diffusers/pr_14252/en/_app/immutable/nodes/34.CB_gloPy.js

import"../chunks/DsnmJJEf.js";import{i as q,h as z,C as F,H as s,a as U,D as o,E as G,s as I}from"../chunks/BtE7mKSK.js";import{p as V,o as E,s as e,f as J,a as k,b as j,c as t,d as D,n,r}from"../chunks/jDjavuwI.js";const N='{"title":"AutoencoderKLAllegro","local":"autoencoderklallegro","sections":[{"title":"AutoencoderKLAllegro","local":"diffusers.AutoencoderKLAllegro","sections":[],"depth":2},{"title":"AutoencoderKLOutput","local":"diffusers.models.modeling_outputs.AutoencoderKLOutput","sections":[],"depth":2},{"title":"DecoderOutput","local":"diffusers.models.autoencoders.vae.DecoderOutput","sections":[],"depth":2}],"depth":1}';var Z=D('<meta name="hf:doc:metadata"/>'),R=D(`<p></p> <!> <!> <p>The 3D variational autoencoder (VAE) model with KL loss used in <a href="https://github.com/rhymes-ai/Allegro" rel="nofollow">Allegro</a> was introduced in <a href="https://huggingface.co/papers/2410.15458" rel="nofollow">Allegro: Open the Black Box of Commercial-Level Video Generation Model</a> by RhymesAI.</p> <p>The model can be loaded with the following code snippet.</p> <!> <!> <div class="docstring border-l-2 border-t-2 pl-4 pt-3.5 border-gray-100 rounded-tl-xl mb-6 mt-8"><!> <p>A VAE model with KL loss for encoding videos into latents and decoding latent representations into videos. Used in <a href="https://github.com/rhymes-ai/Allegro" rel="nofollow">Allegro</a>.</p> <p>This model inherits from <a href="/docs/diffusers/pr_14252/en/api/models/overview#diffusers.ModelMixin">ModelMixin</a>. Check the superclass documentation for it’s generic methods implemented
for all models (such as downloading or saving).</p> <div class="docstring border-l-2 border-t-2 pl-4 pt-3.5 border-gray-100 rounded-tl-xl mb-6 mt-8"><!> <p>Decode a batch of videos.</p></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"><!> <p>Encode a batch of videos into latents.</p></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"><!></div></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"><!> <p>Output of AutoencoderKL encoding method.</p></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"><!> <p>Output of decoding method.</p></div> <!> <p></p>`,1);function H(L,y){V(y,!1),E(()=>{new URLSearchParams(window.location.search).get("fw")}),q();var u=R();z("v9ic1a",A=>{var w=Z();I(w,"content",N),k(A,w)});var p=e(J(u),2);F(p,{containerStyle:"float: right; margin-left: 10px; display: inline-flex; position: relative; z-index: 10;"});var m=e(p,2);s(m,{title:"AutoencoderKLAllegro",local:"autoencoderklallegro",headingTag:"h1"});var f=e(m,6);U(f,{code:"ZnJvbSUyMGRpZmZ1c2VycyUyMGltcG9ydCUyMEF1dG9lbmNvZGVyS0xBbGxlZ3JvJTBBJTBBdmFlJTIwJTNEJTIwQXV0b2VuY29kZXJLTEFsbGVncm8uZnJvbV9wcmV0cmFpbmVkKCUyMnJoeW1lcy1haSUyRkFsbGVncm8lMjIlMkMlMjBzdWJmb2xkZXIlM0QlMjJ2YWUlMjIlMkMlMjB0b3JjaF9kdHlwZSUzRHRvcmNoLmZsb2F0MzIpLnRvKCUyMmN1ZGElMjIp",highlighted:`<span class="hljs-keyword">from</span> diffusers <span class="hljs-keyword">import</span> AutoencoderKLAllegro

