diffusers/pr_10312/en/_app/immutable/nodes/40.e871b393.js

import{s as ce,n as ue,o as le}from"../chunks/scheduler.8c3d61f6.js";import{S as me,i as _e,g as i,s as r,r as m,A as pe,h as a,f as s,c as n,j as L,u as _,x as J,k as I,y as u,a as t,v as p,d as h,t as A,w as P}from"../chunks/index.da70eac4.js";import{D as se}from"../chunks/Docstring.6b390b9a.js";import{C as he}from"../chunks/CodeBlock.00a903b3.js";import{H as te,E as Ae}from"../chunks/EditOnGithub.1e64e623.js";function Pe(re){let f,X,C,R,g,k,y,ne='A Diffusion Transformer model for 3D video-like data was introduced in <a href="https://huggingface.co/papers/2412.03603" rel="nofollow">HunyuanVideo: A Systematic Framework For Large Video Generative Models</a> by Tencent.',z,T,de="The model can be loaded with the following code snippet.",Z,b,j,$,U,d,v,Q,H,ie='A Transformer model for video-like data used in <a href="https://huggingface.co/tencent/HunyuanVideo" rel="nofollow">HunyuanVideo</a>.',B,l,M,ee,F,ae="Sets the attention processor to use to compute attention.",E,V,W,c,x,oe,w,fe='The output of <a href="/docs/diffusers/pr_10312/en/api/models/transformer2d#diffusers.Transformer2DModel">Transformer2DModel</a>.',K,D,N,S,q;return g=new te({props:{title:"HunyuanVideoTransformer3DModel",local:"hunyuanvideotransformer3dmodel",headingTag:"h1"}}),b=new he({props:{code:"ZnJvbSUyMGRpZmZ1c2VycyUyMGltcG9ydCUyMEh1bnl1YW5WaWRlb1RyYW5zZm9ybWVyM0RNb2RlbCUwQSUwQXRyYW5zZm9ybWVyJTIwJTNEJTIwSHVueXVhblZpZGVvVHJhbnNmb3JtZXIzRE1vZGVsLmZyb21fcHJldHJhaW5lZCglMjJ0ZW5jZW50JTJGSHVueXVhblZpZGVvJTIyJTJDJTIwdG9yY2hfZHR5cGUlM0R0b3JjaC5iZmxvYXQxNik=",highlighted:`<span class="hljs-keyword">from</span> diffusers <span class="hljs-keyword">import</span> HunyuanVideoTransformer3DModel

transformer = HunyuanVideoTransformer3DModel.from_pretrained(<span class="hljs-string">&quot;tencent/HunyuanVideo&quot;</span>, torch_dtype=torch.bfloat16)`,wrap:!1}}),$=new te({props:{title:"HunyuanVideoTransformer3DModel",local:"diffusers.HunyuanVideoTransformer3DModel",headingTag:"h2"}}),v=new se({props:{name:"class diffusers.HunyuanVideoTransformer3DModel",anchor:"diffusers.HunyuanVideoTransformer3DModel",parameters:[{name:"in_channels",val:": int = 16"},{name:"out_channels",val:": int = 16"},{name:"num_attention_heads",val:": int = 24"},{name:"attention_head_dim",val:": int = 128"},{name:"num_layers",val:": int = 20"},{name:"num_single_layers",val:": int = 40"},{name:"num_refiner_layers",val:": int = 2"},{name:"mlp_ratio",val:": float = 4.0"},{name:"patch_size",val:": int = 2"},{name:"patch_size_t",val:": int = 1"},{name:"qk_norm",val:": str = 'rms_norm'"},{name:"guidance_embeds",val:": bool = True"},{name:"text_embed_dim",val:": int = 4096"},{name:"pooled_projection_dim",val:": int = 768"},{name:"rope_theta",val:": float = 256.0"},{name:"rope_axes_dim",val:": typing.Tuple[int] = (16, 56, 56)"}],parametersDescription:[{anchor:"diffusers.HunyuanVideoTransformer3DModel.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.HunyuanVideoTransformer3DModel.out_channels",description:`<strong>out_channels</strong> (<code>int</code>, defaults to <code>16</code>) &#x2014;
The number of channels in the output.`,name:"out_channels"},{anchor:"diffusers.HunyuanVideoTransformer3DModel.num_attention_heads",description:`<strong>num_attention_heads</strong> (<code>int</code>, defaults to <code>24</code>) &#x2014;
The number of heads to use for multi-head attention.`,name:"num_attention_heads"},{anchor:"diffusers.HunyuanVideoTransformer3DModel.attention_head_dim",description:`<strong>attention_head_dim</strong> (<code>int</code>, defaults to <code>128</code>) &#x2014;
