diffusers/main/pt/_app/immutable/nodes/5.CNxIJabh.js

import"../chunks/DsnmJJEf.js";import{i as I,h as v,C as E,H as e,a as l,E as X,s as x}from"../chunks/BZtIR7es.js";import{p as _,o as k,s as a,f as q,a as j,b as R,d as i,c as T,n as w,r as t}from"../chunks/_hyVs69O.js";import{D as V}from"../chunks/DbcpnoaO.js";const N='{"title":"Desempenho básico","local":"desempenho-básico","sections":[{"title":"Uso de memória","local":"uso-de-memória","sections":[],"depth":2},{"title":"Velocidade de inferência","local":"velocidade-de-inferência","sections":[],"depth":2},{"title":"Qualidade de geração","local":"qualidade-de-geração","sections":[],"depth":2},{"title":"Próximos passos","local":"próximos-passos","sections":[],"depth":2}],"depth":1}';var S=T('<meta name="hf:doc:metadata"/>'),Y=T('<p></p> <!> <!> <!> <p>Difusão é um processo aleatório que demanda muito processamento. Você pode precisar executar o <code>DiffusionPipeline</code> várias vezes antes de obter o resultado desejado. Por isso é importante equilibrar cuidadosamente a velocidade de geração e o uso de memória para iterar mais rápido.</p> <p>Este guia recomenda algumas dicas básicas de desempenho para usar o <code>DiffusionPipeline</code>. Consulte a seção de documentação sobre Otimização de Inferência, como <a href="./optimization/fp16">Acelerar inferência</a> ou <a href="./optimization/memory">Reduzir uso de memória</a> para guias de desempenho mais detalhados.</p> <!> <p>Reduzir a quantidade de memória usada indiretamente acelera a geração e pode ajudar um modelo a caber no dispositivo.</p> <p>O método <code>enable_model_cpu_offload()</code> move um modelo para a CPU quando não está em uso para economizar memória da GPU.</p> <!> <!> <p>O processo de remoção de ruído é o mais exigente computacionalmente durante a difusão. Métodos que otimizam este processo aceleram a velocidade de inferência. Experimente os seguintes métodos para acelerar.</p> <ul><li>Adicione <code>device_map="cuda"</code> para colocar o pipeline em uma GPU. Colocar um modelo em um acelerador, como uma GPU, aumenta a velocidade porque realiza computações em paralelo.</li> <li>Defina <code>torch_dtype=torch.bfloat16</code> para executar o pipeline em meia-precisão. Reduzir a precisão do tipo de dado aumenta a velocidade porque leva menos tempo para realizar computações em precisão mais baixa.</li></ul> <!> <ul><li>Use um agendador mais rápido, como <code>DPMSolverMultistepScheduler</code>, que requer apenas ~20-25 passos.</li> <li>Defina <code>num_inference_steps</code> para um valor menor. Reduzir o número de passos de inferência reduz o número total de computações. No entanto, isso pode resultar em menor qualidade de geração.</li></ul> <!> <!> <p>Muitos modelos de difusão modernos entregam imagens de alta qualidade imediatamente. No entanto, você ainda pode melhorar a qualidade de geração experimentando o seguinte.</p> <ul><li><p>Experimente um prompt mais detalhado e descritivo. Inclua detalhes como o meio da imagem, assunto, estilo e estética. Um prompt negativo também pode ajudar, guiando um modelo para longe de características indesejáveis usando palavras como baixa qualidade ou desfocado.</p> <!> <p>Para mais detalhes sobre como criar prompts melhores, consulte a documentação sobre <a href="./using-diffusers/weighted_prompts">Técnicas de prompt</a>.</p></li> <li><p>Experimente um agendador diferente, como <code>HeunDiscreteScheduler</code> ou <code>LMSDiscreteScheduler</code>, que sacrifica velocidade de geração por qualidade.</p> <!></li></ul> <!> <p>Diffusers oferece otimizações mais avançadas e poderosas, como <a href="./optimization/memory#group-offloading">group-offloading</a> e <a href="./optimization/fp16#regional-compilation">compilação regional</a>. Para saber mais sobre como maximizar o desempenho, consulte a seção sobre Otimização de Inferência.</p> <!> <p></p>',1);function F(Z,B){_(B,!1),k(()=>{new URLSearchParams(window.location.search).get("fw")}),I();var n=Y();v("1ji58vo",f=>{var g=S();x(g,"content",N),j(f,g)});var p=a(q(n),2);E(p,{containerStyle:"float: right; margin-left: 10px; display: inline-flex; position: relative; z-index: 10;"});var d=a(p,2);V(d,{containerStyle:"float: right; margin-left: 10px; display: inline-flex; position: relative; z-index: 10;",options:[{label:"Mixed",value:"https://colab.research.google.com/github/huggingface/notebooks/blob/main/diffusers_doc/pt/stable_diffusion.ipynb"},{label:"PyTorch",value:"https://colab.research.google.com/github/huggingface/notebooks/blob/main/diffusers_doc/pt/pytorch/stable_diffusion.ipynb"},{label:"TensorFlow",value:"https://colab.research.google.com/github/huggingface/notebooks/blob/main/diffusers_doc/pt/tensorflow/stable_diffusion.ipynb"},{label:"Mixed",value:"https://studiolab.sagemaker.aws/import/github/huggingface/notebooks/blob/main/diffusers_doc/pt/stable_diffusion.ipynb"},{label:"PyTorch",value:"https://studiolab.sagemaker.aws/import/github/huggingface/notebooks/blob/main/diffusers_doc/pt/pytorch/stable_diffusion.ipynb"},{label:"TensorFlow",value:"https://studiolab.sagemaker.aws/import/github/huggingface/notebooks/blob/main/diffusers_doc/pt/tensorflow/stable_diffusion.ipynb"}]});var r=a(d,2);e(r,{title:"Desempenho básico",local:"desempenho-básico",headingTag:"h1"});var m=a(r,6);e(m,{title:"Uso de memória",local:"uso-de-memória",headingTag:"h2"});var c=a(m,6);l(c,{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> DiffusionPipeline

