diffusers/pr_13832/en/_app/immutable/nodes/33.373bb116.js

import{s as Xe,n as Se,o as De}from"../chunks/scheduler.53228c21.js";import{S as Qe,i as qe,e as d,s as a,c as i,h as Fe,a as l,d as n,b as s,f as V,g as c,j as b,k as G,l as g,m as o,n as m,t as p,o as u,p as f}from"../chunks/index.cac5d66a.js";import{C as Ye}from"../chunks/CopyLLMTxtMenu.efae84b2.js";import{D as P}from"../chunks/Docstring.494cc782.js";import{C as Ee}from"../chunks/CodeBlock.606cbaf4.js";import{H,E as He}from"../chunks/MermaidChart.svelte_svelte_type_style_lang.de76e98b.js";function Oe(Ue){let _,ee,O,te,M,ne,v,oe,w,Je='The Representation Autoencoder (RAE) model introduced in <a href="https://huggingface.co/papers/2510.11690" rel="nofollow">Diffusion Transformers with Representation Autoencoders</a> by Boyang Zheng, Nanye Ma, Shengbang Tong, Saining Xie from NYU VISIONx.',ae,$,Ze="RAE combines a frozen pretrained vision encoder (DINOv2, SigLIP2, or MAE) with a trainable ViT-MAE-style decoder. In the two-stage RAE training recipe, the autoencoder is trained in stage 1 (reconstruction), and then a diffusion model is trained on the resulting latent space in stage 2 (generation).",se,T,Ie="The following RAE models are released and supported in Diffusers:",re,A,ze='<thead><tr><th align="left">Model</th> <th align="left">Encoder</th> <th align="left">Latent shape (224px input)</th></tr></thead> <tbody><tr><td align="left"><a href="https://huggingface.co/nyu-visionx/RAE-dinov2-wReg-base-ViTXL-n08" rel="nofollow"><code>nyu-visionx/RAE-dinov2-wReg-base-ViTXL-n08</code></a></td> <td align="left">DINOv2-base</td> <td align="left">768 x 16 x 16</td></tr> <tr><td align="left"><a href="https://huggingface.co/nyu-visionx/RAE-dinov2-wReg-base-ViTXL-n08-i512" rel="nofollow"><code>nyu-visionx/RAE-dinov2-wReg-base-ViTXL-n08-i512</code></a></td> <td align="left">DINOv2-base (512px)</td> <td align="left">768 x 32 x 32</td></tr> <tr><td align="left"><a href="https://huggingface.co/nyu-visionx/RAE-dinov2-wReg-small-ViTXL-n08" rel="nofollow"><code>nyu-visionx/RAE-dinov2-wReg-small-ViTXL-n08</code></a></td> <td align="left">DINOv2-small</td> <td align="left">384 x 16 x 16</td></tr> <tr><td align="left"><a href="https://huggingface.co/nyu-visionx/RAE-dinov2-wReg-large-ViTXL-n08" rel="nofollow"><code>nyu-visionx/RAE-dinov2-wReg-large-ViTXL-n08</code></a></td> <td align="left">DINOv2-large</td> <td align="left">1024 x 16 x 16</td></tr> <tr><td align="left"><a href="https://huggingface.co/nyu-visionx/RAE-siglip2-base-p16-i256-ViTXL-n08" rel="nofollow"><code>nyu-visionx/RAE-siglip2-base-p16-i256-ViTXL-n08</code></a></td> <td align="left">SigLIP2-base</td> <td align="left">768 x 16 x 16</td></tr> <tr><td align="left"><a href="https://huggingface.co/nyu-visionx/RAE-mae-base-p16-ViTXL-n08" rel="nofollow"><code>nyu-visionx/RAE-mae-base-p16-ViTXL-n08</code></a></td> <td align="left">MAE-base</td> <td align="left">768 x 16 x 16</td></tr></tbody>',de,R,le,x,ie,j,ce,E,me,U,pe,J,Ce="Some pretrained checkpoints include per-channel <code>latents_mean</code> and <code>latents_std</code> statistics for normalizing the latent space. When present, <code>encode</code> and <code>decode</code> automatically apply the normalization and denormalization, respectively.",ue,Z,fe,I,ge,r,z,we,X,We="Representation Autoencoder (RAE) model for encoding images to latents and decoding latents to images.",$e,S,Ne=`This model uses a frozen pretrained encoder (DINOv2, SigLIP2, or MAE) with a trainable ViT decoder to reconstruct
images from learned representations.`,Te,D,Le=`This model inherits from <a href="/docs/diffusers/pr_13832/en/api/models/overview#diffusers.ModelMixin">ModelMixin</a>. Check the superclass documentation for its generic methods implemented for
all models (such as downloading or saving).`,Ae,Q,C,Re,q,W,xe,F,N,he,L,_e,y,B,je,Y,Be="Output of decoding method.",ye,k,be,K,Me;return M=new Ye({props:{containerStyle:"float: right; margin-left: 10px; display: inline-flex; position: relative; z-index: 10;"}}),v=new H({props:{title:"AutoencoderRAE",local:"autoencoderrae",headingTag:"h1"}}),R=new H({props:{title:"Loading a pretrained model",local:"loading-a-pretrained-model",headingTag:"h2"}}),x=new Ee({props:{code:"ZnJvbSUyMGRpZmZ1c2VycyUyMGltcG9ydCUyMEF1dG9lbmNvZGVyUkFFJTBBJTBBbW9kZWwlMjAlM0QlMjBBdXRvZW5jb2RlclJBRS5mcm9tX3ByZXRyYWluZWQoJTBBJTIwJTIwJTIwJTIwJTIybnl1LXZpc2lvbnglMkZSQUUtZGlub3YyLXdSZWctYmFzZS1WaVRYTC1uMDglMjIlMEEpLnRvKCUyMmN1ZGElMjIpLmV2YWwoKQ==",highlighted:`<span class="hljs-keyword">from</span> diffusers <span class="hljs-keyword">import</span> AutoencoderRAE

model = AutoencoderRAE.from_pretrained(
    <span class="hljs-string">&quot;nyu-visionx/RAE-dinov2-wReg-base-ViTXL-n08&quot;</span>
).to(<span class="hljs-string">&quot;cuda&quot;</span>).<span class="hljs-built_in">eval</span>()`,lang:"python",wrap:!1}}),j=new H({props:{title:"Encoding and decoding a real image",local:"encoding-and-decoding-a-real-image",headingTag:"h2"}}),E=new Ee({props:{code:"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",highlighted:`<span class="hljs-keyword">import</span> torch
<span class="hljs-keyword">from</span> diffusers <span class="hljs-keyword">import</span> AutoencoderRAE
<span class="hljs-keyword">from</span> diffusers.utils <span class="hljs-keyword">import</span> load_image
<span class="hljs-keyword">from</span> torchvision.transforms.functional <span class="hljs-keyword">import</span> to_tensor, to_pil_image

model = AutoencoderRAE.from_pretrained(
    <span class="hljs-string">&quot;nyu-visionx/RAE-dinov2-wReg-base-ViTXL-n08&quot;</span>
).to(<span class="hljs-string">&quot;cuda&quot;</span>).<span class="hljs-built_in">eval</span>()

image = load_image(<span class="hljs-string">&quot;https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/diffusers/cat.png&quot;</span>)
image = image.convert(<span class="hljs-string">&quot;RGB&quot;</span>).resize((<span class="hljs-number">224</span>, <span class="hljs-number">224</span>))
