Lightning-bolts

Added

- Improved YOLO model includes YOLOv4, YOLOv5, and YOLOX networks and training algorithms ([817](https://github.com/PyTorchLightning/pytorch-lightning-bolts/pull/817))

Changed

- Move SSL transforms to pl_bolts/transforms ([905](https://github.com/PyTorchLightning/pytorch-lightning-bolts/pull/905))
- Reviewed `models.detection.yolo` ([851](https://github.com/PyTorchLightning/pytorch-lightning-bolts/pull/851))
- Reviewed `LogisticRegression` ([950](https://github.com/PyTorchLightning/pytorch-lightning-bolts/pull/950))
- Bumped support of min python version to py3.8+ ([1021](https://github.com/PyTorchLightning/pytorch-lightning-bolts/pull/1021))
- Update `numpy` compatibility to <1.25.0 ([959](https://github.com/PyTorchLightning/pytorch-lightning-bolts/pull/959))
- Update `torchmetrics` compatibility to <0.12.0 ([1016](https://github.com/PyTorchLightning/pytorch-lightning-bolts/pull/1016))
- Update `pytorch-lightning` compatibility to >1.7.0,<2.0.0 (
[965](https://github.com/PyTorchLightning/pytorch-lightning-bolts/pull/965),
[973](https://github.com/PyTorchLightning/pytorch-lightning-bolts/pull/973),
[1006](https://github.com/PyTorchLightning/pytorch-lightning-bolts/pull/1006)
)

Fixed

- Dropped reference to `torch._six` ([993](https://github.com/PyTorchLightning/pytorch-lightning-bolts/pull/993))

Added

- Updated SparseML callback for latest PyTorch Lightning ([822](https://github.com/Lightning-AI/lightning-bolts/pull/822))
- Updated torch version to v1.10.X ([815](https://github.com/Lightning-AI/lightning-bolts/pull/815))
- Dataset specific args method to CIFAR10, ImageNet, MNIST, and STL10 ([890](https://github.com/Lightning-AI/lightning-bolts/pull/890))
- Migrate to use `lightning-utilities` ([907](https://github.com/Lightning-AI/lightning-bolts/pull/907))
- Support PyTorch Lightning v1.8 ([910](https://github.com/Lightning-AI/lightning-bolts/pull/910))

- Major revision of Bolts
* `under_review` flag ([835](https://github.com/Lightning-AI/lightning-bolts/pull/835), [#837](https://github.com/Lightning-AI/lightning-bolts/pull/837))
* Reviewing GAN basics, `VisionDataModule`, `MNISTDataModule`, `CIFAR10DataModule` ([843](https://github.com/Lightning-AI/lightning-bolts/pull/843))
* Added tests, updated doc-strings for Dummy Datasets ([865](https://github.com/Lightning-AI/lightning-bolts/pull/865))
* Binary MNIST/EMNIST Datasets and Datamodules ([866](https://github.com/Lightning-AI/lightning-bolts/pull/866))
* FashionMNIST/EMNIST Datamodules ([871](https://github.com/Lightning-AI/lightning-bolts/pull/871))
* Revision `ArrayDataset` ([872](https://github.com/Lightning-AI/lightning-bolts/pull/872))
* BYOL weight update callback ([867](https://github.com/Lightning-AI/lightning-bolts/pull/867))
* Revision `models.vision.unet`, `models.vision.segmentation` ([880](https://github.com/Lightning-AI/lightning-bolts/pull/880))
* Revision of SimCLR transforms ([857](https://github.com/Lightning-AI/lightning-bolts/pull/857))
* Revision Metrics ([878](https://github.com/Lightning-AI/lightning-bolts/pull/878), [#887](https://github.com/Lightning-AI/lightning-bolts/pull/887))
* Revision of BYOL module and tests ([874](https://github.com/Lightning-AI/lightning-bolts/pull/874))
* Revision of MNIST module ([873](https://github.com/Lightning-AI/lightning-bolts/pull/873))
* Revision of dataset normalizations ([898](https://github.com/Lightning-AI/lightning-bolts/pull/898))
* Revision of SimSiam module and tests ([891](https://github.com/Lightning-AI/lightning-bolts/pull/891))
* Revision `datasets.kitti_dataset.KittiDataset` ([896](https://github.com/Lightning-AI/lightning-bolts/pull/896))
* SWAV improvements ([903](https://github.com/Lightning-AI/lightning-bolts/pull/903))
* minor dcgan-import fix ([921](https://github.com/Lightning-AI/lightning-bolts/pull/921))

Fixed

- Removing extra flatten ([809](https://github.com/Lightning-AI/lightning-bolts/pull/809))
- support number of channels!=3 in YOLOConfiguration ([806](https://github.com/Lightning-AI/lightning-bolts/pull/806))
- CVE-2007-4559 Patch ([894](https://github.com/Lightning-AI/lightning-bolts/pull/894))

