amyeroberts
amyeroberts
# What does this PR do? This is the first of a series of PRs to replace feature extractors with image processors for vision models. Create a new module `image_transforms.py`...
# What does this PR do? At the moment, the return type of our feature extractors isn't always as expected or sometimes fails if a `do_xxx` config flag is set...
# What does this PR do? Adds the TF equivalent for the PyTorch image classification example script. ## Before submitting - [ ] This PR fixes a typo or improves...
# What does this PR do? Add the missing `bool_masked_pos` information in the docstring for vision models. Fixes #21484 ## Before submitting - [x] This PR fixes a typo or...
# What does this PR do? Updates feature selection to ensure returned dataset structure is consistent after merging of datasets PR: https://github.com/huggingface/datasets/pull/5602. The PR makes it possible to return a...
### Model description Add the TensorFlow port of the SwiftFormer model. See related issue: #22685 To be done once the SwiftFormer model has been added: #22686 ### Open source status...
# What does this PR do? `out_features` and `out_indices` are two parameters which control the behaviour of a backbone. `out_indices` was recently [added as a config argument](https://github.com/huggingface/transformers/pull/22493) for the future...
# What does this PR do? Adds a new class `ImageObject` for use as part of the image processing pipeline. It acts as an array container, which we can use...
# What does this PR do? Adds accelerate support to VideoMAE and ViTMAE following the changes made in the [equivalent ViT PR](https://github.com/huggingface/transformers/pull/20174) Fixes #23086 ## Before submitting - [ ]...
# What does this PR do? Abstracts out cropping logic to be a more generic `crop` function which other, more specific cropping functions e.g. `center_crop` can call. Motivation: * The...