diffusers/pr_13921/en/_app/immutable/nodes/63.56cd8856.js

import{s as fe,n as pe,o as ue}from"../chunks/scheduler.53228c21.js";import{S as _e,i as he,e as i,s,c,h as ge,a as d,d as n,b as r,f as C,g as f,j as F,k as H,l as p,m as o,n as u,t as _,o as h,p as g}from"../chunks/index.cac5d66a.js";import{C as Te}from"../chunks/CopyLLMTxtMenu.0ef49226.js";import{D as se}from"../chunks/Docstring.9de32ff4.js";import{C as ye}from"../chunks/CodeBlock.606cbaf4.js";import{H as re,E as be}from"../chunks/MermaidChart.svelte_svelte_type_style_lang.48d5cb47.js";function Me(ae){let m,j,J,V,y,W,b,q,M,ie='A Diffusion Transformer model for 3D data from <a href="https://github.com/aigc-apps/EasyAnimate" rel="nofollow">EasyAnimate</a> was introduced by Alibaba PAI.',P,v,de="The model can be loaded with the following code snippet.",O,$,R,x,Y,a,D,ee,k,me='A Transformer model for video-like data in <a href="https://github.com/aigc-apps/EasyAnimate" rel="nofollow">EasyAnimate</a>.',te,T,E,ne,z,le='The <a href="/docs/diffusers/pr_13921/en/api/models/easyanimate_transformer3d#diffusers.EasyAnimateTransformer3DModel">EasyAnimateTransformer3DModel</a> forward method.',B,A,X,l,w,oe,L,ce='The output of <a href="/docs/diffusers/pr_13921/en/api/models/transformer2d#diffusers.Transformer2DModel">Transformer2DModel</a>.',G,N,S,Z,U;return y=new Te({props:{containerStyle:"float: right; margin-left: 10px; display: inline-flex; position: relative; z-index: 10;"}}),b=new re({props:{title:"EasyAnimateTransformer3DModel",local:"easyanimatetransformer3dmodel",headingTag:"h1"}}),$=new ye({props:{code:"ZnJvbSUyMGRpZmZ1c2VycyUyMGltcG9ydCUyMEVhc3lBbmltYXRlVHJhbnNmb3JtZXIzRE1vZGVsJTBBJTBBdHJhbnNmb3JtZXIlMjAlM0QlMjBFYXN5QW5pbWF0ZVRyYW5zZm9ybWVyM0RNb2RlbC5mcm9tX3ByZXRyYWluZWQoJTIyYWxpYmFiYS1wYWklMkZFYXN5QW5pbWF0ZVY1LjEtMTJiLXpoJTIyJTJDJTIwc3ViZm9sZGVyJTNEJTIydHJhbnNmb3JtZXIlMjIlMkMlMjB0b3JjaF9kdHlwZSUzRHRvcmNoLmZsb2F0MTYpLnRvKCUyMmN1ZGElMjIp",highlighted:`<span class="hljs-keyword">from</span> diffusers <span class="hljs-keyword">import</span> EasyAnimateTransformer3DModel

transformer = EasyAnimateTransformer3DModel.from_pretrained(<span class="hljs-string">&quot;alibaba-pai/EasyAnimateV5.1-12b-zh&quot;</span>, subfolder=<span class="hljs-string">&quot;transformer&quot;</span>, torch_dtype=torch.float16).to(<span class="hljs-string">&quot;cuda&quot;</span>)`,lang:"python",wrap:!1}}),x=new re({props:{title:"EasyAnimateTransformer3DModel",local:"diffusers.EasyAnimateTransformer3DModel",headingTag:"h2"}}),D=new se({props:{name:"class diffusers.EasyAnimateTransformer3DModel",anchor:"diffusers.EasyAnimateTransformer3DModel",parameters:[{name:"num_attention_heads",val:": int = 48"},{name:"attention_head_dim",val:": int = 64"},{name:"in_channels",val:": int | None = None"},{name:"out_channels",val:": int | None = None"},{name:"patch_size",val:": int | None = None"},{name:"sample_width",val:": int = 90"},{name:"sample_height",val:": int = 60"},{name:"activation_fn",val:": str = 'gelu-approximate'"},{name:"timestep_activation_fn",val:": str = 'silu'"},{name:"freq_shift",val:": int = 0"},{name:"num_layers",val:": int = 48"},{name:"mmdit_layers",val:": int = 48"},{name:"dropout",val:": float = 0.0"},{name:"time_embed_dim",val:": int = 512"},{name:"add_norm_text_encoder",val:": bool = False"},{name:"text_embed_dim",val:": int = 3584"},{name:"text_embed_dim_t5",val:": int = None"},{name:"norm_eps",val:": float = 1e-05"},{name:"norm_elementwise_affine",val:": bool = True"},{name:"flip_sin_to_cos",val:": bool = True"},{name:"time_position_encoding_type",val:": str = '3d_rope'"},{name:"after_norm",val:" = False"},{name:"resize_inpaint_mask_directly",val:": bool = True"},{name:"enable_text_attention_mask",val:": bool = True"},{name:"add_noise_in_inpaint_model",val:": bool = True"}],parametersDescription:[{anchor:"diffusers.EasyAnimateTransformer3DModel.num_attention_heads",description:`<strong>num_attention_heads</strong> (<code>int</code>, defaults to <code>48</code>) &#x2014;
The number of heads to use for multi-head attention.`,name:"num_attention_heads"},{anchor:"diffusers.EasyAnimateTransformer3DModel.attention_head_dim",description:`<strong>attention_head_dim</strong> (<code>int</code>, defaults to <code>64</code>) &#x2014;
The number of channels in each head.`,name:"attention_head_dim"},{anchor:"diffusers.EasyAnimateTransformer3DModel.in_channels",description:`<strong>in_channels</strong> (<code>int</code>, defaults to <code>16</code>) &#x2014;
The number of channels in the input.`,name:"in_channels"},{anchor:"diffusers.EasyAnimateTransformer3DModel.out_channels",description:`<strong>out_channels</strong> (<code>int</code>, <em>optional</em>, defaults to <code>16</code>) &#x2014;
The number of channels in the output.`,name:"out_channels"},{anchor:"diffusers.EasyAnimateTransformer3DModel.patch_size",description:`<strong>patch_size</strong> (<code>int</code>, defaults to <code>2</code>) &#x2014;
The size of the patches to use in the patch embedding layer.`,name:"patch_size"},{anchor:"diffusers.EasyAnimateTransformer3DModel.sample_width",description:`<strong>sample_width</strong> (<code>int</code>, defaults to <code>90</code>) &#x2014;
The width of the input latents.`,name:"sample_width"},{anchor:"diffusers.EasyAnimateTransformer3DModel.sample_height",description:`<strong>sample_height</strong> (<code>int</code>, defaults to <code>60</code>) &#x2014;
The height of the input latents.`,name:"sample_height"},{anchor:"diffusers.EasyAnimateTransformer3DModel.activation_fn",description:`<strong>activation_fn</strong> (<code>str</code>, defaults to <code>&quot;gelu-approximate&quot;</code>) &#x2014;
Activation function to use in feed-forward.`,name:"activation_fn"},{anchor:"diffusers.EasyAnimateTransformer3DModel.timestep_activation_fn",description:`<strong>timestep_activation_fn</strong> (<code>str</code>, defaults to <code>&quot;silu&quot;</code>) &#x2014;
Activation function to use when generating the timestep embeddings.`,name:"timestep_activation_fn"},{anchor:"diffusers.EasyAnimateTransformer3DModel.num_layers",description:`<strong>num_layers</strong> (<code>int</code>, defaults to <code>30</code>) &#x2014;
The number of layers of Transformer blocks to use.`,name:"num_layers"},{anchor:"diffusers.EasyAnimateTransformer3DModel.mmdit_layers",description:`<strong>mmdit_layers</strong> (<code>int</code>, defaults to <code>1000</code>) &#x2014;
The number of layers of Multi Modal Transformer blocks to use.`,name:"mmdit_layers"},{anchor:"diffusers.EasyAnimateTransformer3DModel.dropout",description:`<strong>dropout</strong> (<code>float</code>, defaults to <code>0.0</code>) &#x2014;
The dropout probability to use.`,name:"dropout"},{anchor:"diffusers.EasyAnimateTransformer3DModel.time_embed_dim",description:`<strong>time_embed_dim</strong> (<code>int</code>, defaults to <code>512</code>) &#x2014;