The number of channels in each head.`,name:"attention_head_dim"},{anchor:"diffusers.HunyuanVideoTransformer3DModel.num_layers",description:`<strong>num_layers</strong> (<code>int</code>, defaults to <code>20</code>) &#x2014;
The number of layers of dual-stream blocks to use.`,name:"num_layers"},{anchor:"diffusers.HunyuanVideoTransformer3DModel.num_single_layers",description:`<strong>num_single_layers</strong> (<code>int</code>, defaults to <code>40</code>) &#x2014;
The number of layers of single-stream blocks to use.`,name:"num_single_layers"},{anchor:"diffusers.HunyuanVideoTransformer3DModel.num_refiner_layers",description:`<strong>num_refiner_layers</strong> (<code>int</code>, defaults to <code>2</code>) &#x2014;
The number of layers of refiner blocks to use.`,name:"num_refiner_layers"},{anchor:"diffusers.HunyuanVideoTransformer3DModel.mlp_ratio",description:`<strong>mlp_ratio</strong> (<code>float</code>, defaults to <code>4.0</code>) &#x2014;
The ratio of the hidden layer size to the input size in the feedforward network.`,name:"mlp_ratio"},{anchor:"diffusers.HunyuanVideoTransformer3DModel.patch_size",description:`<strong>patch_size</strong> (<code>int</code>, defaults to <code>2</code>) &#x2014;
The size of the spatial patches to use in the patch embedding layer.`,name:"patch_size"},{anchor:"diffusers.HunyuanVideoTransformer3DModel.patch_size_t",description:`<strong>patch_size_t</strong> (<code>int</code>, defaults to <code>1</code>) &#x2014;
The size of the tmeporal patches to use in the patch embedding layer.`,name:"patch_size_t"},{anchor:"diffusers.HunyuanVideoTransformer3DModel.qk_norm",description:`<strong>qk_norm</strong> (<code>str</code>, defaults to <code>rms_norm</code>) &#x2014;
The normalization to use for the query and key projections in the attention layers.`,name:"qk_norm"},{anchor:"diffusers.HunyuanVideoTransformer3DModel.guidance_embeds",description:`<strong>guidance_embeds</strong> (<code>bool</code>, defaults to <code>True</code>) &#x2014;
Whether to use guidance embeddings in the model.`,name:"guidance_embeds"},{anchor:"diffusers.HunyuanVideoTransformer3DModel.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.HunyuanVideoTransformer3DModel.pooled_projection_dim",description:`<strong>pooled_projection_dim</strong> (<code>int</code>, defaults to <code>768</code>) &#x2014;
The dimension of the pooled projection of the text embeddings.`,name:"pooled_projection_dim"},{anchor:"diffusers.HunyuanVideoTransformer3DModel.rope_theta",description:`<strong>rope_theta</strong> (<code>float</code>, defaults to <code>256.0</code>) &#x2014;
The value of theta to use in the RoPE layer.`,name:"rope_theta"},{anchor:"diffusers.HunyuanVideoTransformer3DModel.rope_axes_dim",description:`<strong>rope_axes_dim</strong> (<code>Tuple[int]</code>, defaults to <code>(16, 56, 56)</code>) &#x2014;
The dimensions of the axes to use in the RoPE layer.`,name:"rope_axes_dim"}],source:"https://github.com/huggingface/diffusers/blob/vr_10312/src/diffusers/models/transformers/transformer_hunyuan_video.py#L503"}}),M=new