This does the final conversion step: take locally converted weights and apply a global renaming to them. It splits attention layers, and take

Updates paths inside resnets to the new naming scheme (local renaming)

Removes segments. Positive values shave the first segments, negative shave the last segments.

Takes a state dict and a config, and returns a converted checkpoint.

Export a list of frames to a GIF. Args: - frames (list): List of frames (as numpy arrays or PIL Image objects). - output_gif_path (s

A helper function to create random tensors on the desired `device` with the desired `dtype`. When passing a list of generators, you can seed each

Updates paths inside resnets to the new naming scheme (local renaming)

Predict the sample from the previous timestep by reversing the SDE. This function propagates the diffusion process from the learned m

Updates paths inside attentions to the new naming scheme (local renaming)

Initializes Unit3D module.

Constructs the noise schedule of Karras et al. (2022).

Estimate a covariance matrix given data. Covariance indicates the level to which two variables vary together. If we examine N-dimensional sam

Updates paths inside attentions to the new naming scheme (local renaming)

Gets scaled timesteps from the Karras sigmas for input to the consistency model. Args: sigmas (`float` or `np.ndarray`):

Appends dimensions to the end of a tensor until it has target_dims dimensions.

Compute padding tuple.

Compute padding tuple.

"Dynamic thresholding: At each sampling step we set s to a certain percentile absolute pixel value in xt0 (the prediction of x_0 at t

Appends dimensions to the end of a tensor until it has target_dims dimensions.

Create a 1D, 2D, or 3D average pooling module.

Create a beta schedule that discretizes the given alpha_t_bar function, which defines the cumulative product of (1-beta) over time from t = [

Create a 1D, 2D, or 3D convolution module.

Creates a config for the diffusers based on the config of the LDM model.

Gets the scalings used in the consistency model parameterization (from Appendix C of the [paper](https://huggingface.co/papers/2303.0

This matches the implementation in Denoising Diffusion Probabilistic Models: Create sinusoidal timestep embeddings. :param timesteps: a 1-D