Output dimension of timestep embeddings.`,name:"time_embed_dim"},{anchor:"diffusers.EasyAnimateTransformer3DModel.text_embed_dim",description:`<strong>text_embed_dim</strong> (<code>int</code>, defaults to <code>4096</code>) &#x2014;
Input dimension of text embeddings from the text encoder.`,name:"text_embed_dim"},{anchor:"diffusers.EasyAnimateTransformer3DModel.norm_eps",description:`<strong>norm_eps</strong> (<code>float</code>, defaults to <code>1e-5</code>) &#x2014;
The epsilon value to use in normalization layers.`,name:"norm_eps"},{anchor:"diffusers.EasyAnimateTransformer3DModel.norm_elementwise_affine",description:`<strong>norm_elementwise_affine</strong> (<code>bool</code>, defaults to <code>True</code>) &#x2014;
Whether to use elementwise affine in normalization layers.`,name:"norm_elementwise_affine"},{anchor:"diffusers.EasyAnimateTransformer3DModel.flip_sin_to_cos",description:`<strong>flip_sin_to_cos</strong> (<code>bool</code>, defaults to <code>True</code>) &#x2014;
Whether to flip the sin to cos in the time embedding.`,name:"flip_sin_to_cos"},{anchor:"diffusers.EasyAnimateTransformer3DModel.time_position_encoding_type",description:`<strong>time_position_encoding_type</strong> (<code>str</code>, defaults to <code>3d_rope</code>) &#x2014;
Type of time position encoding.`,name:"time_position_encoding_type"},{anchor:"diffusers.EasyAnimateTransformer3DModel.after_norm",description:`<strong>after_norm</strong> (<code>bool</code>, defaults to <code>False</code>) &#x2014;
Flag to apply normalization after.`,name:"after_norm"},{anchor:"diffusers.EasyAnimateTransformer3DModel.resize_inpaint_mask_directly",description:`<strong>resize_inpaint_mask_directly</strong> (<code>bool</code>, defaults to <code>True</code>) &#x2014;
Flag to resize inpaint mask directly.`,name:"resize_inpaint_mask_directly"},{anchor:"diffusers.EasyAnimateTransformer3DModel.enable_text_attention_mask",description:`<strong>enable_text_attention_mask</strong> (<code>bool</code>, defaults to <code>True</code>) &#x2014;
Flag to enable text attention mask.`,name:"enable_text_attention_mask"},{anchor:"diffusers.EasyAnimateTransformer3DModel.add_noise_in_inpaint_model",description:`<strong>add_noise_in_inpaint_model</strong> (<code>bool</code>, defaults to <code>False</code>) &#x2014;
Flag to add noise in inpaint model.`,name:"add_noise_in_inpaint_model"}],source:"https://github.com/huggingface/diffusers/blob/vr_13921/src/diffusers/models/transformers/transformer_easyanimate.py#L316"}}),E=new se({props:{name:"forward",anchor:"diffusers.EasyAnimateTransformer3DModel.forward",parameters:[{name:"hidden_states",val:": Tensor"},{name:"timestep",val:": Tensor"},{name:"timestep_cond",val:": torch.Tensor | None = None"},{name:"encoder_hidden_states",val:": torch.Tensor | None = None"},{name:"encoder_hidden_states_t5",val:": torch.Tensor | None = None"},{name:"inpaint_latents",val:": torch.Tensor | None = None"},{name:"control_latents",val:": torch.Tensor | None = None"},{name:"return_dict",val:": bool = True"}],parametersDescription:[{anchor:"diffusers.EasyAnimateTransformer3DModel.forward.hidden_states",description:`<strong>hidden_states</strong> (<code>torch.Tensor</code> of shape <code>(batch_size, channels, num_frames, height, width)</code>) &#x2014;
