diffusers/pr_12229/en/api/models/autoencoder_tiny.md

# Tiny AutoEncoder

Tiny AutoEncoder for Stable Diffusion (TAESD) was introduced in [madebyollin/taesd](https://github.com/madebyollin/taesd) by Ollin Boer Bohan. It is a tiny distilled version of Stable Diffusion's VAE that can quickly decode the latents in a [StableDiffusionPipeline](/docs/diffusers/pr_12229/en/api/pipelines/stable_diffusion/text2img#diffusers.StableDiffusionPipeline) or [StableDiffusionXLPipeline](/docs/diffusers/pr_12229/en/api/pipelines/stable_diffusion/stable_diffusion_xl#diffusers.StableDiffusionXLPipeline) almost instantly.

To use with Stable Diffusion v-2.1:

```python
import torch
from diffusers import DiffusionPipeline, AutoencoderTiny

pipe = DiffusionPipeline.from_pretrained(
    "stabilityai/stable-diffusion-2-1-base", torch_dtype=torch.float16
)
pipe.vae = AutoencoderTiny.from_pretrained("madebyollin/taesd", torch_dtype=torch.float16)
pipe = pipe.to("cuda")

prompt = "slice of delicious New York-style berry cheesecake"
image = pipe(prompt, num_inference_steps=25).images[0]
image
```

To use with Stable Diffusion XL 1.0

```python
import torch
from diffusers import DiffusionPipeline, AutoencoderTiny

pipe = DiffusionPipeline.from_pretrained(
    "stabilityai/stable-diffusion-xl-base-1.0", torch_dtype=torch.float16
)
pipe.vae = AutoencoderTiny.from_pretrained("madebyollin/taesdxl", torch_dtype=torch.float16)
pipe = pipe.to("cuda")

prompt = "slice of delicious New York-style berry cheesecake"
image = pipe(prompt, num_inference_steps=25).images[0]
image
```

## AutoencoderTiny[[diffusers.AutoencoderTiny]]

<div class="docstring border-l-2 border-t-2 pl-4 pt-3.5 border-gray-100 rounded-tl-xl mb-6 mt-8">


