Parses a string of key-value pairs separated by spaces. :param text: String of key-value pairs separated by spaces. :return: List of key

Allow calculations to run in mixed precision, which can be faster.

Evaluate a function without caching intermediate activations, allowing for reduced memory at the expense of extra compute in the backward pas

Performs all_gather operation on the provided tensors. *** Warning ***: torch.distributed.all_gather has no gradient.

Downloads weights from huggingface and returns the path to the downloaded file Given a huggingface repo url, folder, and optional filename,

Calculates the noise level (sigma_down) to step down to and the amount of noise to add (sigma_up) when doing an ancestral sampling step.

Ref:https://github.com/1adrianb/face-alignment/blob/master/face_alignment/utils.py

Trace the execution order of a torch module and store full hierarchical state_dict paths. :param module: The module to trace. :param args

Not yet fully cleaned from https://github.com/hhatto/smartcrop.py.

Given an image, return a caption.

Gets a python object from a string reference if it's location. Example: "functools.lru_cache"

Generates and saves image files based on given prompts, with options for animations and videos. Args: prompts (list[ImaginePrompt] |

Resolve weight and config path if they happen to be shortcuts.

Create sinusoidal timestep embeddings. :param timesteps: a 1-D Tensor of N indices, one per batch element. These may be

Upscales an image using a specified super-resolution model. It accepts an image in various forms: a LazyLoadingImage instance, a PIL Image,

Infer the type of a basic layer.

3x3 convolution with padding

decodes the tensor in slices. This results in images that don't exactly match, so we overlap, feather, and merge to reduce (

Replaces {vars} with random samples of corresponding phraselists. Example: p = "a happy {animal}" prompts = expand_prompts(p

Load a diffusion model. Weights location may also be shortcut name, e.g. "SD-1.5"