x = to_tensor(image).unsqueeze(<span class="hljs-number">0</span>).to(<span class="hljs-string">&quot;cuda&quot;</span>)  <span class="hljs-comment"># (1, 3, 224, 224), values in [0, 1]</span>

<span class="hljs-keyword">with</span> torch.no_grad():
    latents = model.encode(x).latent        <span class="hljs-comment"># (1, 768, 16, 16)</span>
    recon = model.decode(latents).sample     <span class="hljs-comment"># (1, 3, 256, 256)</span>

recon_image = to_pil_image(recon[<span class="hljs-number">0</span>].clamp(<span class="hljs-number">0</span>, <span class="hljs-number">1</span>).cpu())
recon_image.save(<span class="hljs-string">&quot;recon.png&quot;</span>)`,lang:"python",wrap:!1}}),U=new H({props:{title:"Latent normalization",local:"latent-normalization",headingTag:"h2"}}),Z=new Ee({props:{code:"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",highlighted:`model = AutoencoderRAE.from_pretrained(
    <span class="hljs-string">&quot;nyu-visionx/RAE-dinov2-wReg-base-ViTXL-n08&quot;</span>
).to(<span class="hljs-string">&quot;cuda&quot;</span>).<span class="hljs-built_in">eval</span>()

<span class="hljs-comment"># Latent normalization is handled automatically inside encode/decode</span>
<span class="hljs-comment"># when the checkpoint config includes latents_mean/latents_std.</span>
<span class="hljs-keyword">with</span> torch.no_grad():
    latents = model.encode(x).latent   <span class="hljs-comment"># normalized latents</span>
    recon = model.decode(latents).sample`,lang:"python",wrap:!1}}),I=new H({props:{title:"AutoencoderRAE",local:"diffusers.AutoencoderRAE",headingTag:"h2"}}),z=new P({props:{name:"class diffusers.AutoencoderRAE",anchor:"diffusers.AutoencoderRAE",parameters:[{name:"encoder_type",val:": str = 'dinov2'"},{name:"encoder_hidden_size",val:": int = 768"},{name:"encoder_patch_size",val:": int = 14"},{name:"encoder_num_hidden_layers",val:": int = 12"},{name:"decoder_hidden_size",val:": int = 512"},{name:"decoder_num_hidden_layers",val:": int = 8"},{name:"decoder_num_attention_heads",val:": int = 16"},{name:"decoder_intermediate_size",val:": int = 2048"},{name:"patch_size",val:": int = 16"},{name:"encoder_input_size",val:": int = 224"},{name:"image_size",val:": int | None = None"},{name:"num_channels",val:": int = 3"},{name:"encoder_norm_mean",val:": list | None = None"},{name:"encoder_norm_std",val:": list | None = None"},{name:"latents_mean",val:": list | tuple | torch.Tensor | None = None"},{name:"latents_std",val:": list | tuple | torch.Tensor | None = None"},{name:"noise_tau",val:": float = 0.0"},{name:"reshape_to_2d",val:": bool = True"},{name:"use_encoder_loss",val:": bool = False"},{name:"scaling_factor",val:": float = 1.0"}],parametersDescription:[{anchor:"diffusers.AutoencoderRAE.encoder_type",description:`<strong>encoder_type</strong> (<code>str</code>, <em>optional</em>, defaults to <code>&quot;dinov2&quot;</code>) &#x2014;
