ViT weight initialization, original timm impl (for reproducibility)

Creates a tensor of shape [num, *start.shape] whose values are evenly spaced from start to end, inclusive. Replicates but the multi-dimension

List all files recursively in a given directory while ignoring given file and directory names. Returns list of tuples containing both absolute and

:param input_im (PIL Image). :return image (H, W, 3) array in [0, 1].

https://github.com/openai/guided-diffusion/blob/27c20a8fab9cb472df5d6bdd6c8d11c8f430b924/guided_diffusion/nn.py#L86 Take the mean over all no

Normalize vector lengths.

Download the given URL and return a binary-mode file object to access the data.

Expects coordinates in shape (batch_size, num_points_per_batch, 3) Expects grid in shape (1, channels, H, W, D) (Also works if grid has b

SSIM only defined at intensity channel. For RGB or YUV or other image format, this function computes SSIm at each channel and averge them. :pa

timing_decorator: record time

Dot product of two tensors.

Left-multiplies MxM @ NxM. Returns NxM.

Calculate the product of the tuple elements.

Zero out the parameters of a module and return it.