diffusers/pr_13881/en/_app/immutable/nodes/100.-OUz6dDP.js

import"../chunks/DsnmJJEf.js";import{i as y,h as N,C as O,H as n,D as s,E as L,s as k}from"../chunks/CmJXCtRL.js";import{p as z,o as A,s as e,f as C,a as g,b as D,c as t,d as T,n as a,r as i}from"../chunks/DK803DsY.js";const q='{"title":"TransformerTemporalModel","local":"transformertemporalmodel","sections":[{"title":"TransformerTemporalModel","local":"diffusers.TransformerTemporalModel","sections":[],"depth":2},{"title":"TransformerTemporalModelOutput","local":"diffusers.models.transformers.transformer_temporal.TransformerTemporalModelOutput","sections":[],"depth":2}],"depth":1}';var P=T('<meta name="hf:doc:metadata"/>'),B=T('<p></p> <!> <!> <p>A Transformer model for video-like data.</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.</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 <code>TransformerTemporal</code> 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_13881/en/api/models/transformer_temporal#diffusers.TransformerTemporalModel">TransformerTemporalModel</a>.</p></div> <!> <p></p>',1);function E(b,v){z(v,!1),A(()=>{new URLSearchParams(window.location.search).get("fw")}),y();var d=B();N("tt6mhx",h=>{var _=P();k(_,"content",q),g(h,_)});var m=e(C(d),2);O(m,{containerStyle:"float: right; margin-left: 10px; display: inline-flex; position: relative; z-index: 10;"});var l=e(m,2);n(l,{title:"TransformerTemporalModel",local:"transformertemporalmodel",headingTag:"h1"});var f=e(l,4);n(f,{title:"TransformerTemporalModel",local:"diffusers.TransformerTemporalModel",headingTag:"h2"});var o=e(f,2),c=t(o);s(c,{name:"class diffusers.TransformerTemporalModel",anchor:"diffusers.TransformerTemporalModel",source:"https://github.com/huggingface/diffusers/blob/vr_13881/src/diffusers/models/transformers/transformer_temporal.py#L41",parameters:[{name:"num_attention_heads",val:": int = 16"},{name:"attention_head_dim",val:": int = 88"},{name:"in_channels",val:": int | None = None"},{name:"out_channels",val:": int | None = None"},{name:"num_layers",val:": int = 1"},{name:"dropout",val:": float = 0.0"},{name:"norm_num_groups",val:": int = 32"},{name:"cross_attention_dim",val:": int | None = None"},{name:"attention_bias",val:": bool = False"},{name:"sample_size",val:": int | None = None"},{name:"activation_fn",val:": str = 'geglu'"},{name:"norm_elementwise_affine",val:": bool = True"},{name:"double_self_attention",val:": bool = True"},{name:"positional_embeddings",val:": str | None = None"},{name:"num_positional_embeddings",val:": int | None = None"}],parametersDescription:[{anchor:"diffusers.TransformerTemporalModel.num_attention_heads",description:"<strong>num_attention_heads</strong> (<code>int</code>, <em>optional</em>, defaults to 16) &#x2014; The number of heads to use for multi-head attention.",name:"num_attention_heads"},{anchor:"diffusers.TransformerTemporalModel.attention_head_dim",description:"<strong>attention_head_dim</strong> (<code>int</code>, <em>optional</em>, defaults to 88) &#x2014; The number of channels in each head.",name:"attention_head_dim"},{anchor:"diffusers.TransformerTemporalModel.in_channels",description:`<strong>in_channels</strong> (<code>int</code>, <em>optional</em>) &#x2014;
The number of channels in the input and output (specify if the input is <strong>continuous</strong>).`,name:"in_channels"},{anchor:"diffusers.TransformerTemporalModel.num_layers",description:"<strong>num_layers</strong> (<code>int</code>, <em>optional</em>, defaults to 1) &#x2014; The number of layers of Transformer blocks to use.",name:"num_layers"},{anchor:"diffusers.TransformerTemporalModel.dropout",description:"<strong>dropout</strong> (<code>float</code>, <em>optional</em>, defaults to 0.0) &#x2014; The dropout probability to use.",name:"dropout"},{anchor:"diffusers.TransformerTemporalModel.cross_attention_dim",description:"<strong>cross_attention_dim</strong> (<code>int</code>, <em>optional</em>) &#x2014; The number of <code>encoder_hidden_states</code> dimensions to use.",name:"cross_attention_dim"},{anchor:"diffusers.TransformerTemporalModel.attention_bias",description:`<strong>attention_bias</strong> (<code>bool</code>, <em>optional</em>) &#x2014;
Configure if the <code>TransformerBlock</code> attention should contain a bias parameter.`,name:"attention_bias"},{anchor:"diffusers.TransformerTemporalModel.sample_size",description:`<strong>sample_size</strong> (<code>int</code>, <em>optional</em>) &#x2014; The width of the latent images (specify if the input is <strong>discrete</strong>).
