs to load the pretrained models from Clara Train MMARs

docs/source/whatsnew_0_7.md

@@ -0,0 +1,62 @@
+# What's new in 0.7 🎉🎉
+
+- Performance enhancements with profiling and tuning guides
+- Major usability improvements in `monai.transforms`
+- Reimplementing state-of-the-art Kaggle solutions
+- Vision-language multimodal transformer architectures
+
+## Performance enhancements with profiling and tuning guides
+
+Model training is often a time-consuming step during deep learning development,
+especially for medical imaging applications. Even with powerful hardware (e.g.
+CPU/GPU with large RAM), the workflows often require careful profiling and
+tuning to achieve high performance. MONAI has been focusing on performance
+enhancements, and in this version, [a fast model training
+guide](https://github.com/Project-MONAI/tutorials/blob/master/acceleration/fast_model_training_guide.md)
+is provided to help build highly performant workflows, with a comprehensive
+overview of the profiling tools and practical strategies.  The following figure
+shows the use of [Nvidia Nsight™ Systems](https://developer.nvidia.com/nsight-systems) for system-wide performance analysis during
+a performance enhancement study.
+![nsight_vis](../images/nsight_comparison.png)
+
+With the performance profiling and enhancements, several typical use cases were studied to
+improve the training efficiency.  The following figure shows that fast
+training using MONAI can be 20 times faster than a regular baseline ([learn
+more](https://github.com/Project-MONAI/tutorials/blob/master/acceleration/fast_training_tutorial.ipynb)).
+![fast_training](../images/fast_training.png)
+
+## Major usability improvements in `monai.transforms` for NumPy/PyTorch inputs and backends
+
+ MONAI starts to roll out major usability enhancements for the
+ `monai.transforms` module. Many transforms are now supporting both NumPy and
+ PyTorch, as input types and computational backends.
+
+One benefit of these enhancements is that the users can now better leverage the
+GPUs for preprocessing. By transferring the input data onto GPU using
+`ToTensor` or `EnsureType`, and applying the GPU-based transforms to the data,
+[the tutorial of spleen
+segmentation](https://github.com/Project-MONAI/tutorials/blob/master/acceleration/fast_training_tutorial.ipynb)
+shows the great potential of using the flexible modules for fast and efficient
+training.
+
+## Reimplementing state-of-the-art Kaggle solutions
+
+With this release, we actively evaluate and enhance the quality and flexibility
+of the MONAI core modules, using the public Kaggle challenge as a testbed. [A
+reimplementation](https://github.com/Project-MONAI/tutorials/tree/master/kaggle/RANZCR/4th_place_solution)
+of a state-of-the-art solution at [Kaggle RANZCR CLiP - Catheter and Line
+Position
+Challenge](https://www.kaggle.com/c/ranzcr-clip-catheter-line-classification)
+is made available in this version.
+
+##  Vision-language multimodal transformers
+
+In this release, MONAI adds support for training multimodal (vision + language)
+transformers that can handle both image and textual data. MONAI introduces the
+`TransCheX` model which consists of vision, language, and mixed-modality
+transformer layers for processing chest X-ray and their corresponding
+radiological reports within a unified framework. In addition to `TransCheX`,
+users have the flexibility to alter the architecture by varying the number of
+vision, language and mixed-modality layers and customizing the classification
+head. In addition, the model can be initialized from pre-trained BERT language
+models for fine-tuning.