diffusers/pr_14229/en/_app/immutable/nodes/182.86eYY2qb.js

import"../chunks/DsnmJJEf.js";import{i as W,h as Z,H as t,a as s,D as r,E as j,s as B}from"../chunks/BtE7mKSK.js";import{p as P,o as Y,s as e,f as v,a as l,b as N,c as d,d as c,r as p,n as J}from"../chunks/jDjavuwI.js";import{E as C}from"../chunks/SrSJA0zO.js";const E='{"title":"Lumina2","local":"lumina2","sections":[{"title":"Using Single File loading with Lumina Image 2.0","local":"using-single-file-loading-with-lumina-image-20","sections":[],"depth":2},{"title":"Using GGUF Quantized Checkpoints with Lumina Image 2.0","local":"using-gguf-quantized-checkpoints-with-lumina-image-20","sections":[],"depth":2},{"title":"Lumina2Pipeline","local":"diffusers.Lumina2Pipeline","sections":[],"depth":2}],"depth":1}';var z=c('<meta name="hf:doc:metadata"/>'),F=c("<p>Examples:</p> <!>",1),R=c(`<p></p> <!> <div class="flex flex-wrap space-x-1"><img alt="LoRA" src="https://img.shields.io/badge/LoRA-d8b4fe?style=flat"/></div> <p><a href="https://huggingface.co/Alpha-VLLM/Lumina-Image-2.0" rel="nofollow">Lumina Image 2.0: A Unified and Efficient Image Generative Model</a> is a 2 billion parameter flow-based diffusion transformer capable of generating diverse images from text descriptions.</p> <p>The abstract from the paper is:</p> <p><em>We introduce Lumina-Image 2.0, an advanced text-to-image model that surpasses previous state-of-the-art methods across multiple benchmarks, while also shedding light on its potential to evolve into a generalist vision intelligence model. Lumina-Image 2.0 exhibits three key properties: (1) Unification – it adopts a unified architecture that treats text and image tokens as a joint sequence, enabling natural cross-modal interactions and facilitating task expansion. Besides, since high-quality captioners can provide semantically better-aligned text-image training pairs, we introduce a unified captioning system, UniCaptioner, which generates comprehensive and precise captions for the model. This not only accelerates model convergence but also enhances prompt adherence, variable-length prompt handling, and task generalization via prompt templates. (2) Efficiency – to improve the efficiency of the unified architecture, we develop a set of optimization techniques that improve semantic learning and fine-grained texture generation during training while incorporating inference-time acceleration strategies without compromising image quality. (3) Transparency – we open-source all training details, code, and models to ensure full reproducibility, aiming to bridge the gap between well-resourced closed-source research teams and independent developers.</em></p> <blockquote class="tip"><p>Make sure to check out the Schedulers <a href="../../using-diffusers/schedulers">guide</a> to learn how to explore the tradeoff between scheduler speed and quality, and see the <a href="../../using-diffusers/loading#reuse-a-pipeline">reuse components across pipelines</a> section to learn how to efficiently load the same components into multiple pipelines.</p></blockquote> <!> <p>Single file loading for Lumina Image 2.0 is available for the <code>Lumina2Transformer2DModel</code></p> <!> <!> <p>GGUF Quantized checkpoints for the <code>Lumina2Transformer2DModel</code> can be loaded via <code>from_single_file</code> with the <code>GGUFQuantizationConfig</code></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>Pipeline for text-to-image generation using Lumina-T2I.</p> <p>This model inherits from <a href="/docs/diffusers/pr_14229/en/api/pipelines/overview#diffusers.DiffusionPipeline">DiffusionPipeline</a>. Check the superclass documentation for the generic methods the
