Fused `Conv2d()` followed by `downsample_2d()`. Padding is performed only once at the beginning, not between the operations. The fused op is c

r""" Encodes the prompt into text encoder hidden states. Args: prompt (`str` or `List[str]`, *optional*):

extract features from the model :param batch: batch to extract features for. Has shape BxCxHxW. :param layers: layer to extra

Generate a dataset from evaluation data. Args: data (List[Dict[str, Any]]): The data to generate a dataset for. confi

Helper that gets a list of the model's attention modules. Each module has a `bias` buffer used for causal masking. The Prefix LM conv

Initialize the step_index counter for the scheduler.

Optionally removes offloading in case the pipeline has been already sequentially offloaded to CPU. Args: _pipeline (`Dif

Translates coordinates in pixel space to coordinates in latent space

Reorders the elements in the array based on the sorting function. Parameters: - arr (Union[List, Tuple[Tuple[int, Any], ...]

Swap the `Transformer2DModel` blocks between the image and text UNets

"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

"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