Input <code>hidden_states</code>.`,name:"hidden_states"},{anchor:"diffusers.EasyAnimateTransformer3DModel.forward.timestep",description:`<strong>timestep</strong> (<code>torch.LongTensor</code>) &#x2014;
Used to indicate denoising step.`,name:"timestep"},{anchor:"diffusers.EasyAnimateTransformer3DModel.forward.timestep_cond",description:`<strong>timestep_cond</strong> (<code>torch.Tensor</code>, <em>optional</em>) &#x2014;
Conditional embeddings for timestep. If provided, the embeddings will be summed with the samples passed
through the <code>self.time_embedding</code> layer to obtain the final timestep embeddings.`,name:"timestep_cond"},{anchor:"diffusers.EasyAnimateTransformer3DModel.forward.encoder_hidden_states",description:`<strong>encoder_hidden_states</strong> (<code>torch.Tensor</code>, <em>optional</em>) &#x2014;
Conditional embeddings (embeddings computed from the input conditions such as prompts) to use.`,name:"encoder_hidden_states"},{anchor:"diffusers.EasyAnimateTransformer3DModel.forward.encoder_hidden_states_t5",description:`<strong>encoder_hidden_states_t5</strong> (<code>torch.Tensor</code>, <em>optional</em>) &#x2014;
Additional conditional embeddings computed from a T5 text encoder.`,name:"encoder_hidden_states_t5"},{anchor:"diffusers.EasyAnimateTransformer3DModel.forward.inpaint_latents",description:`<strong>inpaint_latents</strong> (<code>torch.Tensor</code>, <em>optional</em>) &#x2014;
Latents concatenated to <code>hidden_states</code> for inpainting variants of the model.`,name:"inpaint_latents"},{anchor:"diffusers.EasyAnimateTransformer3DModel.forward.control_latents",description:`<strong>control_latents</strong> (<code>torch.Tensor</code>, <em>optional</em>) &#x2014;
Latents concatenated to <code>hidden_states</code> for control variants of the model.`,name:"control_latents"},{anchor:"diffusers.EasyAnimateTransformer3DModel.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>~models.transformer_2d.Transformer2DModelOutput</code> instead of a plain
tuple.`,name:"return_dict"}],source:"https://github.com/huggingface/diffusers/blob/vr_13921/src/diffusers/models/transformers/transformer_easyanimate.py#L461",returnDescription:`<script context="module">export const metadata = 'undefined';<\/script>


<p>If <code>return_dict</code> is True, an <code>~models.transformer_2d.Transformer2DModelOutput</code> is returned, otherwise a
<code>tuple</code> where the first element is the sample tensor.</p>
`}}),A=new re({props:{title:"Transformer2DModelOutput",local:"diffusers.models.modeling_outputs.Transformer2DModelOutput",headingTag:"h2"}}),w=new se({props:{name:"class diffusers.models.modeling_outputs.Transformer2DModelOutput",anchor:"diffusers.models.modeling_outputs.Transformer2DModelOutput",parameters:[{name:"sample",val:": torch.Tensor"}],parametersDescription:[{anchor:"diffusers.models.modeling_outputs.Transformer2DModelOutput.sample",description:`<strong>sample</strong> (<code>torch.Tensor</code> of shape <code>(batch_size, num_channels, height, width)</code> or <code>(batch size, num_vector_embeds - 1, num_latent_pixels)</code> if <a href="/docs/diffusers/pr_13921/en/api/models/transformer2d#diffusers.Transformer2DModel">Transformer2DModel</a> is discrete) &#x2014;
The hidden states output conditioned on the <code>encoder_hidden_states</code> input. If discrete, returns probability