library implements for all the pipelines (such as downloading or saving, running on a particular device, etc.)</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>Function invoked when calling the pipeline for generation.</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>Encodes the prompt into text encoder hidden states.</p></div></div> <!> <p></p>`,1);function X(x,U){P(U,!1),Y(()=>{new URLSearchParams(window.location.search).get("fw")}),W();var m=R();Z("1q88c3q",n=>{var i=z();B(i,"content",E),l(n,i)});var u=e(v(m),2);t(u,{title:"Lumina2",local:"lumina2",headingTag:"h1"});var g=e(u,12);t(g,{title:"Using Single File loading with Lumina Image 2.0",local:"using-single-file-loading-with-lumina-image-20",headingTag:"h2"});var h=e(g,4);s(h,{code:"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",highlighted:`<span class="hljs-keyword">import</span> torch
<span class="hljs-keyword">from</span> diffusers <span class="hljs-keyword">import</span> Lumina2Transformer2DModel, Lumina2Pipeline

ckpt_path = <span class="hljs-string">&quot;https://huggingface.co/Alpha-VLLM/Lumina-Image-2.0/blob/main/consolidated.00-of-01.pth&quot;</span>
transformer = Lumina2Transformer2DModel.from_single_file(
    ckpt_path, torch_dtype=torch.bfloat16
)

pipe = Lumina2Pipeline.from_pretrained(
    <span class="hljs-string">&quot;Alpha-VLLM/Lumina-Image-2.0&quot;</span>, transformer=transformer, torch_dtype=torch.bfloat16
)
pipe.enable_model_cpu_offload()
image = pipe(
    <span class="hljs-string">&quot;a cat holding a sign that says hello&quot;</span>,
    generator=torch.Generator(<span class="hljs-string">&quot;cpu&quot;</span>).manual_seed(<span class="hljs-number">0</span>),
).images[<span class="hljs-number">0</span>]
image.save(<span class="hljs-string">&quot;lumina-single-file.png&quot;</span>)
`,lang:"python",wrap:!1});var f=e(h,2);t(f,{title:"Using GGUF Quantized Checkpoints with Lumina Image 2.0",local:"using-gguf-quantized-checkpoints-with-lumina-image-20",headingTag:"h2"});var _=e(f,4);s(_,{code:"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",highlighted:`<span class="hljs-keyword">from</span> diffusers <span class="hljs-keyword">import</span> Lumina2Transformer2DModel, Lumina2Pipeline, GGUFQuantizationConfig 

ckpt_path = <span class="hljs-string">&quot;https://huggingface.co/calcuis/lumina-gguf/blob/main/lumina2-q4_0.gguf&quot;</span>
transformer = Lumina2Transformer2DModel.from_single_file(
    ckpt_path,
    quantization_config=GGUFQuantizationConfig(compute_dtype=torch.bfloat16),
    torch_dtype=torch.bfloat16,
)

pipe = Lumina2Pipeline.from_pretrained(
    <span class="hljs-string">&quot;Alpha-VLLM/Lumina-Image-2.0&quot;</span>, transformer=transformer, torch_dtype=torch.bfloat16
)
pipe.enable_model_cpu_offload()
image = pipe(
    <span class="hljs-string">&quot;a cat holding a sign that says hello&quot;</span>,
    generator=torch.Generator(<span class="hljs-string">&quot;cpu&quot;</span>).manual_seed(<span class="hljs-number">0</span>),
).images[<span class="hljs-number">0</span>]
