Philip Meier

This is #6517

> If i understand well, the issue is restricted to the edge of the image and the rest of the grid sampling/interpolation works correctly. Is that right? Correct. And to...

Here is adapted script to show the difference between nearest and bilinear: ```python import torch import torchvision from torchvision.transforms.v2.functional import affine from torchvision.tv_tensors import Image from torchvision.transforms import InterpolationMode image...

I've run a few quick benchmarks whether or not it is useful to compile kernels in the first place. I've used a simple classification pipeline (random_resized_crop, horizontal_flip, to_dtype, normalize) and...

`torch.compile` doesn't yet handle tensor subclasses. From this error message ``` Argument displacement shape should be (1, 1, 4, 2), but given torch.Size([1, 17, 11, 2]) ``` you can see...

There are two sources of graph breaks in the way we currently dispatch: 1. We use the dispatcher and the input type directly as dictionary keys: https://github.com/pytorch/vision/blob/15c166ac127db5c8d1541b3485ef5730d34bb68a/torchvision/transforms/v2/functional/_utils.py#L15-L16 This is currently...

One thing that I noticed while playing around with the benchmarks is that dynamo does not give us a strict improvement for individual ops. ### `random_resized_crop` ``` [------------------ -----------------] |...

One disadvantage of the solution provided in https://github.com/pydantic/pydantic/issues/8603#issuecomment-1904939403 is that the static annotation is lost, because it is set at runtime. Running `mypy` on ```python from typing import Annotated, Optional...

Improved version of https://github.com/pydantic/pydantic/issues/8603#issuecomment-2264988605 that also works with models that use `validate_default=True` by default, e.g. `pydantic_settings.BaseSettings`: ```python from __future__ import annotations from typing import TYPE_CHECKING, Any, Callable from pydantic import...

@asmeurer I've added you to the `module: python array api` as per request in https://github.com/pytorch/pytorch/issues/58743#issuecomment-859377579.