se({props:{name:"set_attn_processor",anchor:"diffusers.HunyuanVideoTransformer3DModel.set_attn_processor",parameters:[{name:"processor",val:": typing.Union[diffusers.models.attention_processor.AttnProcessor, diffusers.models.attention_processor.CustomDiffusionAttnProcessor, diffusers.models.attention_processor.AttnAddedKVProcessor, diffusers.models.attention_processor.AttnAddedKVProcessor2_0, diffusers.models.attention_processor.JointAttnProcessor2_0, diffusers.models.attention_processor.PAGJointAttnProcessor2_0, diffusers.models.attention_processor.PAGCFGJointAttnProcessor2_0, diffusers.models.attention_processor.FusedJointAttnProcessor2_0, diffusers.models.attention_processor.AllegroAttnProcessor2_0, diffusers.models.attention_processor.AuraFlowAttnProcessor2_0, diffusers.models.attention_processor.FusedAuraFlowAttnProcessor2_0, diffusers.models.attention_processor.FluxAttnProcessor2_0, diffusers.models.attention_processor.FluxAttnProcessor2_0_NPU, diffusers.models.attention_processor.FusedFluxAttnProcessor2_0, diffusers.models.attention_processor.FusedFluxAttnProcessor2_0_NPU, diffusers.models.attention_processor.CogVideoXAttnProcessor2_0, diffusers.models.attention_processor.FusedCogVideoXAttnProcessor2_0, diffusers.models.attention_processor.XFormersAttnAddedKVProcessor, diffusers.models.attention_processor.XFormersAttnProcessor, diffusers.models.attention_processor.XLAFlashAttnProcessor2_0, diffusers.models.attention_processor.AttnProcessorNPU, diffusers.models.attention_processor.AttnProcessor2_0, diffusers.models.attention_processor.MochiVaeAttnProcessor2_0, diffusers.models.attention_processor.MochiAttnProcessor2_0, diffusers.models.attention_processor.StableAudioAttnProcessor2_0, diffusers.models.attention_processor.HunyuanAttnProcessor2_0, diffusers.models.attention_processor.FusedHunyuanAttnProcessor2_0, diffusers.models.attention_processor.PAGHunyuanAttnProcessor2_0, diffusers.models.attention_processor.PAGCFGHunyuanAttnProcessor2_0, diffusers.models.attention_processor.LuminaAttnProcessor2_0, diffusers.models.attention_processor.FusedAttnProcessor2_0, diffusers.models.attention_processor.CustomDiffusionXFormersAttnProcessor, diffusers.models.attention_processor.CustomDiffusionAttnProcessor2_0, diffusers.models.attention_processor.SlicedAttnProcessor, diffusers.models.attention_processor.SlicedAttnAddedKVProcessor, diffusers.models.attention_processor.SanaLinearAttnProcessor2_0, diffusers.models.attention_processor.PAGCFGSanaLinearAttnProcessor2_0, diffusers.models.attention_processor.PAGIdentitySanaLinearAttnProcessor2_0, diffusers.models.attention_processor.SanaMultiscaleLinearAttention, diffusers.models.attention_processor.SanaMultiscaleAttnProcessor2_0, diffusers.models.attention_processor.SanaMultiscaleAttentionProjection, diffusers.models.attention_processor.IPAdapterAttnProcessor, diffusers.models.attention_processor.IPAdapterAttnProcessor2_0, diffusers.models.attention_processor.IPAdapterXFormersAttnProcessor, diffusers.models.attention_processor.SD3IPAdapterJointAttnProcessor2_0, diffusers.models.attention_processor.PAGIdentitySelfAttnProcessor2_0, diffusers.models.attention_processor.PAGCFGIdentitySelfAttnProcessor2_0, diffusers.models.attention_processor.LoRAAttnProcessor, diffusers.models.attention_processor.LoRAAttnProcessor2_0, diffusers.models.attention_processor.LoRAXFormersAttnProcessor, diffusers.models.attention_processor.LoRAAttnAddedKVProcessor, typing.Dict[str, typing.Union[diffusers.models.attention_processor.AttnProcessor, diffusers.models.attention_processor.CustomDiffusionAttnProcessor, diffusers.models.attention_processor.AttnAddedKVProcessor, diffusers.models.attention_processor.AttnAddedKVProcessor2_0, diffusers.models.attention_processor.JointAttnProcessor2_0, diffusers.models.attention_processor.PAGJointAttnProcessor2_0, diffusers.models.attention_processor.PAGCFGJointAttnProcessor2_0, diffusers.models.attention