diffusers/pr_13881/en/api/models/consistency_decoder_vae.md

Consistency Decoder

Consistency decoder can be used to decode the latents from the denoising UNet in the StableDiffusionPipeline. This decoder was introduced in the DALL-E 3 technical report.

The original codebase can be found at openai/consistencydecoder.

Inference is only supported for 2 iterations as of now.

The pipeline could not have been contributed without the help of madebyollin and mrsteyk from this issue.

ConsistencyDecoderVAE[[diffusers.ConsistencyDecoderVAE]]

The consistency decoder used with DALL-E 3.

Examples:

>>> import torch
>>> from diffusers import StableDiffusionPipeline, ConsistencyDecoderVAE

>>> vae = ConsistencyDecoderVAE.from_pretrained("openai/consistency-decoder", torch_dtype=torch.float16)
>>> pipe = StableDiffusionPipeline.from_pretrained(
...     "stable-diffusion-v1-5/stable-diffusion-v1-5", vae=vae, torch_dtype=torch.float16
... ).to("cuda")

>>> image = pipe("horse", generator=torch.manual_seed(0)).images[0]
>>> image

• z (torch.Tensor) -- The input latent vector.
• generator (torch.Generator | None) -- The random number generator. Default is None.
• return_dict (bool) -- Whether to return the output as a dictionary. Default is True.
• num_inference_steps (int) -- The number of inference steps. Default is 2.DecoderOutput | tuple[torch.Tensor]The decoded output.

Decodes the input latent vector z using the consistency decoder VAE model.

• x (torch.Tensor) -- Input batch of images.
• return_dict (bool, optional, defaults to True) -- Whether to return a ConsistencyDecoderVAEOutput instead of a plain tuple.The latent representations of the encoded images. If return_dict is True, a ConsistencyDecoderVAEOutput is returned, otherwise a plain tuple is returned.

Encode a batch of images into latents.

• sample (torch.Tensor) -- Input sample.
• sample_posterior (bool, optional, defaults to False) -- Whether to sample from the posterior.
• return_dict (bool, optional, defaults to True) -- Whether or not to return a DecoderOutput instead of a plain tuple.
• generator (torch.Generator, optional, defaults to None) -- Generator to use for sampling.DecoderOutput or tupleIf return_dict is True, a DecoderOutput is returned, otherwise a plain tuple is returned.

Disables custom attention processors and sets the default attention implementation.

• x (torch.Tensor) -- Input batch of images.
• return_dict (bool, optional, defaults to True) -- Whether or not to return a ConsistencyDecoderVAEOutput instead of a plain tuple.ConsistencyDecoderVAEOutput or tupleIf return_dict is True, a ConsistencyDecoderVAEOutput is returned, otherwise a plain tuple is returned. Encode a batch of images using a tiled encoder.

When this option is enabled, the VAE will split the input tensor into tiles to compute encoding in several steps. This is useful to keep memory use constant regardless of image size. The end result of tiled encoding is different from non-tiled encoding because each tile uses a different encoder. To avoid tiling artifacts, the tiles overlap and are blended together to form a smooth output. You may still see tile-sized changes in the output, but they should be much less noticeable.