Type of frozen encoder to use. One of <code>&quot;dinov2&quot;</code>, <code>&quot;siglip2&quot;</code>, or <code>&quot;mae&quot;</code>.`,name:"encoder_type"},{anchor:"diffusers.AutoencoderRAE.encoder_hidden_size",description:`<strong>encoder_hidden_size</strong> (<code>int</code>, <em>optional</em>, defaults to <code>768</code>) &#x2014;
Hidden size of the encoder model.`,name:"encoder_hidden_size"},{anchor:"diffusers.AutoencoderRAE.encoder_patch_size",description:`<strong>encoder_patch_size</strong> (<code>int</code>, <em>optional</em>, defaults to <code>14</code>) &#x2014;
Patch size of the encoder model.`,name:"encoder_patch_size"},{anchor:"diffusers.AutoencoderRAE.encoder_num_hidden_layers",description:`<strong>encoder_num_hidden_layers</strong> (<code>int</code>, <em>optional</em>, defaults to <code>12</code>) &#x2014;
Number of hidden layers in the encoder model.`,name:"encoder_num_hidden_layers"},{anchor:"diffusers.AutoencoderRAE.patch_size",description:`<strong>patch_size</strong> (<code>int</code>, <em>optional</em>, defaults to <code>16</code>) &#x2014;
Decoder patch size (used for unpatchify and decoder head).`,name:"patch_size"},{anchor:"diffusers.AutoencoderRAE.encoder_input_size",description:`<strong>encoder_input_size</strong> (<code>int</code>, <em>optional</em>, defaults to <code>224</code>) &#x2014;
Input size expected by the encoder.`,name:"encoder_input_size"},{anchor:"diffusers.AutoencoderRAE.image_size",description:`<strong>image_size</strong> (<code>int</code>, <em>optional</em>) &#x2014;
Decoder output image size. If <code>None</code>, it is derived from encoder token count and <code>patch_size</code> like
RAE-main: <code>image_size = patch_size * sqrt(num_patches)</code>, where <code>num_patches = (encoder_input_size // encoder_patch_size) ** 2</code>.`,name:"image_size"},{anchor:"diffusers.AutoencoderRAE.num_channels",description:`<strong>num_channels</strong> (<code>int</code>, <em>optional</em>, defaults to <code>3</code>) &#x2014;
Number of input/output channels.`,name:"num_channels"},{anchor:"diffusers.AutoencoderRAE.encoder_norm_mean",description:`<strong>encoder_norm_mean</strong> (<code>list</code>, <em>optional</em>, defaults to <code>[0.485, 0.456, 0.406]</code>) &#x2014;
Channel-wise mean for encoder input normalization (ImageNet defaults).`,name:"encoder_norm_mean"},{anchor:"diffusers.AutoencoderRAE.encoder_norm_std",description:`<strong>encoder_norm_std</strong> (<code>list</code>, <em>optional</em>, defaults to <code>[0.229, 0.224, 0.225]</code>) &#x2014;
Channel-wise std for encoder input normalization (ImageNet defaults).`,name:"encoder_norm_std"},{anchor:"diffusers.AutoencoderRAE.latents_mean",description:`<strong>latents_mean</strong> (<code>list</code> or <code>tuple</code>, <em>optional</em>) &#x2014;
Optional mean for latent normalization. Tensor inputs are accepted and converted to config-serializable
lists.`,name:"latents_mean"},{anchor:"diffusers.AutoencoderRAE.latents_std",description:`<strong>latents_std</strong> (<code>list</code> or <code>tuple</code>, <em>optional</em>) &#x2014;
Optional standard deviation for latent normalization. Tensor inputs are accepted and converted to