vae = AutoencoderKLAllegro.from_pretrained(<span class="hljs-string">&quot;rhymes-ai/Allegro&quot;</span>, subfolder=<span class="hljs-string">&quot;vae&quot;</span>, torch_dtype=torch.float32).to(<span class="hljs-string">&quot;cuda&quot;</span>)`,lang:"python",wrap:!1});var g=e(f,2);s(g,{title:"AutoencoderKLAllegro",local:"diffusers.AutoencoderKLAllegro",headingTag:"h2"});var a=e(g,2),h=t(a);o(h,{name:"class diffusers.AutoencoderKLAllegro",anchor:"diffusers.AutoencoderKLAllegro",source:"https://github.com/huggingface/diffusers/blob/vr_14252/src/diffusers/models/autoencoders/autoencoder_kl_allegro.py#L676",parameters:[{name:"in_channels",val:": int = 3"},{name:"out_channels",val:": int = 3"},{name:"down_block_types",val:": tuple = ('AllegroDownBlock3D', 'AllegroDownBlock3D', 'AllegroDownBlock3D', 'AllegroDownBlock3D')"},{name:"up_block_types",val:": tuple = ('AllegroUpBlock3D', 'AllegroUpBlock3D', 'AllegroUpBlock3D', 'AllegroUpBlock3D')"},{name:"block_out_channels",val:": tuple = (128, 256, 512, 512)"},{name:"temporal_downsample_blocks",val:": tuple = (True, True, False, False)"},{name:"temporal_upsample_blocks",val:": tuple = (False, True, True, False)"},{name:"latent_channels",val:": int = 4"},{name:"layers_per_block",val:": int = 2"},{name:"act_fn",val:": str = 'silu'"},{name:"norm_num_groups",val:": int = 32"},{name:"temporal_compression_ratio",val:": float = 4"},{name:"sample_size",val:": int = 320"},{name:"scaling_factor",val:": float = 0.13"},{name:"force_upcast",val:": bool = True"}],parametersDescription:[{anchor:"diffusers.AutoencoderKLAllegro.in_channels",description:`<strong>in_channels</strong> (int, defaults to <code>3</code>) &#x2014;
Number of channels in the input image.`,name:"in_channels"},{anchor:"diffusers.AutoencoderKLAllegro.out_channels",description:`<strong>out_channels</strong> (int, defaults to <code>3</code>) &#x2014;
Number of channels in the output.`,name:"out_channels"},{anchor:"diffusers.AutoencoderKLAllegro.down_block_types",description:`<strong>down_block_types</strong> (<code>tuple[str, ...]</code>, defaults to <code>(&quot;AllegroDownBlock3D&quot;, &quot;AllegroDownBlock3D&quot;, &quot;AllegroDownBlock3D&quot;, &quot;AllegroDownBlock3D&quot;)</code>) &#x2014;
tuple of strings denoting which types of down blocks to use.`,name:"down_block_types"},{anchor:"diffusers.AutoencoderKLAllegro.up_block_types",description:`<strong>up_block_types</strong> (<code>tuple[str, ...]</code>, defaults to <code>(&quot;AllegroUpBlock3D&quot;, &quot;AllegroUpBlock3D&quot;, &quot;AllegroUpBlock3D&quot;, &quot;AllegroUpBlock3D&quot;)</code>) &#x2014;
tuple of strings denoting which types of up blocks to use.`,name:"up_block_types"},{anchor:"diffusers.AutoencoderKLAllegro.block_out_channels",description:`<strong>block_out_channels</strong> (<code>tuple[int, ...]</code>, defaults to <code>(128, 256, 512, 512)</code>) &#x2014;
tuple of integers denoting number of output channels in each block.`,name:"block_out_channels"},{anchor:"diffusers.AutoencoderKLAllegro.temporal_downsample_blocks",description:`<strong>temporal_downsample_blocks</strong> (<code>tuple[bool, ...]</code>, defaults to <code>(True, True, False, False)</code>) &#x2014;