This is fixed during training since it is used to learn a number of position embeddings.`,name:"sample_size"},{anchor:"diffusers.TransformerTemporalModel.activation_fn",description:`<strong>activation_fn</strong> (<code>str</code>, <em>optional</em>, defaults to <code>&quot;geglu&quot;</code>) &#x2014;
Activation function to use in feed-forward. See <code>diffusers.models.activations.get_activation</code> for supported
activation functions.`,name:"activation_fn"},{anchor:"diffusers.TransformerTemporalModel.norm_elementwise_affine",description:`<strong>norm_elementwise_affine</strong> (<code>bool</code>, <em>optional</em>) &#x2014;
Configure if the <code>TransformerBlock</code> should use learnable elementwise affine parameters for normalization.`,name:"norm_elementwise_affine"},{anchor:"diffusers.TransformerTemporalModel.double_self_attention",description:`<strong>double_self_attention</strong> (<code>bool</code>, <em>optional</em>) &#x2014;
Configure if each <code>TransformerBlock</code> should contain two self-attention layers.`,name:"double_self_attention"},{anchor:"diffusers.TransformerTemporalModel.positional_embeddings",description:`<strong>positional_embeddings</strong> &#x2014; (<code>str</code>, <em>optional</em>):
The type of positional embeddings to apply to the sequence input before passing use.`,name:"positional_embeddings"},{anchor:"diffusers.TransformerTemporalModel.num_positional_embeddings",description:`<strong>num_positional_embeddings</strong> &#x2014; (<code>int</code>, <em>optional</em>):
The maximum length of the sequence over which to apply positional embeddings.`,name:"num_positional_embeddings"}]});var p=e(c,4),M=t(p);s(M,{name:"forward",anchor:"diffusers.TransformerTemporalModel.forward",source:"https://github.com/huggingface/diffusers/blob/vr_13881/src/diffusers/models/transformers/transformer_temporal.py#L123",parameters:[{name:"hidden_states",val:": Tensor"},{name:"encoder_hidden_states",val:": torch.LongTensor | None = None"},{name:"timestep",val:": torch.LongTensor | None = None"},{name:"class_labels",val:": LongTensor = None"},{name:"num_frames",val:": int = 1"},{name:"cross_attention_kwargs",val:": dict[str, typing.Any] | None = None"},{name:"return_dict",val:": bool = True"}],parametersDescription:[{anchor:"diffusers.TransformerTemporalModel.forward.hidden_states",description:`<strong>hidden_states</strong> (<code>torch.LongTensor</code> of shape <code>(batch size, num latent pixels)</code> if discrete, <code>torch.Tensor</code> of shape <code>(batch size, channel, height, width)</code> if continuous) &#x2014;
Input hidden_states.`,name:"hidden_states"},{anchor:"diffusers.TransformerTemporalModel.forward.encoder_hidden_states",description:`<strong>encoder_hidden_states</strong> ( <code>torch.LongTensor</code> of shape <code>(batch size, encoder_hidden_states dim)</code>, <em>optional</em>) &#x2014;
Conditional embeddings for cross attention layer. If not given, cross-attention defaults to
self-attention.`,name:"encoder_hidden_states"},{anchor:"diffusers.TransformerTemporalModel.forward.timestep",description:`<strong>timestep</strong> ( <code>torch.LongTensor</code>, <em>optional</em>) &#x2014;