config-serializable lists.`,name:"latents_std"},{anchor:"diffusers.AutoencoderRAE.noise_tau",description:`<strong>noise_tau</strong> (<code>float</code>, <em>optional</em>, defaults to <code>0.0</code>) &#x2014;
Noise level for training (adds noise to latents during training).`,name:"noise_tau"},{anchor:"diffusers.AutoencoderRAE.reshape_to_2d",description:`<strong>reshape_to_2d</strong> (<code>bool</code>, <em>optional</em>, defaults to <code>True</code>) &#x2014;
Whether to reshape latents to 2D (B, C, H, W) format.`,name:"reshape_to_2d"},{anchor:"diffusers.AutoencoderRAE.use_encoder_loss",description:`<strong>use_encoder_loss</strong> (<code>bool</code>, <em>optional</em>, defaults to <code>False</code>) &#x2014;
Whether to use encoder hidden states in the loss (for advanced training).`,name:"use_encoder_loss"}],source:"https://github.com/huggingface/diffusers/blob/vr_13832/src/diffusers/models/autoencoders/autoencoder_rae.py#L393"}}),C=new P({props:{name:"wrapper",anchor:"diffusers.AutoencoderRAE.encode",parameters:[{name:"*args",val:""},{name:"**kwargs",val:""}],source:"https://github.com/huggingface/diffusers/blob/vr_13832/src/diffusers/utils/accelerate_utils.py#L43"}}),W=new P({props:{name:"wrapper",anchor:"diffusers.AutoencoderRAE.decode",parameters:[{name:"*args",val:""},{name:"**kwargs",val:""}],source:"https://github.com/huggingface/diffusers/blob/vr_13832/src/diffusers/utils/accelerate_utils.py#L43"}}),N=new P({props:{name:"forward",anchor:"diffusers.AutoencoderRAE.forward",parameters:[{name:"sample",val:": Tensor"},{name:"return_dict",val:": bool = True"},{name:"generator",val:": torch._C.Generator | None = None"}],parametersDescription:[{anchor:"diffusers.AutoencoderRAE.forward.sample",description:"<strong>sample</strong> (<code>torch.Tensor</code>) &#x2014; Input sample.",name:"sample"},{anchor:"diffusers.AutoencoderRAE.forward.return_dict",description:`<strong>return_dict</strong> (<code>bool</code>, <em>optional</em>, defaults to <code>True</code>) &#x2014;
Whether or not to return a <code>DecoderOutput</code> instead of a plain tuple.`,name:"return_dict"},{anchor:"diffusers.AutoencoderRAE.forward.generator",description:`<strong>generator</strong> (<code>torch.Generator</code>, <em>optional</em>) &#x2014;
A <a href="https://pytorch.org/docs/stable/generated/torch.Generator.html" rel="nofollow"><code>torch.Generator</code></a> to make sampling
deterministic.`,name:"generator"}],source:"https://github.com/huggingface/diffusers/blob/vr_13832/src/diffusers/models/autoencoders/autoencoder_rae.py#L682",returnDescription:`<script context="module">export const metadata = 'undefined';<\/script>


<p>If <code>return_dict</code> is True, a <code>DecoderOutput</code> is returned, otherwise a plain <code>tuple</code> is returned.</p>
`,returnType:`<script context="module">export const metadata = 'undefined';<\/script>


<p><code>DecoderOutput</code> or <code>tuple</code></p>