tuple of booleans denoting which blocks to enable temporal downsampling in.`,name:"temporal_downsample_blocks"},{anchor:"diffusers.AutoencoderKLAllegro.latent_channels",description:`<strong>latent_channels</strong> (<code>int</code>, defaults to <code>4</code>) &#x2014;
Number of channels in latents.`,name:"latent_channels"},{anchor:"diffusers.AutoencoderKLAllegro.layers_per_block",description:`<strong>layers_per_block</strong> (<code>int</code>, defaults to <code>2</code>) &#x2014;
Number of resnet or attention or temporal convolution layers per down/up block.`,name:"layers_per_block"},{anchor:"diffusers.AutoencoderKLAllegro.act_fn",description:`<strong>act_fn</strong> (<code>str</code>, defaults to <code>&quot;silu&quot;</code>) &#x2014;
The activation function to use.`,name:"act_fn"},{anchor:"diffusers.AutoencoderKLAllegro.norm_num_groups",description:`<strong>norm_num_groups</strong> (<code>int</code>, defaults to <code>32</code>) &#x2014;
Number of groups to use in normalization layers.`,name:"norm_num_groups"},{anchor:"diffusers.AutoencoderKLAllegro.temporal_compression_ratio",description:`<strong>temporal_compression_ratio</strong> (<code>int</code>, defaults to <code>4</code>) &#x2014;
Ratio by which temporal dimension of samples are compressed.`,name:"temporal_compression_ratio"},{anchor:"diffusers.AutoencoderKLAllegro.sample_size",description:`<strong>sample_size</strong> (<code>int</code>, defaults to <code>320</code>) &#x2014;
Default latent size.`,name:"sample_size"},{anchor:"diffusers.AutoencoderKLAllegro.scaling_factor",description:`<strong>scaling_factor</strong> (<code>float</code>, defaults to <code>0.13235</code>) &#x2014;
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 <code>z = z * scaling_factor</code> before being passed to the
diffusion model. When decoding, the latents are scaled back to the original scale with the formula: <code>z = 1 / scaling_factor * z</code>. For more details, refer to sections 4.3.2 and D.1 of the <a href="https://huggingface.co/papers/2112.10752" rel="nofollow">High-Resolution Image
Synthesis with Latent Diffusion Models</a> paper.`,name:"scaling_factor"},{anchor:"diffusers.AutoencoderKLAllegro.force_upcast",description:`<strong>force_upcast</strong> (<code>bool</code>, default to <code>True</code>) &#x2014;
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 <code>force_upcast</code>
can be set to <code>False</code> - see: <a href="https://huggingface.co/madebyollin/sdxl-vae-fp16-fix" rel="nofollow">https://huggingface.co/madebyollin/sdxl-vae-fp16-fix</a>`,name:"force_upcast"}]});var l=e(h,6),x=t(l);o(x,{name:"decode",anchor:"diffusers.AutoencoderKLAllegro.decode",source:"https://github.com/huggingface/diffusers/blob/vr_14252/src/diffusers/models/autoencoders/autoencoder_kl_allegro.py#L843",parameters:[{name:"z",val:": Tensor"},{name:"return_dict",val:": bool = True"}],parametersDescription:[{anchor:"diffusers.AutoencoderKLAllegro.decode.z",description:`<strong>z</strong> (<code>torch.Tensor</code>) &#x2014;
Input batch of latent vectors.`,name:"z"},{anchor:"diffusers.AutoencoderKLAllegro.decode.return_dict",description:`<strong>return_dict</strong> (<code>bool</code>, defaults to <code>True</code>) &#x2014;
Whether to return a <code>~models.vae.DecoderOutput</code> instead of a plain tuple.`,name:"return_dict"}],returnDescription:`<script context="module">export const metadata = 'undefined';<\/script>