_processor.FusedJointAttnProcessor2_0, diffusers.models.attention_processor.AllegroAttnProcessor2_0, diffusers.models.attention_processor.AuraFlowAttnProcessor2_0, diffusers.models.attention_processor.FusedAuraFlowAttnProcessor2_0, diffusers.models.attention_processor.FluxAttnProcessor2_0, diffusers.models.attention_processor.FluxAttnProcessor2_0_NPU, diffusers.models.attention_processor.FusedFluxAttnProcessor2_0, diffusers.models.attention_processor.FusedFluxAttnProcessor2_0_NPU, diffusers.models.attention_processor.CogVideoXAttnProcessor2_0, diffusers.models.attention_processor.FusedCogVideoXAttnProcessor2_0, diffusers.models.attention_processor.XFormersAttnAddedKVProcessor, diffusers.models.attention_processor.XFormersAttnProcessor, diffusers.models.attention_processor.XLAFlashAttnProcessor2_0, diffusers.models.attention_processor.AttnProcessorNPU, diffusers.models.attention_processor.AttnProcessor2_0, diffusers.models.attention_processor.MochiVaeAttnProcessor2_0, diffusers.models.attention_processor.MochiAttnProcessor2_0, diffusers.models.attention_processor.StableAudioAttnProcessor2_0, diffusers.models.attention_processor.HunyuanAttnProcessor2_0, diffusers.models.attention_processor.FusedHunyuanAttnProcessor2_0, diffusers.models.attention_processor.PAGHunyuanAttnProcessor2_0, diffusers.models.attention_processor.PAGCFGHunyuanAttnProcessor2_0, diffusers.models.attention_processor.LuminaAttnProcessor2_0, diffusers.models.attention_processor.FusedAttnProcessor2_0, diffusers.models.attention_processor.CustomDiffusionXFormersAttnProcessor, diffusers.models.attention_processor.CustomDiffusionAttnProcessor2_0, diffusers.models.attention_processor.SlicedAttnProcessor, diffusers.models.attention_processor.SlicedAttnAddedKVProcessor, diffusers.models.attention_processor.SanaLinearAttnProcessor2_0, diffusers.models.attention_processor.PAGCFGSanaLinearAttnProcessor2_0, diffusers.models.attention_processor.PAGIdentitySanaLinearAttnProcessor2_0, diffusers.models.attention_processor.SanaMultiscaleLinearAttention, diffusers.models.attention_processor.SanaMultiscaleAttnProcessor2_0, diffusers.models.attention_processor.SanaMultiscaleAttentionProjection, diffusers.models.attention_processor.IPAdapterAttnProcessor, diffusers.models.attention_processor.IPAdapterAttnProcessor2_0, diffusers.models.attention_processor.IPAdapterXFormersAttnProcessor, diffusers.models.attention_processor.SD3IPAdapterJointAttnProcessor2_0, diffusers.models.attention_processor.PAGIdentitySelfAttnProcessor2_0, diffusers.models.attention_processor.PAGCFGIdentitySelfAttnProcessor2_0, diffusers.models.attention_processor.LoRAAttnProcessor, diffusers.models.attention_processor.LoRAAttnProcessor2_0, diffusers.models.attention_processor.LoRAXFormersAttnProcessor, diffusers.models.attention_processor.LoRAAttnAddedKVProcessor]]]"}],parametersDescription:[{anchor:"diffusers.HunyuanVideoTransformer3DModel.set_attn_processor.processor",description:`<strong>processor</strong> (<code>dict</code> of <code>AttentionProcessor</code> or only <code>AttentionProcessor</code>) &#x2014;
The instantiated processor class or a dictionary of processor classes that will be set as the processor
for <strong>all</strong> <code>Attention</code> layers.</p>
<p>If <code>processor</code> is a dict, the key needs to define the path to the corresponding cross attention
