diffusers/pr_13966/en/_app/immutable/nodes/80.5danukOt.js

import"../chunks/DsnmJJEf.js";import{i as w,h as k,C as q,H as r,a as Z,D as t,E as z,s as N}from"../chunks/BtE7mKSK.js";import{p as J,o as j,s as e,f as X,a as M,b as B,c as s,d as T,n as d,r as a}from"../chunks/jDjavuwI.js";const O='{"title":"MochiTransformer3DModel","local":"mochitransformer3dmodel","sections":[{"title":"MochiTransformer3DModel","local":"diffusers.MochiTransformer3DModel","sections":[],"depth":2},{"title":"Transformer2DModelOutput","local":"diffusers.models.modeling_outputs.Transformer2DModelOutput","sections":[],"depth":2}],"depth":1}';var I=T('<meta name="hf:doc:metadata"/>'),L=T('<p></p> <!> <!> <p>A Diffusion Transformer model for 3D video-like data was introduced in <a href="https://huggingface.co/genmo/mochi-1-preview" rel="nofollow">Mochi-1 Preview</a> by Genmo.</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 Transformer model for video-like data introduced in <a href="https://huggingface.co/genmo/mochi-1-preview" rel="nofollow">Mochi</a>.</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>The <a href="/docs/diffusers/pr_13966/en/api/models/mochi_transformer3d#diffusers.MochiTransformer3DModel">MochiTransformer3DModel</a> forward method.</p></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>The output of <a href="/docs/diffusers/pr_13966/en/api/models/transformer2d#diffusers.Transformer2DModel">Transformer2DModel</a>.</p></div> <!> <p></p>',1);function A(b,v){J(v,!1),j(()=>{new URLSearchParams(window.location.search).get("fw")}),w();var i=L();k("14dr7i2",_=>{var g=I();N(g,"content",O),M(_,g)});var c=e(X(i),2);q(c,{containerStyle:"float: right; margin-left: 10px; display: inline-flex; position: relative; z-index: 10;"});var m=e(c,2);r(m,{title:"MochiTransformer3DModel",local:"mochitransformer3dmodel",headingTag:"h1"});var l=e(m,6);Z(l,{code:"ZnJvbSUyMGRpZmZ1c2VycyUyMGltcG9ydCUyME1vY2hpVHJhbnNmb3JtZXIzRE1vZGVsJTBBJTBBdHJhbnNmb3JtZXIlMjAlM0QlMjBNb2NoaVRyYW5zZm9ybWVyM0RNb2RlbC5mcm9tX3ByZXRyYWluZWQoJTIyZ2VubW8lMkZtb2NoaS0xLXByZXZpZXclMjIlMkMlMjBzdWJmb2xkZXIlM0QlMjJ0cmFuc2Zvcm1lciUyMiUyQyUyMHRvcmNoX2R0eXBlJTNEdG9yY2guZmxvYXQxNikudG8oJTIyY3VkYSUyMik=",highlighted:`<span class="hljs-keyword">from</span> diffusers <span class="hljs-keyword">import</span> MochiTransformer3DModel