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


<p><code>~models.vae.DecoderOutput</code> or <code>tuple</code></p>
`}),n(2),r(l);var d=e(l,2),K=t(d);o(K,{name:"encode",anchor:"diffusers.AutoencoderKLAllegro.encode",source:"https://github.com/huggingface/diffusers/blob/vr_14252/src/diffusers/models/autoencoders/autoencoder_kl_allegro.py#L806",parameters:[{name:"x",val:": Tensor"},{name:"return_dict",val:": bool = True"}],parametersDescription:[{anchor:"diffusers.AutoencoderKLAllegro.encode.x",description:`<strong>x</strong> (<code>torch.Tensor</code>) &#x2014;
Input batch of videos.`,name:"x"},{anchor:"diffusers.AutoencoderKLAllegro.encode.return_dict",description:`<strong>return_dict</strong> (<code>bool</code>, defaults to <code>True</code>) &#x2014;
Whether to return a <code>~models.autoencoder_kl.AutoencoderKLOutput</code> instead of a plain tuple.`,name:"return_dict"}],returnDescription:`<script context="module">export const metadata = 'undefined';<\/script>


<p>The latent representations of the encoded videos. If <code>return_dict</code> is True, a
<code>~models.autoencoder_kl.AutoencoderKLOutput</code> is returned, otherwise a plain <code>tuple</code> is returned.</p>
`}),n(2),r(d);var _=e(d,2),T=t(_);o(T,{name:"forward",anchor:"diffusers.AutoencoderKLAllegro.forward",source:"https://github.com/huggingface/diffusers/blob/vr_14252/src/diffusers/models/autoencoders/autoencoder_kl_allegro.py#L1041",parameters:[{name:"sample",val:": Tensor"},{name:"sample_posterior",val:": bool = False"},{name:"return_dict",val:": bool = True"},{name:"generator",val:": typing.Optional[torch.Generator] = None"}],parametersDescription:[{anchor:"diffusers.AutoencoderKLAllegro.forward.sample",description:"<strong>sample</strong> (<code>torch.Tensor</code>) &#x2014; Input sample.",name:"sample"},{anchor:"diffusers.AutoencoderKLAllegro.forward.sample_posterior",description:`<strong>sample_posterior</strong> (<code>bool</code>, <em>optional</em>, defaults to <code>False</code>) &#x2014;
Whether to sample from the posterior.`,name:"sample_posterior"},{anchor:"diffusers.AutoencoderKLAllegro.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.AutoencoderKLAllegro.forward.generator",description:`<strong>generator</strong> (<code>torch.Generator</code>, <em>optional</em>) &#x2014;
PyTorch random number generator.`,name:"generator"}],returnDescription:`<script context="module">export const metadata = 'undefined';<\/script>


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


<p><code>~models.vae.DecoderOutput</code> or <code>tuple</code></p>
`}),r(_),r(a);var b=e(a,2);s(b,{title:"AutoencoderKLOutput",local:"diffusers.models.modeling_outputs.AutoencoderKLOutput",headingTag:"h2"});var c=e(b,2),O=t(c);o(O,{name:"class diffusers.models.modeling_outputs.AutoencoderKLOutput",anchor:"diffusers.models.modeling_outputs.AutoencoderKLOutput",source:"https://github.com/huggingface/diffusers/blob/vr_14252/src/diffusers/models/modeling_outputs.py#L7",parameters:[{name:"latent_dist",val:": DiagonalGaussianDistribution"}],parametersDescription:[{anchor:"diffusers.models.modeling_outputs.AutoencoderKLOutput.latent_dist",description:`<strong>latent_dist</strong> (<code>DiagonalGaussianDistribution</code>) &#x2014;
Encoded outputs of <code>Encoder</code> represented as the mean and logvar of <code>DiagonalGaussianDistribution</code>.
<code>DiagonalGaussianDistribution</code> allows for sampling latents from the distribution.`,name:"latent_dist"}]}),n(2),r(c);var v=e(c,2);s(v,{title:"DecoderOutput",local:"diffusers.models.autoencoders.vae.DecoderOutput",headingTag:"h2"});var i=e(v,2),B=t(i);o(B,{name:"class diffusers.models.autoencoders.vae.DecoderOutput",anchor:"diffusers.models.autoencoders.vae.DecoderOutput",source:"https://github.com/huggingface/diffusers/blob/vr_14252/src/diffusers/models/autoencoders/vae.py#L46",parameters:[{name:"sample",val:": Tensor"},{name:"commit_loss",val:": typing.Optional[torch.FloatTensor] = 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"}]}),n(2),r(i);var M=e(i,2);G(M,{source:"https://github.com/huggingface/diffusers/blob/main/docs/source/en/api/models/autoencoderkl_allegro.md"}),n(2),k(L,u),j()}export{H as component};