diffusers/pr_13881/en/_app/immutable/nodes/109.CldOSnCY.js

import"../chunks/DsnmJJEf.js";import{i as z,h as D,C as Z,H as h,D as a,E as w,s as k}from"../chunks/BtE7mKSK.js";import{p as L,o as N,s as e,f as P,a as u,b as A,c as o,d as v,n as r,r as s}from"../chunks/jDjavuwI.js";const S='{"title":"ZImageTransformer2DModel","local":"zimagetransformer2dmodel","sections":[{"title":"ZImageTransformer2DModel","local":"diffusers.ZImageTransformer2DModel","sections":[],"depth":2}],"depth":1}';var C=v('<meta name="hf:doc:metadata"/>'),O=v(`<p></p> <!> <!> <p>A Transformer model for image-like data from <a href="https://huggingface.co/Tongyi-MAI/Z-Image-Turbo" rel="nofollow">Z-Image</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"><!> <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_13881/en/api/models/z_image_transformer2d#diffusers.ZImageTransformer2DModel">ZImageTransformer2DModel</a> forward method.</p> <p>Flow: patchify -> t_embed -> x_embed -> x_refine -> cap_embed -> cap_refine
-> [siglip_embed -> siglip_refine] -> build_unified -> main_layers -> final_layer -> unpatchify</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>Patchify for basic mode: single image per batch item.</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>Patchify for omni mode: multiple images per batch item with noise masks.</p></div></div> <!> <p></p>`,1);function q(b,T){L(T,!1),N(()=>{new URLSearchParams(window.location.search).get("fw")}),z();var d=O();D("13z97x7",_=>{var g=C();k(g,"content",S),u(_,g)});var m=e(P(d),2);Z(m,{containerStyle:"float: right; margin-left: 10px; display: inline-flex; position: relative; z-index: 10;"});var l=e(m,2);h(l,{title:"ZImageTransformer2DModel",local:"zimagetransformer2dmodel",headingTag:"h1"});var c=e(l,4);h(c,{title:"ZImageTransformer2DModel",local:"diffusers.ZImageTransformer2DModel",headingTag:"h2"});var n=e(c,2),f=o(n);a(f,{name:"class diffusers.ZImageTransformer2DModel",anchor:"diffusers.ZImageTransformer2DModel",source:"https://github.com/huggingface/diffusers/blob/vr_13881/src/diffusers/models/transformers/transformer_z_image.py#L359",parameters:[{name:"all_patch_size",val:" = (2,)"},{name:"all_f_patch_size",val:" = (1,)"},{name:"in_channels",val:" = 16"},{name:"dim",val:" = 3840"},{name:"n_layers",val:" = 30"},{name:"n_refiner_layers",val:" = 2"},{name:"n_heads",val:" = 30"},{name:"n_kv_heads",val:" = 30"},{name:"norm_eps",val:" = 1e-05"},{name:"qk_norm",val:" = True"},{name:"cap_feat_dim",val:" = 2560"},{name:"siglip_feat_dim",val:" = None"},{name:"rope_theta",val:" = 256.0"},{name:"t_scale",val:" = 1000.0"},{name:"axes_dims",val:" = [32, 48, 48]"},{name:"axes_lens",val:" = [1024, 512, 512]"}]});var t=e(f,2),x=o(t);a(x,{name:"forward",anchor:"diffusers.ZImageTransformer2DModel.forward",source:"https://github.com/huggingface/diffusers/blob/vr_13881/src/diffusers/models/transformers/transformer_z_image.py#L894",parameters:[{name:"x",val:": list"},{name:"t",val:""},{name:"cap_feats",val:": list"},{name:"return_dict",val:": bool = True"},{name:"controlnet_block_samples",val:": dict[int, torch.Tensor] | None = None"},{name:"siglip_feats",val:": list[list[torch.Tensor]] | None = None"},{name:"image_noise_mask",val:": list[list[int]] | None = None"},{name:"patch_size",val:": int = 2"},{name:"f_patch_size",val:": int = 1"}],parametersDescription:[{anchor:"diffusers.ZImageTransformer2DModel.forward.x",description:`<strong>x</strong> (<code>list</code> of <code>torch.Tensor</code> or nested <code>list</code> of <code>torch.Tensor</code>) &#x2014;