distributions for the unnoised latent pixels.`,name:"sample"}],source:"https://github.com/huggingface/diffusers/blob/vr_13921/src/diffusers/models/modeling_outputs.py#L21"}}),N=new be({props:{source:"https://github.com/huggingface/diffusers/blob/main/docs/source/en/api/models/easyanimate_transformer3d.md"}}),{c(){m=i("meta"),j=s(),J=i("p"),V=s(),c(y.$$.fragment),W=s(),c(b.$$.fragment),q=s(),M=i("p"),M.innerHTML=ie,P=s(),v=i("p"),v.textContent=de,O=s(),c($.$$.fragment),R=s(),c(x.$$.fragment),Y=s(),a=i("div"),c(D.$$.fragment),ee=s(),k=i("p"),k.innerHTML=me,te=s(),T=i("div"),c(E.$$.fragment),ne=s(),z=i("p"),z.innerHTML=le,B=s(),c(A.$$.fragment),X=s(),l=i("div"),c(w.$$.fragment),oe=s(),L=i("p"),L.innerHTML=ce,G=s(),c(N.$$.fragment),S=s(),Z=i("p"),this.h()},l(e){const t=ge("svelte-u9bgzb",document.head);m=d(t,"META",{name:!0,content:!0}),t.forEach(n),j=r(e),J=d(e,"P",{}),C(J).forEach(n),V=r(e),f(y.$$.fragment,e),W=r(e),f(b.$$.fragment,e),q=r(e),M=d(e,"P",{"data-svelte-h":!0}),F(M)!=="svelte-14a7gv2"&&(M.innerHTML=ie),P=r(e),v=d(e,"P",{"data-svelte-h":!0}),F(v)!=="svelte-1vuni30"&&(v.textContent=de),O=r(e),f($.$$.fragment,e),R=r(e),f(x.$$.fragment,e),Y=r(e),a=d(e,"DIV",{class:!0});var I=C(a);f(D.$$.fragment,I),ee=r(I),k=d(I,"P",{"data-svelte-h":!0}),F(k)!=="svelte-xn9wgl"&&(k.innerHTML=me),te=r(I),T=d(I,"DIV",{class:!0});var Q=C(T);f(E.$$.fragment,Q),ne=r(Q),z=d(Q,"P",{"data-svelte-h":!0}),F(z)!=="svelte-112ctud"&&(z.innerHTML=le),Q.forEach(n),I.forEach(n),B=r(e),f(A.$$.fragment,e),X=r(e),l=d(e,"DIV",{class:!0});var K=C(l);f(w.$$.fragment,K),oe=r(K),L=d(K,"P",{"data-svelte-h":!0}),F(L)!=="svelte-2clpd6"&&(L.innerHTML=ce),K.forEach(n),G=r(e),f(N.$$.fragment,e),S=r(e),Z=d(e,"P",{}),C(Z).forEach(n),this.h()},h(){H(m,"name","hf:doc:metadata"),H(m,"content",ve),H(T,"class","docstring border-l-2 border-t-2 pl-4 pt-3.5 border-gray-100 rounded-tl-xl mb-6 mt-8"),H(a,"class","docstring border-l-2 border-t-2 pl-4 pt-3.5 border-gray-100 rounded-tl-xl mb-6 mt-8"),H(l,"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){p(document.head,m),o(e,j,t),o(e,J,t),o(e,V,t),u(y,e,t),o(e,W,t),u(b,e,t),o(e,q,t),o(e,M,t),o(e,P,t),o(e,v,t),o(e,O,t),u($,e,t),o(e,R,t),u(x,e,t),o(e,Y,t),o(e,a,t),u(D,a,null),p(a,ee),p(a,k),p(a,te),p(a,T),u(E,T,null),p(T,ne),p(T,z),o(e,B,t),u(A,e,t),o(e,X,t),o(e,l,t),u(w,l,null),p(l,oe),p(l,L),o(e,G,t),u(N,e,t),o(e,S,t),o(e,Z,t),U=!0},p:pe,i(e){U||(_(y.$$.fragment,e),_(b.$$.fragment,e),_($.$$.fragment,e),_(x.$$.fragment,e),_(D.$$.fragment,e),_(E.$$.fragment,e),_(A.$$.fragment,e),_(w.$$.fragment,e),_(N.$$.fragment,e),U=!0)},o(e){h(y.$$.fragment,e),h(b.$$.fragment,e),h($.$$.fragment,e),h(x.$$.fragment,e),h(D.$$.fragment,e),h(E.$$.fragment,e),h(A.$$.fragment,e),h(w.$$.fragment,e),h(N.$$.fragment,e),U=!1},d(e){e&&(n(j),n(J),n(V),n(W),n(q),n(M),n(P),n(v),n(O),n(R),n(Y),n(a),n(B),n(X),n(l),n(G),n(S),n(Z)),n(m),g(y,e),g(b,e),g($,e),g(x,e),g(D),g(E),g(A,e),g(w),g(N,e)}}}const ve='{"title":"EasyAnimateTransformer3DModel","local":"easyanimatetransformer3dmodel","sections":[{"title":"EasyAnimateTransformer3DModel","local":"diffusers.EasyAnimateTransformer3DModel","sections":[],"depth":2},{"title":"Transformer2DModelOutput","local":"diffusers.models.modeling_outputs.Transformer2DModelOutput","sections":[],"depth":2}],"depth":1}';function $e(ae){return ue(()=>{new URLSearchParams(window.location.search).get("fw")}),[]}class ke extends _e{constructor(m){super(),he(this,m,$e,Me,fe,{})}}export{ke as component};