image.save(<span class="hljs-string">&quot;lumina-gguf.png&quot;</span>)`,lang:"python",wrap:!1});var b=e(_,2);t(b,{title:"Lumina2Pipeline",local:"diffusers.Lumina2Pipeline",headingTag:"h2"});var o=e(b,2),y=d(o);r(y,{name:"class diffusers.Lumina2Pipeline",anchor:"diffusers.Lumina2Pipeline",source:"https://github.com/huggingface/diffusers/blob/vr_14229/src/diffusers/pipelines/lumina2/pipeline_lumina2.py#L137",parameters:[{name:"transformer",val:": Lumina2Transformer2DModel"},{name:"scheduler",val:": FlowMatchEulerDiscreteScheduler"},{name:"vae",val:": AutoencoderKL"},{name:"text_encoder",val:": Gemma2PreTrainedModel"},{name:"tokenizer",val:": GemmaTokenizer"}],parametersDescription:[{anchor:"diffusers.Lumina2Pipeline.vae",description:`<strong>vae</strong> (<a href="/docs/diffusers/pr_14229/en/api/models/autoencoderkl#diffusers.AutoencoderKL">AutoencoderKL</a>) &#x2014;
Variational Auto-Encoder (VAE) Model to encode and decode images to and from latent representations.`,name:"vae"},{anchor:"diffusers.Lumina2Pipeline.text_encoder",description:`<strong>text_encoder</strong> (<code>Gemma2PreTrainedModel</code>) &#x2014;
Frozen Gemma2 text-encoder.`,name:"text_encoder"},{anchor:"diffusers.Lumina2Pipeline.tokenizer",description:`<strong>tokenizer</strong> (<code>GemmaTokenizer</code> or <code>GemmaTokenizerFast</code>) &#x2014;
Gemma tokenizer.`,name:"tokenizer"},{anchor:"diffusers.Lumina2Pipeline.transformer",description:`<strong>transformer</strong> (<a href="/docs/diffusers/pr_14229/en/api/models/transformer2d#diffusers.Transformer2DModel">Transformer2DModel</a>) &#x2014;
A text conditioned <code>Transformer2DModel</code> to denoise the encoded image latents.`,name:"transformer"},{anchor:"diffusers.Lumina2Pipeline.scheduler",description:`<strong>scheduler</strong> (<a href="/docs/diffusers/pr_14229/en/api/schedulers/overview#diffusers.SchedulerMixin">SchedulerMixin</a>) &#x2014;
A scheduler to be used in combination with <code>transformer</code> to denoise the encoded image latents.`,name:"scheduler"}]});var a=e(y,6),w=d(a);r(w,{name:"__call__",anchor:"diffusers.Lumina2Pipeline.__call__",source:"https://github.com/huggingface/diffusers/blob/vr_14229/src/diffusers/pipelines/lumina2/pipeline_lumina2.py#L471",parameters:[{name:"prompt",val:": str | list[str] = None"},{name:"width",val:": int | None = None"},{name:"height",val:": int | None = None"},{name:"num_inference_steps",val:": int = 30"},{name:"guidance_scale",val:": float = 4.0"},{name:"negative_prompt",val:": str | list[str] = None"},{name:"sigmas",val:": list = None"},{name:"num_images_per_prompt",val:": int | None = 1"},{name:"generator",val:": typing.Union[torch.Generator, list[torch.Generator], NoneType] = None"},{name:"latents",val:": typing.Optional[torch.Tensor] = None"},{name:"prompt_embeds",val:": typing.Optional[torch.Tensor] = None"},{name:"negative_prompt_embeds",val:": typing.Optional[torch.Tensor] = None"},{name:"prompt_attention_mask",val:": typing.Optional[torch.Tensor] = None"},{name:"negative_prompt_attention_mask",val:": typing.Optional[torch.Tensor] = None"},{name:"output_type",val:": str | None = 'pil'"},{name:"return_dict",val:": bool = True"},{name:"attention_kwargs",val:": dict[str, typing.Any] | None = None"},{name:"callback_on_step_end",val:": typing.Optional[typing.Callable[[int, int], NoneType]] = None"},{name:"callback_on_step_end_tensor_inputs",val:": list = ['latents']"},{name:"system_prompt",val:": str | None = None"},{name:"cfg_trunc_ratio",val:": float = 1.0"},{name:"cfg_normalization",val:": bool = True"},{name:"max_sequence_length",val:": int = 256"}],parametersDescription:[{anchor:"diffusers.Lumina2Pipeline.__call__.prompt",description:`<strong>prompt</strong> (<code>str</code> or <code>list[str]</code>, <em>optional</em>) &#x2014;
The prompt or prompts to guide the image generation. If not defined, one has to pass <code>prompt_embeds</code>.