Added

- Added YOLO model ([552](https://github.com/Lightning-AI/lightning-bolts/pull/552))
- Added `SRGAN`, `SRImageLoggerCallback`, `TVTDataModule`, `SRCelebA`, `SRMNIST`, `SRSTL10` ([466](https://github.com/Lightning-AI/lightning-bolts/pull/466))
- Added nn.Module support for FasterRCNN backbone ([661](https://github.com/Lightning-AI/lightning-bolts/pull/661))
- Added `RetinaNet` with torchvision backbones ([529](https://github.com/Lightning-AI/lightning-bolts/pull/529))
- Added Python 3.9 support ([786](https://github.com/Lightning-AI/lightning-bolts/pull/786))

Changed

- VAE now uses deterministic KL divergence during training, previously estimated KL divergence by random sampling ([760](https://github.com/Lightning-AI/lightning-bolts/pull/760))

Removed

- Removed PyTorch 1.6 support ([786](https://github.com/Lightning-AI/lightning-bolts/pull/786))
- Removed Python 3.6 support ([785](https://github.com/Lightning-AI/lightning-bolts/pull/785))

Fixed

- Fixed doctest fails with ImportError: cannot import name 'Env' from 'gym' ([751](https://github.com/Lightning-AI/lightning-bolts/pull/751))
- Fixed MoCo v2 missing Cosine Annealing learning scheduler ([757](https://github.com/Lightning-AI/lightning-bolts/pull/757))

Added

- Added Soft Actor Critic (SAC) Model ([627](https://github.com/Lightning-AI/lightning-bolts/pull/627))
- Added `EMNISTDataModule`, `BinaryEMNISTDataModule`, and `BinaryEMNIST` dataset ([676](https://github.com/Lightning-AI/lightning-bolts/pull/676))
- Added Advantage Actor-Critic (A2C) Model ([598](https://github.com/Lightning-AI/lightning-bolts/pull/598))
- Added Torch ORT Callback ([720](https://github.com/Lightning-AI/lightning-bolts/pull/720))
- Added SparseML Callback ([724](https://github.com/Lightning-AI/lightning-bolts/pull/724))

Changed

- Changed the default values `pin_memory=False`, `shuffle=False` and `num_workers=16` to `pin_memory=True`, `shuffle=True` and `num_workers=0` of datamodules ([701](https://github.com/Lightning-AI/lightning-bolts/pull/701))
- Supporting deprecated attribute usage ([699](https://github.com/Lightning-AI/lightning-bolts/pull/699))

Fixed

- Fixed ImageNet val loader to use val transform instead of train transform ([713](https://github.com/Lightning-AI/lightning-bolts/pull/713))
- Fixed the MNIST download giving HTTP 404 with `torchvision>=0.9.1` ([674](https://github.com/Lightning-AI/lightning-bolts/pull/674))
- Removed momentum updating from val step and add separate val queue ([631](https://github.com/Lightning-AI/lightning-bolts/pull/631))
- Fixed moving the queue to GPU when resuming checkpoint for SwAV model ([684](https://github.com/Lightning-AI/lightning-bolts/pull/684))
- Fixed FP16 support with vision GPT model ([694](https://github.com/Lightning-AI/lightning-bolts/pull/694))
- Removing bias from linear model regularisation ([669](https://github.com/Lightning-AI/lightning-bolts/pull/669))
- Fixed CPC module issue ([680](https://github.com/Lightning-AI/lightning-bolts/pull/680))

Changed

- Replaced `load_boston` with `load_diabetes` in the docs and tests ([629](https://github.com/Lightning-AI/lightning-bolts/pull/629))
- Added base encoder and MLP dimension arguments to BYOL constructor ([637](https://github.com/Lightning-AI/lightning-bolts/pull/637))

Fixed

- Fixed the MNIST download giving HTTP 503 ([633](https://github.com/Lightning-AI/lightning-bolts/pull/633))
- Fixed type annotation of `ExperienceSource.__iter__` ([645](https://github.com/Lightning-AI/lightning-bolts/pull/645))
- Fixed `pretrained_urls` on Windows ([652](https://github.com/Lightning-AI/lightning-bolts/pull/652))
- Fixed logistic regression ([655](https://github.com/Lightning-AI/lightning-bolts/pull/655), [#664](https://github.com/Lightning-AI/lightning-bolts/pull/664))
- Fixed double softmax in `SSLEvaluator` ([663](https://github.com/Lightning-AI/lightning-bolts/pull/663))

Changed

- Suppressed missing package warnings, conditioned by `WARN_MISSING_PACKAGE="1"` ([617](https://github.com/Lightning-AI/lightning-bolts/pull/617))
- Updated all scripts to LARS ([613](https://github.com/Lightning-AI/lightning-bolts/pull/613))

Fixed

- Add missing `dataclass` requirements ([618](https://github.com/Lightning-AI/lightning-bolts/pull/618))