<docstring><name>class diffusers.AutoencoderTiny</name><anchor>diffusers.AutoencoderTiny</anchor><source>https://github.com/huggingface/diffusers/blob/vr_12229/src/diffusers/models/autoencoders/autoencoder_tiny.py#L41</source><parameters>[{"name": "in_channels", "val": ": int = 3"}, {"name": "out_channels", "val": ": int = 3"}, {"name": "encoder_block_out_channels", "val": ": typing.Tuple[int, ...] = (64, 64, 64, 64)"}, {"name": "decoder_block_out_channels", "val": ": typing.Tuple[int, ...] = (64, 64, 64, 64)"}, {"name": "act_fn", "val": ": str = 'relu'"}, {"name": "upsample_fn", "val": ": str = 'nearest'"}, {"name": "latent_channels", "val": ": int = 4"}, {"name": "upsampling_scaling_factor", "val": ": int = 2"}, {"name": "num_encoder_blocks", "val": ": typing.Tuple[int, ...] = (1, 3, 3, 3)"}, {"name": "num_decoder_blocks", "val": ": typing.Tuple[int, ...] = (3, 3, 3, 1)"}, {"name": "latent_magnitude", "val": ": int = 3"}, {"name": "latent_shift", "val": ": float = 0.5"}, {"name": "force_upcast", "val": ": bool = False"}, {"name": "scaling_factor", "val": ": float = 1.0"}, {"name": "shift_factor", "val": ": float = 0.0"}]</parameters><paramsdesc>- **in_channels** (`int`, *optional*, defaults to 3) -- Number of channels in the input image.
- **out_channels** (`int`,  *optional*, defaults to 3) -- Number of channels in the output.
- **encoder_block_out_channels** (`Tuple[int]`, *optional*, defaults to `(64, 64, 64, 64)`) --
  Tuple of integers representing the number of output channels for each encoder block. The length of the
  tuple should be equal to the number of encoder blocks.
- **decoder_block_out_channels** (`Tuple[int]`, *optional*, defaults to `(64, 64, 64, 64)`) --
  Tuple of integers representing the number of output channels for each decoder block. The length of the
  tuple should be equal to the number of decoder blocks.
- **act_fn** (`str`, *optional*, defaults to `"relu"`) --
  Activation function to be used throughout the model.
- **latent_channels** (`int`, *optional*, defaults to 4) --
  Number of channels in the latent representation. The latent space acts as a compressed representation of
  the input image.
- **upsampling_scaling_factor** (`int`, *optional*, defaults to 2) --
  Scaling factor for upsampling in the decoder. It determines the size of the output image during the
  upsampling process.
- **num_encoder_blocks** (`Tuple[int]`, *optional*, defaults to `(1, 3, 3, 3)`) --
  Tuple of integers representing the number of encoder blocks at each stage of the encoding process. The
  length of the tuple should be equal to the number of stages in the encoder. Each stage has a different
  number of encoder blocks.
- **num_decoder_blocks** (`Tuple[int]`, *optional*, defaults to `(3, 3, 3, 1)`) --
  Tuple of integers representing the number of decoder blocks at each stage of the decoding process. The
  length of the tuple should be equal to the number of stages in the decoder. Each stage has a different
  number of decoder blocks.
- **latent_magnitude** (`float`, *optional*, defaults to 3.0) --
  Magnitude of the latent representation. This parameter scales the latent representation values to control
  the extent of information preservation.
- **latent_shift** (float, *optional*, defaults to 0.5) --
  Shift applied to the latent representation. This parameter controls the center of the latent space.
- **scaling_factor** (`float`, *optional*, defaults to 1.0) --
  The component-wise standard deviation of the trained latent space computed using the first batch of the
  training set. This is used to scale the latent space to have unit variance when training the diffusion
  model. The latents are scaled with the formula `z = z * scaling_factor` before being passed to the
  diffusion model. When decoding, the latents are scaled back to the original scale with the formula: `z = 1
  / scaling_factor * z`. For more details, refer to sections 4.3.2 and D.1 of the [High-Resolution Image
  Synthesis with Latent Diffusion Models](https://huggingface.co/papers/2112.10752) paper. For this
  Autoencoder, however, no such scaling factor was used, hence the value of 1.0 as the default.
- **force_upcast** (`bool`, *optional*, default to `False`) --
  If enabled it will force the VAE to run in float32 for high image resolution pipelines, such as SD-XL. VAE
  can be fine-tuned / trained to a lower range without losing too much precision, in which case
  `force_upcast` can be set to `False` (see this fp16-friendly
  [AutoEncoder](https://huggingface.co/madebyollin/sdxl-vae-fp16-fix)).</paramsdesc><paramgroups>0</paramgroups></docstring>

A tiny distilled VAE model for encoding images into latents and decoding latent representations into images.

[AutoencoderTiny](/docs/diffusers/pr_12229/en/api/models/autoencoder_tiny#diffusers.AutoencoderTiny) is a wrapper around the original implementation of `TAESD`.

This model inherits from [ModelMixin](/docs/diffusers/pr_12229/en/api/models/overview#diffusers.ModelMixin). Check the superclass documentation for its generic methods implemented for
all models (such as downloading or saving).





<div class="docstring border-l-2 border-t-2 pl-4 pt-3.5 border-gray-100 rounded-tl-xl mb-6 mt-8">


<docstring><name>disable_slicing</name><anchor>diffusers.AutoencoderTiny.disable_slicing</anchor><source>https://github.com/huggingface/diffusers/blob/vr_12229/src/diffusers/models/autoencoders/autoencoder_tiny.py#L172</source><parameters>[]</parameters></docstring>

Disable sliced VAE decoding. If `enable_slicing` was previously enabled, this method will go back to computing
decoding in one step.


</div>
<div class="docstring border-l-2 border-t-2 pl-4 pt-3.5 border-gray-100 rounded-tl-xl mb-6 mt-8">


<docstring><name>disable_tiling</name><anchor>diffusers.AutoencoderTiny.disable_tiling</anchor><source>https://github.com/huggingface/diffusers/blob/vr_12229/src/diffusers/models/autoencoders/autoencoder_tiny.py#L187</source><parameters>[]</parameters></docstring>

Disable tiled VAE decoding. If `enable_tiling` was previously enabled, this method will go back to computing
decoding in one step.