instead.`,name:"prompt"},{anchor:"diffusers.Lumina2Pipeline.__call__.negative_prompt",description:`<strong>negative_prompt</strong> (<code>str</code> or <code>list[str]</code>, <em>optional</em>) &#x2014;
The prompt or prompts not to guide the image generation. If not defined, one has to pass
<code>negative_prompt_embeds</code> instead. Ignored when not using guidance (i.e., ignored if <code>guidance_scale</code> is
less than <code>1</code>).`,name:"negative_prompt"},{anchor:"diffusers.Lumina2Pipeline.__call__.num_inference_steps",description:`<strong>num_inference_steps</strong> (<code>int</code>, <em>optional</em>, defaults to 30) &#x2014;
The number of denoising steps. More denoising steps usually lead to a higher quality image at the
expense of slower inference.`,name:"num_inference_steps"},{anchor:"diffusers.Lumina2Pipeline.__call__.sigmas",description:`<strong>sigmas</strong> (<code>list[float]</code>, <em>optional</em>) &#x2014;
Custom sigmas to use for the denoising process with schedulers which support a <code>sigmas</code> argument in
their <code>set_timesteps</code> method. If not defined, the default behavior when <code>num_inference_steps</code> is passed
will be used.`,name:"sigmas"},{anchor:"diffusers.Lumina2Pipeline.__call__.guidance_scale",description:`<strong>guidance_scale</strong> (<code>float</code>, <em>optional</em>, defaults to 4.0) &#x2014;
Guidance scale as defined in <a href="https://huggingface.co/papers/2207.12598" rel="nofollow">Classifier-Free Diffusion
Guidance</a>. <code>guidance_scale</code> is defined as <code>w</code> of equation 2.
of <a href="https://huggingface.co/papers/2205.11487" rel="nofollow">Imagen Paper</a>. Guidance scale is enabled by setting
<code>guidance_scale &gt; 1</code>. Higher guidance scale encourages to generate images that are closely linked to
the text <code>prompt</code>, usually at the expense of lower image quality.`,name:"guidance_scale"},{anchor:"diffusers.Lumina2Pipeline.__call__.num_images_per_prompt",description:`<strong>num_images_per_prompt</strong> (<code>int</code>, <em>optional</em>, defaults to 1) &#x2014;
The number of images to generate per prompt.`,name:"num_images_per_prompt"},{anchor:"diffusers.Lumina2Pipeline.__call__.height",description:`<strong>height</strong> (<code>int</code>, <em>optional</em>, defaults to self.unet.config.sample_size) &#x2014;
The height in pixels of the generated image.`,name:"height"},{anchor:"diffusers.Lumina2Pipeline.__call__.width",description:`<strong>width</strong> (<code>int</code>, <em>optional</em>, defaults to self.unet.config.sample_size) &#x2014;
The width in pixels of the generated image.`,name:"width"},{anchor:"diffusers.Lumina2Pipeline.__call__.generator",description:`<strong>generator</strong> (<code>torch.Generator</code> or <code>list[torch.Generator]</code>, <em>optional</em>) &#x2014;
One or a list of <a href="https://pytorch.org/docs/stable/generated/torch.Generator.html" rel="nofollow">torch generator(s)</a>
to make generation deterministic.`,name:"generator"},{anchor:"diffusers.Lumina2Pipeline.__call__.latents",description:`<strong>latents</strong> (<code>torch.Tensor</code>, <em>optional</em>) &#x2014;
Pre-generated noisy latents, sampled from a Gaussian distribution, to be used as inputs for image
generation. Can be used to tweak the same generation with different prompts. If not provided, a latents
