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

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

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

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

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

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

Update d coefficients.

Update the latent according to the computed loss.

Fused `upsample_2d()` followed by `Conv2d()`. Padding is performed only once at the beginning, not between the operations. The fused op is

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 =

Returns list of utf-8 byte and a corresponding list of unicode strings. The reversible bpe codes work on unicode strings. This means y

Calculate performance for Clotho+AudioCaps for model selection.