</div>
<div class="docstring border-l-2 border-t-2 pl-4 pt-3.5 border-gray-100 rounded-tl-xl mb-6 mt-8">


<docstring><name>enable_slicing</name><anchor>diffusers.AutoencoderTiny.enable_slicing</anchor><source>https://github.com/huggingface/diffusers/blob/vr_12229/src/diffusers/models/autoencoders/autoencoder_tiny.py#L165</source><parameters>[]</parameters></docstring>

Enable sliced VAE decoding. When this option is enabled, the VAE will split the input tensor in slices to
compute decoding in several steps. This is useful to save some memory and allow larger batch sizes.


</div>
<div class="docstring border-l-2 border-t-2 pl-4 pt-3.5 border-gray-100 rounded-tl-xl mb-6 mt-8">


<docstring><name>enable_tiling</name><anchor>diffusers.AutoencoderTiny.enable_tiling</anchor><source>https://github.com/huggingface/diffusers/blob/vr_12229/src/diffusers/models/autoencoders/autoencoder_tiny.py#L179</source><parameters>[{"name": "use_tiling", "val": ": bool = True"}]</parameters></docstring>

Enable tiled VAE decoding. When this option is enabled, the VAE will split the input tensor into tiles to
compute decoding and encoding in several steps. This is useful for saving a large amount of memory and to allow
processing larger images.


</div>
<div class="docstring border-l-2 border-t-2 pl-4 pt-3.5 border-gray-100 rounded-tl-xl mb-6 mt-8">


<docstring><name>forward</name><anchor>diffusers.AutoencoderTiny.forward</anchor><source>https://github.com/huggingface/diffusers/blob/vr_12229/src/diffusers/models/autoencoders/autoencoder_tiny.py#L321</source><parameters>[{"name": "sample", "val": ": Tensor"}, {"name": "return_dict", "val": ": bool = True"}]</parameters><paramsdesc>- **sample** (`torch.Tensor`) -- Input sample.
- **return_dict** (`bool`, *optional*, defaults to `True`) --
  Whether or not to return a `DecoderOutput` instead of a plain tuple.</paramsdesc><paramgroups>0</paramgroups></docstring>




</div>
<div class="docstring border-l-2 border-t-2 pl-4 pt-3.5 border-gray-100 rounded-tl-xl mb-6 mt-8">


<docstring><name>scale_latents</name><anchor>diffusers.AutoencoderTiny.scale_latents</anchor><source>https://github.com/huggingface/diffusers/blob/vr_12229/src/diffusers/models/autoencoders/autoencoder_tiny.py#L157</source><parameters>[{"name": "x", "val": ": Tensor"}]</parameters></docstring>
raw latents -> [0, 1]

</div>
<div class="docstring border-l-2 border-t-2 pl-4 pt-3.5 border-gray-100 rounded-tl-xl mb-6 mt-8">


<docstring><name>unscale_latents</name><anchor>diffusers.AutoencoderTiny.unscale_latents</anchor><source>https://github.com/huggingface/diffusers/blob/vr_12229/src/diffusers/models/autoencoders/autoencoder_tiny.py#L161</source><parameters>[{"name": "x", "val": ": Tensor"}]</parameters></docstring>
[0, 1] -> raw latents

</div></div>

## AutoencoderTinyOutput[[diffusers.models.autoencoders.autoencoder_tiny.AutoencoderTinyOutput]]

<div class="docstring border-l-2 border-t-2 pl-4 pt-3.5 border-gray-100 rounded-tl-xl mb-6 mt-8">


<docstring><name>class diffusers.models.autoencoders.autoencoder_tiny.AutoencoderTinyOutput</name><anchor>diffusers.models.autoencoders.autoencoder_tiny.AutoencoderTinyOutput</anchor><source>https://github.com/huggingface/diffusers/blob/vr_12229/src/diffusers/models/autoencoders/autoencoder_tiny.py#L29</source><parameters>[{"name": "latents", "val": ": Tensor"}]</parameters><paramsdesc>- **latents** (`torch.Tensor`) -- Encoded outputs of the `Encoder`.</paramsdesc><paramgroups>0</paramgroups></docstring>

Output of AutoencoderTiny encoding method.




</div>

<EditOnGithub source="https://github.com/huggingface/diffusers/blob/main/docs/source/en/api/models/autoencoder_tiny.md" />