tensor will be generated by sampling using the supplied random <code>generator</code>.`,name:"latents"},{anchor:"diffusers.Lumina2Pipeline.__call__.prompt_embeds",description:`<strong>prompt_embeds</strong> (<code>torch.Tensor</code>, <em>optional</em>) &#x2014;
Pre-generated text embeddings. Can be used to easily tweak text inputs, <em>e.g.</em> prompt weighting. If not
provided, text embeddings will be generated from <code>prompt</code> input argument.`,name:"prompt_embeds"},{anchor:"diffusers.Lumina2Pipeline.__call__.prompt_attention_mask",description:"<strong>prompt_attention_mask</strong> (<code>torch.Tensor</code>, <em>optional</em>) &#x2014; Pre-generated attention mask for text embeddings.",name:"prompt_attention_mask"},{anchor:"diffusers.Lumina2Pipeline.__call__.negative_prompt_embeds",description:`<strong>negative_prompt_embeds</strong> (<code>torch.Tensor</code>, <em>optional</em>) &#x2014;
Pre-generated negative text embeddings. For Lumina-T2I this negative prompt should be &quot;&quot;. If not
provided, negative_prompt_embeds will be generated from <code>negative_prompt</code> input argument.`,name:"negative_prompt_embeds"},{anchor:"diffusers.Lumina2Pipeline.__call__.negative_prompt_attention_mask",description:`<strong>negative_prompt_attention_mask</strong> (<code>torch.Tensor</code>, <em>optional</em>) &#x2014;
Pre-generated attention mask for negative text embeddings.`,name:"negative_prompt_attention_mask"},{anchor:"diffusers.Lumina2Pipeline.__call__.output_type",description:`<strong>output_type</strong> (<code>str</code>, <em>optional</em>, defaults to <code>&quot;pil&quot;</code>) &#x2014;
The output format of the generate image. Choose between
<a href="https://pillow.readthedocs.io/en/stable/" rel="nofollow">PIL</a>: <code>PIL.Image.Image</code> or <code>np.array</code>.`,name:"output_type"},{anchor:"diffusers.Lumina2Pipeline.__call__.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>~pipelines.stable_diffusion.IFPipelineOutput</code> instead of a plain tuple.`,name:"return_dict"},{anchor:"diffusers.Lumina2Pipeline.__call__.attention_kwargs",description:`<strong>attention_kwargs</strong> &#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.Lumina2Pipeline.__call__.callback_on_step_end",description:`<strong>callback_on_step_end</strong> (<code>Callable</code>, <em>optional</em>) &#x2014;
A function that calls at the end of each denoising steps during the inference. The function is called
with the following arguments: <code>callback_on_step_end(self: DiffusionPipeline, step: int, timestep: int, callback_kwargs: Dict)</code>. <code>callback_kwargs</code> will include a list of all tensors as specified by
<code>callback_on_step_end_tensor_inputs</code>.`,name:"callback_on_step_end"},{anchor:"diffusers.Lumina2Pipeline.__call__.callback_on_step_end_tensor_inputs",description:`<strong>callback_on_step_end_tensor_inputs</strong> (<code>list</code>, <em>optional</em>) &#x2014;
The list of tensor inputs for the <code>callback_on_step_end</code> function. The tensors specified in the list
will be passed as <code>callback_kwargs</code> argument. You will only be able to include variables listed in the
<code>._callback_tensor_inputs</code> attribute of your pipeline class.`,name:"callback_on_step_end_tensor_inputs"},{anchor:"diffusers.Lumina2Pipeline.__call__.system_prompt",description:`<strong>system_prompt</strong> (<code>str</code>, <em>optional</em>) &#x2014;