processor. This is strongly recommended when setting trainable attention processors.`,name:"processor"}],source:"https://github.com/huggingface/diffusers/blob/vr_10312/src/diffusers/models/transformers/transformer_hunyuan_video.py#L631"}}),V=new te({props:{title:"Transformer2DModelOutput",local:"diffusers.models.modeling_outputs.Transformer2DModelOutput",headingTag:"h2"}}),x=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_10312/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_10312/src/diffusers/models/modeling_outputs.py#L20"}}),D=new Ae({props:{source:"https://github.com/huggingface/diffusers/blob/main/docs/source/en/api/models/hunyuan_video_transformer_3d.md"}}),{c(){f=i("meta"),X=r(),C=i("p"),R=r(),m(g.$$.fragment),k=r(),y=i("p"),y.innerHTML=ne,z=r(),T=i("p"),T.textContent=de,Z=r(),m(b.$$.fragment),j=r(),m($.$$.fragment),U=r(),d=i("div"),m(v.$$.fragment),Q=r(),H=i("p"),H.innerHTML=ie,B=r(),l=i("div"),m(M.$$.fragment),ee=r(),F=i("p"),F.textContent=ae,E=r(),m(V.$$.fragment),W=r(),c=i("div"),m(x.$$.fragment),oe=r(),w=i("p"),w.innerHTML=fe,K=r(),m(D.$$.fragment),N=r(),S=i("p"),this.h()},l(e){const o=pe("svelte-u9bgzb",document.head);f=a(o,"META",{name:!0,content:!0}),o.forEach(s),X=n(e),C=a(e,"P",{}),L(C).forEach(s),R=n(e),_(g.$$.fragment,e),k=n(e),y=a(e,"P",{"data-svelte-h":!0}),J(y)!=="svelte-fdngxv"&&(y.innerHTML=ne),z=n(e),T=a(e,"P",{"data-svelte-h":!0}),J(T)!=="svelte-1vuni30"&&(T.textContent=de),Z=n(e),_(b.$$.fragment,e),j=n(e),_($.$$.fragment,e),U=n(e),d=a(e,"DIV",{class:!0});var G=L(d);_(v.$$.fragment,G),Q=n(G),H=a(G,"P",{"data-svelte-h":!0}),J(H)!=="svelte-wu7xtw"&&(H.innerHTML=ie),B=n(G),l=a(G,"DIV",{class:!0});var O=L(l);_(M.$$.fragment,O),ee=n(O),F=a(O,"P",{"data-svelte-h":!0}),J(F)!=="svelte-1o77hl2"&&(F.textContent=ae),O.forEach(s),G.forEach(s),E=n(e),_(V.$$.fragment,e),W=n(e),c=a(e,"DIV",{class:!0});var Y=L(c);_(x.$$.fragment,Y),oe=n(Y),w=a(Y,"P",{"data-svelte-h":!0}),J(w)!=="svelte-yagord"&&(w.innerHTML=fe),Y.forEach(s),K=n(e),_(D.$$.fragment,e),N=n(e),S=a(e,"P",{}),L(S).forEach(s),this.h()},h(){I(f,"name","hf:doc:metadata"),I(f,"content",ge),I(l,"class","docstring border-l-2 border-t-2 pl-4 pt-3.5 border-gray-100 rounded-tl-xl mb-6 mt-8"),I(d,"class","docstring border-l-2 border-t-2 pl-4 pt-3.5 border-gray-100 rounded-tl-xl mb-6 mt-8"),I(c,"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,o){u(document.head,f),t(e,X,o),t(e,C,o),t(e,R,o),p(g,e,o),t(e,k,o),t(e,y,o),t(e,z,o),t(e,T,o),t(e,Z,o),p(b,e,o),t(e,j,o),p($,e,o),t(e,U,o),t(e,d,o),p(v,d,null),u(d,Q),u(d,H),u(d,B),u(d,l),p(M,l,null),u(l,ee),u(l,F),t(e,E,o),p(V,e,o),t(e,W,o),t(e,c,o),p(x,c,null),u(c,oe),u(c,w),t(e,K,o),p(D,e,o),t(e,N,o),t(e,S,o),q=!0},p:ue,i(e){q||(h(g.$$.fragment,e),h(b.$$.fragment,e),h($.$$.fragment,e),h(v.$$.fragment,e),h(M.$$.fragment,e),h(V.$$.fragment,e),h(x.$$.fragment,e),h(D.$$.fragment,e),q=!0)},o(e){A(g.$$.fragment,e),A(b.$$.fragment,e),A($.$$.fragment,e),A(v.$$.fragment,e),A(M.$$.fragment,e),A(V.$$.fragment,e),A(x.$$.fragment,e),A(D.$$.fragment,e),q=!1},d(e){e&&(s(X),s(C),s(R),s(k),s(y),s(z),s(T),s(Z),s(j),s(U),s(d),s(E),s(W),s(c),s(K),s(N),s(S)),s(f),P(g,e),P(b,e),P($,e),P(v),P(M),P(V,e),P(x),P(D,e)}}}const ge='{"title":"HunyuanVideoTransformer3DModel","local":"hunyuanvideotransformer3dmodel","sections":[{"title":"HunyuanVideoTransformer3DModel","local":"diffusers.HunyuanVideoTransformer3DModel","sections":[],"depth":2},{"title":"Transformer2DModelOutput","local":"diffusers.models.modeling_outputs.Transformer2DModelOutput","sections":[],"depth":2}],"depth":1}';function ye(re){return le(()=>{new URLSearchParams(window.location.search).get("fw")}),[]}class Ve extends me{constructor(f){super(),_e(this,f,ye,Pe,ce,{})}}export{Ve as component};