to_grayscale gives non-actionable deprecation warning

[pmeier]

Originally reported in the user forum by @function2.

When I use to_grayscale, there’s a deprecation warning:

UserWarning: The function `to_grayscale(...)` is deprecated in will be removed in a future release. Instead, please use `convert_color_space(..., color_space=datapoints.ColorSpace.GRAY)`.

However, I can’t find this function in the current code base

---

Note that this only applies to torchvision.transforms.v2.function

vision/torchvision/transforms/v2/functional/_deprecated.py

Lines 12 to 22 in 52eb503

	def to_grayscale(inpt: PIL.Image.Image, num_output_channels: int = 1) -> PIL.Image.Image:
	call = ", num_output_channels=3" if num_output_channels == 3 else ""
	replacement = "convert_color_space(..., color_space=datapoints.ColorSpace.GRAY)"
	if num_output_channels == 3:
	replacement = f"convert_color_space({replacement}, color_space=datapoints.ColorSpace.RGB)"
	warnings.warn(
	f"The function `to_grayscale(...{call})` is deprecated in will be removed in a future release. "
	f"Instead, please use `{replacement}`.",
	)
	return _F.to_grayscale(inpt, num_output_channels=num_output_channels)

since the v1 version, i.e. torchvision.transforms.functional does not emit the warning

vision/torchvision/transforms/functional.py

Lines 1249 to 1253 in 52eb503

	# Looks like to_grayscale() is a stand-alone functional that is never called
	# from the transform classes. Perhaps it's still here for BC? I can't be
	# bothered to dig. Anyway, this can be deprecated as we migrate to V2.
	@torch.jit.unused
	def to_grayscale(img, num_output_channels=1):

Fixing the v2 warning was forgotten in #7120.

cc @vfdev-5