Used to indicate denoising step. Optional timestep to be applied as an embedding in <code>AdaLayerNorm</code>.`,name:"timestep"},{anchor:"diffusers.TransformerTemporalModel.forward.class_labels",description:`<strong>class_labels</strong> ( <code>torch.LongTensor</code> of shape <code>(batch size, num classes)</code>, <em>optional</em>) &#x2014;
Used to indicate class labels conditioning. Optional class labels to be applied as an embedding in
<code>AdaLayerZeroNorm</code>.`,name:"class_labels"},{anchor:"diffusers.TransformerTemporalModel.forward.num_frames",description:`<strong>num_frames</strong> (<code>int</code>, <em>optional</em>, defaults to 1) &#x2014;
The number of frames to be processed per batch. This is used to reshape the hidden states.`,name:"num_frames"},{anchor:"diffusers.TransformerTemporalModel.forward.cross_attention_kwargs",description:`<strong>cross_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:"cross_attention_kwargs"},{anchor:"diffusers.TransformerTemporalModel.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 <a href="/docs/diffusers/pr_13881/en/api/models/transformer_temporal#diffusers.models.transformers.transformer_temporal.TransformerTemporalModelOutput">TransformerTemporalModelOutput</a>
instead of a plain tuple.`,name:"return_dict"}],returnDescription:`<script context="module">export const metadata = 'undefined';<\/script>


<p>If <code>return_dict</code> is True, an
<a
  href="/docs/diffusers/pr_13881/en/api/models/transformer_temporal#diffusers.models.transformers.transformer_temporal.TransformerTemporalModelOutput"
>TransformerTemporalModelOutput</a> is returned, otherwise a
<code>tuple</code> where the first element is the sample tensor.</p>
`,returnType:`<script context="module">export const metadata = 'undefined';<\/script>


<p><a
  href="/docs/diffusers/pr_13881/en/api/models/transformer_temporal#diffusers.models.transformers.transformer_temporal.TransformerTemporalModelOutput"
>TransformerTemporalModelOutput</a> or <code>tuple</code></p>
`}),a(2),i(p),i(o);var u=e(o,2);n(u,{title:"TransformerTemporalModelOutput",local:"diffusers.models.transformers.transformer_temporal.TransformerTemporalModelOutput",headingTag:"h2"});var r=e(u,2),x=t(r);s(x,{name:"class diffusers.models.transformers.transformer_temporal.TransformerTemporalModelOutput",anchor:"diffusers.models.transformers.transformer_temporal.TransformerTemporalModelOutput",source:"https://github.com/huggingface/diffusers/blob/vr_13881/src/diffusers/models/transformers/transformer_temporal.py#L29",parameters:[{name:"sample",val:": Tensor"}],parametersDescription:[{anchor:"diffusers.models.transformers.transformer_temporal.TransformerTemporalModelOutput.sample",description:`<strong>sample</strong> (<code>torch.Tensor</code> of shape <code>(batch_size x num_frames, num_channels, height, width)</code>) &#x2014;
The hidden states output conditioned on <code>encoder_hidden_states</code> input.`,name:"sample"}]}),a(2),i(r);var w=e(r,2);L(w,{source:"https://github.com/huggingface/diffusers/blob/main/docs/source/en/api/models/transformer_temporal.md"}),a(2),g(b,d),D()}export{E as component};