Nicolas Hug
Nicolas Hug
If we're going to expose them as public, we need unit tests for them.
https://github.com/pytorch/vision/pull/8647 introduced a new way of determining whether the code is running internally. We should remove the existing https://github.com/pytorch/vision/blob/7eb09df3901091b08882c1b6df82fcbd0b710cd3/test/common_utils.py#L29 to avoid any confusion. (That one is only used in tests).
Dear contributors and users, I wanted to share an update on what we’ll be prioritizing in TorchVision from now on. In short:  In details: - The `transforms.v2` module, and...
The functional does support `inplace`: https://github.com/pytorch/vision/blob/f7097666f0241548d354b338552e7b3c0d2956cd/torchvision/transforms/v2/functional/_meta.py#L200-L205 But for some reason the class doesn't. I don't remember if there was any limitation to this, but we should try. CC @antoinesimoulin
Thanks to @antoinesimoulin for uncovering this issue in https://github.com/pytorch/vision/pull/8841/commits/5321f234fdac29fa3c1c708dfdeeb9e0b1ae5e0a ```python old_format = tv_tensors.BoundingBoxFormat.CXCYWH new_format = tv_tensors.BoundingBoxFormat.XYXY dtype = torch.int64 fn_type = "functional" device = torch.device("cpu") # bounding_boxes = make_bounding_boxes(format=old_format, dtype=dtype,...