Input latents. A flat list when running in standard mode, or a nested list when running in omni mode.`,name:"x"},{anchor:"diffusers.ZImageTransformer2DModel.forward.t",description:`<strong>t</strong> (<code>torch.Tensor</code>) &#x2014;
Used to indicate denoising step.`,name:"t"},{anchor:"diffusers.ZImageTransformer2DModel.forward.cap_feats",description:`<strong>cap_feats</strong> (<code>list</code> of <code>torch.Tensor</code> or nested <code>list</code> of <code>torch.Tensor</code>) &#x2014;
Conditional caption embeddings (embeddings computed from the input conditions such as prompts) to use.`,name:"cap_feats"},{anchor:"diffusers.ZImageTransformer2DModel.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"},{anchor:"diffusers.ZImageTransformer2DModel.forward.controlnet_block_samples",description:`<strong>controlnet_block_samples</strong> (<code>dict</code> of <code>int</code> to <code>torch.Tensor</code>, <em>optional</em>) &#x2014;
A mapping from block index to tensor that if specified are added to the residuals of transformer
blocks.`,name:"controlnet_block_samples"},{anchor:"diffusers.ZImageTransformer2DModel.forward.siglip_feats",description:`<strong>siglip_feats</strong> (<code>list</code> of <code>list</code> of <code>torch.Tensor</code>, <em>optional</em>) &#x2014;
Optional SigLIP image features used as additional conditioning.`,name:"siglip_feats"},{anchor:"diffusers.ZImageTransformer2DModel.forward.image_noise_mask",description:`<strong>image_noise_mask</strong> (<code>list</code> of <code>list</code> of <code>int</code>, <em>optional</em>) &#x2014;
Per-image noise masks indicating noisy vs. clean tokens in omni mode.`,name:"image_noise_mask"},{anchor:"diffusers.ZImageTransformer2DModel.forward.patch_size",description:`<strong>patch_size</strong> (<code>int</code>, <em>optional</em>, defaults to 2) &#x2014;
Spatial patch size used to patchify the input latents.`,name:"patch_size"},{anchor:"diffusers.ZImageTransformer2DModel.forward.f_patch_size",description:`<strong>f_patch_size</strong> (<code>int</code>, <em>optional</em>, defaults to 1) &#x2014;
Temporal patch size used to patchify the input latents.`,name:"f_patch_size"}]}),r(4),s(t);var i=e(t,2),y=o(i);a(y,{name:"patchify_and_embed",anchor:"diffusers.ZImageTransformer2DModel.patchify_and_embed",source:"https://github.com/huggingface/diffusers/blob/vr_13881/src/diffusers/models/transformers/transformer_z_image.py#L588",parameters:[{name:"all_image",val:": list"},{name:"all_cap_feats",val:": list"},{name:"patch_size",val:": int"},{name:"f_patch_size",val:": int"}]}),r(2),s(i);var p=e(i,2),I=o(p);a(I,{name:"patchify_and_embed_omni",anchor:"diffusers.ZImageTransformer2DModel.patchify_and_embed_omni",source:"https://github.com/huggingface/diffusers/blob/vr_13881/src/diffusers/models/transformers/transformer_z_image.py#L625",parameters:[{name:"all_x",val:": list"},{name:"all_cap_feats",val:": list"},{name:"all_siglip_feats",val:": list"},{name:"patch_size",val:": int"},{name:"f_patch_size",val:": int"},{name:"images_noise_mask",val:": list"}]}),r(2),s(p),s(n);var M=e(n,2);w(M,{source:"https://github.com/huggingface/diffusers/blob/main/docs/source/en/api/models/z_image_transformer2d.md"}),r(2),u(b,d),A()}export{q as component};