`}}),L=new H({props:{title:"DecoderOutput",local:"diffusers.models.autoencoders.vae.DecoderOutput",headingTag:"h2"}}),B=new P({props:{name:"class diffusers.models.autoencoders.vae.DecoderOutput",anchor:"diffusers.models.autoencoders.vae.DecoderOutput",parameters:[{name:"sample",val:": Tensor"},{name:"commit_loss",val:": torch.FloatTensor | None = None"}],parametersDescription:[{anchor:"diffusers.models.autoencoders.vae.DecoderOutput.sample",description:`<strong>sample</strong> (<code>torch.Tensor</code> of shape <code>(batch_size, num_channels, height, width)</code>) &#x2014;
The decoded output sample from the last layer of the model.`,name:"sample"}],source:"https://github.com/huggingface/diffusers/blob/vr_13832/src/diffusers/models/autoencoders/vae.py#L46"}}),k=new He({props:{source:"https://github.com/huggingface/diffusers/blob/main/docs/source/en/api/models/autoencoder_rae.md"}}),{c(){_=d("meta"),ee=a(),O=d("p"),te=a(),i(M.$$.fragment),ne=a(),i(v.$$.fragment),oe=a(),w=d("p"),w.innerHTML=Je,ae=a(),$=d("p"),$.textContent=Ze,se=a(),T=d("p"),T.textContent=Ie,re=a(),A=d("table"),A.innerHTML=ze,de=a(),i(R.$$.fragment),le=a(),i(x.$$.fragment),ie=a(),i(j.$$.fragment),ce=a(),i(E.$$.fragment),me=a(),i(U.$$.fragment),pe=a(),J=d("p"),J.innerHTML=Ce,ue=a(),i(Z.$$.fragment),fe=a(),i(I.$$.fragment),ge=a(),r=d("div"),i(z.$$.fragment),we=a(),X=d("p"),X.textContent=We,$e=a(),S=d("p"),S.textContent=Ne,Te=a(),D=d("p"),D.innerHTML=Le,Ae=a(),Q=d("div"),i(C.$$.fragment),Re=a(),q=d("div"),i(W.$$.fragment),xe=a(),F=d("div"),i(N.$$.fragment),he=a(),i(L.$$.fragment),_e=a(),y=d("div"),i(B.$$.fragment),je=a(),Y=d("p"),Y.textContent=Be,ye=a(),i(k.$$.fragment),be=a(),K=d("p"),this.h()},l(e){const t=Fe("svelte-u9bgzb",document.head);_=l(t,"META",{name:!0,content:!0}),t.forEach(n),ee=s(e),O=l(e,"P",{}),V(O).forEach(n),te=s(e),c(M.$$.fragment,e),ne=s(e),c(v.$$.fragment,e),oe=s(e),w=l(e,"P",{"data-svelte-h":!0}),b(w)!=="svelte-1i7ai8n"&&(w.innerHTML=Je),ae=s(e),$=l(e,"P",{"data-svelte-h":!0}),b($)!=="svelte-1k2g9sg"&&($.textContent=Ze),se=s(e),T=l(e,"P",{"data-svelte-h":!0}),b(T)!=="svelte-nl10ag"&&(T.textContent=Ie),re=s(e),A=l(e,"TABLE",{"data-svelte-h":!0}),b(A)!=="svelte-9jdi6e"&&(A.innerHTML=ze),de=s(e),c(R.$$.fragment,e),le=s(e),c(x.$$.fragment,e),ie=s(e),c(j.$$.fragment,e),ce=s(e),c(E.$$.fragment,e),me=s(e),c(U.$$.fragment,e),pe=s(e),J=l(e,"P",{"data-svelte-h":!0}),b(J)!=="svelte-1jwlhl7"&&(J.innerHTML=Ce),ue=s(e),c(Z.$$.fragment,e),fe=s(e),c(I.$$.fragment,e),ge=s(e),r=l(e,"DIV",{class:!0});var h=V(r);c(z.$$.fragment,h),we=s(h),X=l(h,"P",{"data-svelte-h":!0}),b(X)!=="svelte-1a91lnw"&&(X.textContent=We),$e=s(h),S=l(h,"P",{"data-svelte-h":!0}),b(S)!=="svelte-z7pq73"&&(S.textContent=Ne),Te=s(h),D=l(h,"P",{"data-svelte-h":!0}),b(D)!=="svelte-gv2zfw"&&(D.innerHTML=Le),Ae=s(h),Q=l(h,"DIV",{class:!0});var ke=V(Q);c(C.$$.fragment,ke),ke.forEach(n),Re=s(h),q=l(h,"DIV",{class:!0});var Ve=V(q);c(W.$$.fragment,Ve),Ve.forEach(n),xe=s(h),F=l(h,"DIV",{class:!0});var Ge=V(F);c(N.$$.fragment,Ge),Ge.forEach(n),h.forEach(n),he=s(e),c(L.$$.fragment,e),_e=s(e),y=l(e,"DIV",{class:!0});var ve=V(y);c(B.