Philip Meier
Philip Meier
The transforms in `torchvision.transforms` are fully JIT scriptable. To achieve this we needed a lot of helper methods and unnecessary strict or plain wrong annotations. The revamped transforms in `torchvision.prototype.transforms`...
If you don't set the respective env vars https://github.com/pytorch/vision/blob/d5bd8b728f14c33b339fc45c90ca39be339bce3f/references/classification/utils.py#L255-L258 training will not be distributed and in turn the backend will not be initialized. However, during evaluation we check https://github.com/pytorch/vision/blob/d5bd8b728f14c33b339fc45c90ca39be339bce3f/references/classification/train.py#L88 unguarded,...
This PR is just here to showcase what ops are not working with 2d image input. Although we might cherry-pick the utility changes later on. cc @vfdev-5
`torchvision` is currently building https://github.com/pytorch/vision/blob/cac4e228c9ca9e7564cb34406e7ebccfdd736976/setup.py#L321 https://github.com/pytorch/vision/blob/cac4e228c9ca9e7564cb34406e7ebccfdd736976/packaging/torchvision/meta.yaml#L13 and testing against `libjpeg` https://github.com/pytorch/vision/blob/cac4e228c9ca9e7564cb34406e7ebccfdd736976/.circleci/unittest/linux/scripts/environment.yml#L10 https://github.com/pytorch/vision/blob/cac4e228c9ca9e7564cb34406e7ebccfdd736976/.circleci/unittest/windows/scripts/environment.yml#L10 `Pillow` is building against `libjpeg-turbo` on Windows for some time now since [`Pillow=9` on all platforms](https://pillow.readthedocs.io/en/stable/releasenotes/9.0.0.html#switched-to-libjpeg-turbo-in-macos-and-linux-wheels) (Jan 2022)....
### 🚀 The feature *Note: To track the progress of the project check out [this board](https://github.com/pytorch/vision/projects/5).* The current transforms in the `torchvision.transforms` name space are limited to images. This makes...
Most of our prototype transforms that override `forward` do this only to perform some input checks before passing them on to `_get_params` and `_transform`. This PR adds a `_check` method...
Our current implementation of `query_chw` requires an image somewhere in the sample, since that is the only way to extract the number of channels. https://github.com/pytorch/vision/blob/f9966d228c5b809b0b68671ef168004f1e06d3fe/torchvision/prototype/transforms/_utils.py#L39 However, there are quite a...
```py from torchvision.prototype import features from copy import copy, deepcopy import torch image = features.Image(torch.rand(3, 16, 16)) image_copy = copy(image) # works image_deepcopy = deepcopy(image) # fails ``` ``` RuntimeError:...
Fixes #206. Although the installation instructions install CUDA and a compatible PyTorch version https://github.com/libffcv/ffcv/blob/f25386557e213711cc8601833add36ff966b80b2/README.md?plain=1#L38 I didn't find a hint or the like that CUDA is required. The presence of a...
Although the coverage is 78%, the tests cover very little apart from the regular cases. A lot of `if` branches especially in error cases are not covered. Additionally, the following...