The system prompt to use for the image generation.`,name:"system_prompt"},{anchor:"diffusers.Lumina2Pipeline.__call__.cfg_trunc_ratio",description:`<strong>cfg_trunc_ratio</strong> (<code>float</code>, <em>optional</em>, defaults to <code>1.0</code>) &#x2014;
The ratio of the timestep interval to apply normalization-based guidance scale.`,name:"cfg_trunc_ratio"},{anchor:"diffusers.Lumina2Pipeline.__call__.cfg_normalization",description:`<strong>cfg_normalization</strong> (<code>bool</code>, <em>optional</em>, defaults to <code>True</code>) &#x2014;
Whether to apply normalization-based guidance scale.`,name:"cfg_normalization"},{anchor:"diffusers.Lumina2Pipeline.__call__.max_sequence_length",description:`<strong>max_sequence_length</strong> (<code>int</code>, defaults to <code>256</code>) &#x2014;
Maximum sequence length to use with the <code>prompt</code>.`,name:"max_sequence_length"}],returnDescription:`<script context="module">export const metadata = 'undefined';<\/script>


<p>If <code>return_dict</code> is <code>True</code>, <a
  href="/docs/diffusers/pr_14229/en/api/pipelines/latent_diffusion#diffusers.ImagePipelineOutput"
>ImagePipelineOutput</a> is returned, otherwise a <code>tuple</code> is
returned where the first element is a list with the generated images</p>
`,returnType:`<script context="module">export const metadata = 'undefined';<\/script>


<p><a
  href="/docs/diffusers/pr_14229/en/api/pipelines/latent_diffusion#diffusers.ImagePipelineOutput"
>ImagePipelineOutput</a> or <code>tuple</code></p>
`});var k=e(w,4);C(k,{anchor:"diffusers.Lumina2Pipeline.__call__.example",children:(n,i)=>{var T=F(),G=e(v(T),2);s(G,{code:"aW1wb3J0JTIwdG9yY2glMEFmcm9tJTIwZGlmZnVzZXJzJTIwaW1wb3J0JTIwTHVtaW5hMlBpcGVsaW5lJTBBJTBBcGlwZSUyMCUzRCUyMEx1bWluYTJQaXBlbGluZS5mcm9tX3ByZXRyYWluZWQoJTIyQWxwaGEtVkxMTSUyRkx1bWluYS1JbWFnZS0yLjAlMjIlMkMlMjB0b3JjaF9kdHlwZSUzRHRvcmNoLmJmbG9hdDE2KSUwQSUyMyUyMEVuYWJsZSUyMG1lbW9yeSUyMG9wdGltaXphdGlvbnMuJTBBcGlwZS5lbmFibGVfbW9kZWxfY3B1X29mZmxvYWQoKSUwQSUwQXByb21wdCUyMCUzRCUyMCUyMlVwcGVyJTIwYm9keSUyMG9mJTIwYSUyMHlvdW5nJTIwd29tYW4lMjBpbiUyMGElMjBWaWN0b3JpYW4tZXJhJTIwb3V0Zml0JTIwd2l0aCUyMGJyYXNzJTIwZ29nZ2xlcyUyMGFuZCUyMGxlYXRoZXIlMjBzdHJhcHMuJTIwQmFja2dyb3VuZCUyMHNob3dzJTIwYW4lMjBpbmR1c3RyaWFsJTIwcmV2b2x1dGlvbiUyMGNpdHlzY2FwZSUyMHdpdGglMjBzbW9reSUyMHNraWVzJTIwYW5kJTIwdGFsbCUyQyUyMG1ldGFsJTIwc3RydWN0dXJlcyUyMiUwQWltYWdlJTIwJTNEJTIwcGlwZShwcm9tcHQpLmltYWdlcyU1QjAlNUQ=",highlighted:`<span class="hljs-meta">&gt;&gt;&gt; </span><span class="hljs-keyword">import</span> torch
<span class="hljs-meta">&gt;&gt;&gt; </span><span class="hljs-keyword">from</span> diffusers <span class="hljs-keyword">import</span> Lumina2Pipeline

<span class="hljs-meta">&gt;&gt;&gt; </span>pipe = Lumina2Pipeline.from_pretrained(<span class="hljs-string">&quot;Alpha-VLLM/Lumina-Image-2.0&quot;</span>, torch_dtype=torch.bfloat16)
<span class="hljs-meta">&gt;&gt;&gt; </span><span class="hljs-comment"># Enable memory optimizations.</span>
<span class="hljs-meta">&gt;&gt;&gt; </span>pipe.enable_model_cpu_offload()

<span class="hljs-meta">&gt;&gt;&gt; </span>prompt = <span class="hljs-string">&quot;Upper body of a young woman in a Victorian-era outfit with brass goggles and leather straps. Background shows an industrial revolution cityscape with smoky skies and tall, metal structures&quot;</span>