$$.fragment,ve),je=s(ve),Y=l(ve,"P",{"data-svelte-h":!0}),b(Y)!=="svelte-18u8upa"&&(Y.textContent=Be),ve.forEach(n),ye=s(e),c(k.$$.fragment,e),be=s(e),K=l(e,"P",{}),V(K).forEach(n),this.h()},h(){G(_,"name","hf:doc:metadata"),G(_,"content",Ke),G(Q,"class","docstring border-l-2 border-t-2 pl-4 pt-3.5 border-gray-100 rounded-tl-xl mb-6 mt-8"),G(q,"class","docstring border-l-2 border-t-2 pl-4 pt-3.5 border-gray-100 rounded-tl-xl mb-6 mt-8"),G(F,"class","docstring border-l-2 border-t-2 pl-4 pt-3.5 border-gray-100 rounded-tl-xl mb-6 mt-8"),G(r,"class","docstring border-l-2 border-t-2 pl-4 pt-3.5 border-gray-100 rounded-tl-xl mb-6 mt-8"),G(y,"class","docstring border-l-2 border-t-2 pl-4 pt-3.5 border-gray-100 rounded-tl-xl mb-6 mt-8")},m(e,t){g(document.head,_),o(e,ee,t),o(e,O,t),o(e,te,t),m(M,e,t),o(e,ne,t),m(v,e,t),o(e,oe,t),o(e,w,t),o(e,ae,t),o(e,$,t),o(e,se,t),o(e,T,t),o(e,re,t),o(e,A,t),o(e,de,t),m(R,e,t),o(e,le,t),m(x,e,t),o(e,ie,t),m(j,e,t),o(e,ce,t),m(E,e,t),o(e,me,t),m(U,e,t),o(e,pe,t),o(e,J,t),o(e,ue,t),m(Z,e,t),o(e,fe,t),m(I,e,t),o(e,ge,t),o(e,r,t),m(z,r,null),g(r,we),g(r,X),g(r,$e),g(r,S),g(r,Te),g(r,D),g(r,Ae),g(r,Q),m(C,Q,null),g(r,Re),g(r,q),m(W,q,null),g(r,xe),g(r,F),m(N,F,null),o(e,he,t),m(L,e,t),o(e,_e,t),o(e,y,t),m(B,y,null),g(y,je),g(y,Y),o(e,ye,t),m(k,e,t),o(e,be,t),o(e,K,t),Me=!0},p:Se,i(e){Me||(p(M.$$.fragment,e),p(v.$$.fragment,e),p(R.$$.fragment,e),p(x.$$.fragment,e),p(j.$$.fragment,e),p(E.$$.fragment,e),p(U.$$.fragment,e),p(Z.$$.fragment,e),p(I.$$.fragment,e),p(z.$$.fragment,e),p(C.$$.fragment,e),p(W.$$.fragment,e),p(N.$$.fragment,e),p(L.$$.fragment,e),p(B.$$.fragment,e),p(k.$$.fragment,e),Me=!0)},o(e){u(M.$$.fragment,e),u(v.$$.fragment,e),u(R.$$.fragment,e),u(x.$$.fragment,e),u(j.$$.fragment,e),u(E.$$.fragment,e),u(U.$$.fragment,e),u(Z.$$.fragment,e),u(I.$$.fragment,e),u(z.$$.fragment,e),u(C.$$.fragment,e),u(W.$$.fragment,e),u(N.$$.fragment,e),u(L.$$.fragment,e),u(B.$$.fragment,e),u(k.$$.fragment,e),Me=!1},d(e){e&&(n(ee),n(O),n(te),n(ne),n(oe),n(w),n(ae),n($),n(se),n(T),n(re),n(A),n(de),n(le),n(ie),n(ce),n(me),n(pe),n(J),n(ue),n(fe),n(ge),n(r),n(he),n(_e),n(y),n(ye),n(be),n(K)),n(_),f(M,e),f(v,e),f(R,e),f(x,e),f(j,e),f(E,e),f(U,e),f(Z,e),f(I,e),f(z),f(C),f(W),f(N),f(L,e),f(B),f(k,e)}}}const Ke='{"title":"AutoencoderRAE","local":"autoencoderrae","sections":[{"title":"Loading a pretrained model","local":"loading-a-pretrained-model","sections":[],"depth":2},{"title":"Encoding and decoding a real image","local":"encoding-and-decoding-a-real-image","sections":[],"depth":2},{"title":"Latent normalization","local":"latent-normalization","sections":[],"depth":2},{"title":"AutoencoderRAE","local":"diffusers.AutoencoderRAE","sections":[],"depth":2},{"title":"DecoderOutput","local":"diffusers.models.autoencoders.vae.DecoderOutput","sections":[],"depth":2}],"depth":1}';function Pe(Ue){return De(()=>{new URLSearchParams(window.location.search).get("fw")}),[]}class rt extends Qe{constructor(_){super(),qe(this,_,Pe,Oe,Xe,{})}}export{rt as component};