PiperOrigin-RevId: 463812468

PiperOrigin-RevId: 463764367

PiperOrigin-RevId: 463617057

PiperOrigin-RevId: 463467924

PiperOrigin-RevId: 463429471

https://github.com/tensorflow/models/issues/10074

PiperOrigin-RevId: 463195812

PiperOrigin-RevId: 463193774

PiperOrigin-RevId: 463166393

PiperOrigin-RevId: 463160317

PiperOrigin-RevId: 463012965

Move `output_range` argument to call(). After evaluating several use cases, `output_range` is better to be specified at `call` time which allows us the train with full sequence and serve with certain tokens.

PiperOrigin-RevId: 462885854

PiperOrigin-RevId: 462851231

PiperOrigin-RevId: 462680315

Fix attention dropout rate bug. The output_dropout rate was used for attention dropout by mistake.

PiperOrigin-RevId: 462470287

PiperOrigin-RevId: 462275036

PiperOrigin-RevId: 461964298

PiperOrigin-RevId: 461948867

PiperOrigin-RevId: 461738130

PiperOrigin-RevId: 461706948

Support ViT dynamic positional embedding interpolation. This is useful to finetune a model pre-trained on a different resolution.

PiperOrigin-RevId: 461666803

PiperOrigin-RevId: 461655578

Enable to construct decoder only in DETRTransformer and refactor DETR model. Also fix a typo in DetectionTask.

PiperOrigin-RevId: 461376875

PiperOrigin-RevId: 461254505

PiperOrigin-RevId: 461239167

PiperOrigin-RevId: 461063108

PiperOrigin-RevId: 460979064

PiperOrigin-RevId: 460912624

PiperOrigin-RevId: 460836806

PiperOrigin-RevId: 460830316

PiperOrigin-RevId: 460756776

PiperOrigin-RevId: 460731093

Adding FAVOR++ mechanism from the paper: "Chefs' Random Tables: Non-Trigonometric Random Features" to TF MODEL GARDEN.

PiperOrigin-RevId: 460625574

PiperOrigin-RevId: 460570094

PiperOrigin-RevId: 460499239

PiperOrigin-RevId: 460280458

PiperOrigin-RevId: 460263057

Allow setting hyperparameters of augmentation for video classification in the config, and extend MixupAndCutmix for videos.

PiperOrigin-RevId: 459615324

PiperOrigin-RevId: 459611611