<span class="hljs-meta">&gt;&gt;&gt; </span>image = pipe(prompt).images[<span class="hljs-number">0</span>]`,lang:"py",wrap:!1}),l(n,T)},$$slots:{default:!0}}),p(a);var M=e(a,2),L=d(M);r(L,{name:"encode_prompt",anchor:"diffusers.Lumina2Pipeline.encode_prompt",source:"https://github.com/huggingface/diffusers/blob/vr_14229/src/diffusers/pipelines/lumina2/pipeline_lumina2.py#L238",parameters:[{name:"prompt",val:": str | list[str]"},{name:"do_classifier_free_guidance",val:": bool = True"},{name:"negative_prompt",val:": str | list[str] = None"},{name:"num_images_per_prompt",val:": int = 1"},{name:"device",val:": typing.Optional[torch.device] = None"},{name:"prompt_embeds",val:": typing.Optional[torch.Tensor] = None"},{name:"negative_prompt_embeds",val:": typing.Optional[torch.Tensor] = None"},{name:"prompt_attention_mask",val:": typing.Optional[torch.Tensor] = None"},{name:"negative_prompt_attention_mask",val:": typing.Optional[torch.Tensor] = None"},{name:"system_prompt",val:": str | None = None"},{name:"max_sequence_length",val:": int = 256"}],parametersDescription:[{anchor:"diffusers.Lumina2Pipeline.encode_prompt.prompt",description:`<strong>prompt</strong> (<code>str</code> or <code>list[str]</code>, <em>optional</em>) &#x2014;
prompt to be encoded`,name:"prompt"},{anchor:"diffusers.Lumina2Pipeline.encode_prompt.negative_prompt",description:`<strong>negative_prompt</strong> (<code>str</code> or <code>list[str]</code>, <em>optional</em>) &#x2014;
The prompt not to guide the image generation. If not defined, one has to pass <code>negative_prompt_embeds</code>
instead. Ignored when not using guidance (i.e., ignored if <code>guidance_scale</code> is less than <code>1</code>). For
Lumina-T2I, this should be &quot;&quot;.`,name:"negative_prompt"},{anchor:"diffusers.Lumina2Pipeline.encode_prompt.do_classifier_free_guidance",description:`<strong>do_classifier_free_guidance</strong> (<code>bool</code>, <em>optional</em>, defaults to <code>True</code>) &#x2014;
whether to use classifier free guidance or not`,name:"do_classifier_free_guidance"},{anchor:"diffusers.Lumina2Pipeline.encode_prompt.num_images_per_prompt",description:`<strong>num_images_per_prompt</strong> (<code>int</code>, <em>optional</em>, defaults to 1) &#x2014;
number of images that should be generated per prompt`,name:"num_images_per_prompt"},{anchor:"diffusers.Lumina2Pipeline.encode_prompt.device",description:`<strong>device</strong> &#x2014; (<code>torch.device</code>, <em>optional</em>):
torch device to place the resulting embeddings on`,name:"device"},{anchor:"diffusers.Lumina2Pipeline.encode_prompt.prompt_embeds",description:`<strong>prompt_embeds</strong> (<code>torch.Tensor</code>, <em>optional</em>) &#x2014;
Pre-generated text embeddings. Can be used to easily tweak text inputs, <em>e.g.</em> prompt weighting. If not
provided, text embeddings will be generated from <code>prompt</code> input argument.`,name:"prompt_embeds"},{anchor:"diffusers.Lumina2Pipeline.encode_prompt.negative_prompt_embeds",description:`<strong>negative_prompt_embeds</strong> (<code>torch.Tensor</code>, <em>optional</em>) &#x2014;
Pre-generated negative text embeddings. For Lumina-T2I, it&#x2019;s should be the embeddings of the &quot;&quot; string.`,name:"negative_prompt_embeds"},{anchor:"diffusers.Lumina2Pipeline.encode_prompt.max_sequence_length",description:`<strong>max_sequence_length</strong> (<code>int</code>, defaults to <code>256</code>) &#x2014;
Maximum sequence length to use for the prompt.`,name:"max_sequence_length"}]}),J(2),p(M),p(o);var I=e(o,2);j(I,{source:"https://github.com/huggingface/diffusers/blob/main/docs/source/en/api/pipelines/lumina2.md"}),J(2),l(x,m),N()}export{X as component};