transformer = MochiTransformer3DModel.from_pretrained(<span class="hljs-string">&quot;genmo/mochi-1-preview&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});var f=e(l,2);r(f,{title:"MochiTransformer3DModel",local:"diffusers.MochiTransformer3DModel",headingTag:"h2"});var o=e(f,2),h=s(o);t(h,{name:"class diffusers.MochiTransformer3DModel",anchor:"diffusers.MochiTransformer3DModel",source:"https://github.com/huggingface/diffusers/blob/vr_13966/src/diffusers/models/transformers/transformer_mochi.py#L309",parameters:[{name:"patch_size",val:": int = 2"},{name:"num_attention_heads",val:": int = 24"},{name:"attention_head_dim",val:": int = 128"},{name:"num_layers",val:": int = 48"},{name:"pooled_projection_dim",val:": int = 1536"},{name:"in_channels",val:": int = 12"},{name:"out_channels",val:": int | None = None"},{name:"qk_norm",val:": str = 'rms_norm'"},{name:"text_embed_dim",val:": int = 4096"},{name:"time_embed_dim",val:": int = 256"},{name:"activation_fn",val:": str = 'swiglu'"},{name:"max_sequence_length",val:": int = 256"}],parametersDescription:[{anchor:"diffusers.MochiTransformer3DModel.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.MochiTransformer3DModel.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.MochiTransformer3DModel.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.MochiTransformer3DModel.num_layers",description:`<strong>num_layers</strong> (<code>int</code>, defaults to <code>48</code>) &#x2014;
The number of layers of Transformer blocks to use.`,name:"num_layers"},{anchor:"diffusers.MochiTransformer3DModel.in_channels",description:`<strong>in_channels</strong> (<code>int</code>, defaults to <code>12</code>) &#x2014;
The number of channels in the input.`,name:"in_channels"},{anchor:"diffusers.MochiTransformer3DModel.out_channels",description:`<strong>out_channels</strong> (<code>int</code>, <em>optional</em>, defaults to <code>None</code>) &#x2014;
The number of channels in the output.`,name:"out_channels"},{anchor:"diffusers.MochiTransformer3DModel.qk_norm",description:`<strong>qk_norm</strong> (<code>str</code>, defaults to <code>&quot;rms_norm&quot;</code>) &#x2014;
The normalization layer to use.`,name:"qk_norm"},{anchor:"diffusers.MochiTransformer3DModel.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.MochiTransformer3DModel.time_embed_dim",description:`<strong>time_embed_dim</strong> (<code>int</code>, defaults to <code>256</code>) &#x2014;
Output dimension of timestep embeddings.`,name:"time_embed_dim"},{anchor:"diffusers.MochiTransformer3DModel.activation_fn",description:`<strong>activation_fn</strong> (<code>str</code>, defaults to <code>&quot;swiglu&quot;</code>) &#x2014;
Activation function to use in feed-forward.`,name:"activation_fn"},{anchor:"diffusers.MochiTransformer3DModel.max_sequence_length",description:`<strong>max_sequence_length</strong> (<code>int</code>, defaults to <code>256</code>) &#x2014;
The maximum sequence length of text embeddings supported.`,name:"max_sequence_length"}]});var u=e(h,4),D=s(u);t(D,{name:"forward",anchor:"diffusers.MochiTransformer3DModel.forward",source:"https://github.com/huggingface/diffusers/blob/vr_13966/src/diffusers/models/transformers/transformer_mochi.py#L407",parameters:[{name:"hidden_states",val:": Tensor"},{name:"encoder_hidden_states",val:": Tensor"},{name:"timestep",val:": LongTensor"},{name:"encoder_attention_mask",val:": Tensor"},{name:"attention_kwargs",val:": dict[str, typing.Any] | None = None"},{name:"return_dict",val:": bool = True"}],parametersDescription:[{anchor:"diffusers.MochiTransformer3DModel.forward.hidden_states",description:`<strong>hidden_states</strong> (<code>torch.Tensor</code> of shape <code>(batch_size, num_channels, num_frames, height, width)</code>) &#x2014;
Input <code>hidden_states</code>.`,name:"hidden_states"},{anchor:"diffusers.MochiTransformer3DModel.forward.encoder_hidden_states",description:`<strong>encoder_hidden_states</strong> (<code>torch.Tensor</code> of shape <code>(batch_size, sequence_len, embed_dims)</code>) &#x2014;
Conditional embeddings (embeddings computed from the input conditions such as prompts) to use.`,name:"encoder_hidden_states"},{anchor:"diffusers.MochiTransformer3DModel.forward.timestep",description:`<strong>timestep</strong> (<code>torch.LongTensor</code>) &#x2014;
Used to indicate denoising step.`,name:"timestep"},{anchor:"diffusers.MochiTransformer3DModel.forward.encoder_attention_mask",description:`<strong>encoder_attention_mask</strong> (<code>torch.Tensor</code>) &#x2014;
Mask applied to <code>encoder_hidden_states</code> during attention.`,name:"encoder_attention_mask"},{anchor:"diffusers.MochiTransformer3DModel.forward.attention_kwargs",description:`<strong>attention_kwargs</strong> (<code>dict</code>, <em>optional</em>) &#x2014;
A kwargs dictionary that if specified is passed along to the <code>AttentionProcessor</code> as defined under
<code>self.processor</code> in
<a href="https://github.com/huggingface/diffusers/blob/main/src/diffusers/models/attention_processor.py" rel="nofollow">diffusers.models.attention_processor</a>.`,name:"attention_kwargs"},{anchor:"diffusers.MochiTransformer3DModel.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"}],returnDescription:`<script context="module">export const metadata = 'undefined';<\/script>


<p>The denoised output tensor of shape <code>(batch_size, out_channels, num_frames, height, width)</code>.</p>
`,returnType:`<script context="module">export const metadata = 'undefined';<\/script>


<p><code>torch.Tensor</code></p>
`}),d(2),a(u),a(o);var p=e(o,2);r(p,{title:"Transformer2DModelOutput",local:"diffusers.models.modeling_outputs.Transformer2DModelOutput",headingTag:"h2"});var n=e(p,2),x=s(n);t(x,{name:"class diffusers.models.modeling_outputs.Transformer2DModelOutput",anchor:"diffusers.models.modeling_outputs.Transformer2DModelOutput",source:"https://github.com/huggingface/diffusers/blob/vr_13966/src/diffusers/models/modeling_outputs.py#L21",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_13966/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"}]}),d(2),a(n);var y=e(n,2);z(y,{source:"https://github.com/huggingface/diffusers/blob/main/docs/source/en/api/models/mochi_transformer3d.md"}),d(2),M(b,i),B()}export{A as component};