pipeline = DiffusionPipeline.from_pretrained(
  <span class="hljs-string">&quot;stabilityai/stable-diffusion-xl-base-1.0&quot;</span>,
  torch_dtype=torch.bfloat16,
  device_map=<span class="hljs-string">&quot;cuda&quot;</span>
)
pipeline.enable_model_cpu_offload()

prompt = <span class="hljs-string">&quot;&quot;&quot;
cinematic film still of a cat sipping a margarita in a pool in Palm Springs, California
highly detailed, high budget hollywood movie, cinemascope, moody, epic, gorgeous, film grain
&quot;&quot;&quot;</span>
pipeline(prompt).images[<span class="hljs-number">0</span>]
<span class="hljs-built_in">print</span>(<span class="hljs-string">f&quot;Memória máxima reservada: <span class="hljs-subst">{torch.cuda.max_memory_allocated() / <span class="hljs-number">1024</span>**<span class="hljs-number">3</span>:<span class="hljs-number">.2</span>f}</span> GB&quot;</span>)`,lang:"py",wrap:!1});var u=a(c,2);e(u,{title:"Velocidade de inferência",local:"velocidade-de-inferência",headingTag:"h2"});var M=a(u,6);l(M,{code:"aW1wb3J0JTIwdG9yY2glMEFpbXBvcnQlMjB0aW1lJTBBZnJvbSUyMGRpZmZ1c2VycyUyMGltcG9ydCUyMERpZmZ1c2lvblBpcGVsaW5lJTJDJTIwRFBNU29sdmVyTXVsdGlzdGVwU2NoZWR1bGVyJTBBJTBBcGlwZWxpbmUlMjAlM0QlMjBEaWZmdXNpb25QaXBlbGluZS5mcm9tX3ByZXRyYWluZWQoJTBBJTIwJTIwJTIyc3RhYmlsaXR5YWklMkZzdGFibGUtZGlmZnVzaW9uLXhsLWJhc2UtMS4wJTIyJTJDJTBBJTIwJTIwdG9yY2hfZHR5cGUlM0R0b3JjaC5iZmxvYXQxNiUyQyUwQSUyMCUyMGRldmljZV9tYXAlM0QlMjJjdWRhJTIyJTBBKQ==",highlighted:`<span class="hljs-keyword">import</span> torch
<span class="hljs-keyword">import</span> time
<span class="hljs-keyword">from</span> diffusers <span class="hljs-keyword">import</span> DiffusionPipeline, DPMSolverMultistepScheduler

pipeline = DiffusionPipeline.from_pretrained(
  <span class="hljs-string">&quot;stabilityai/stable-diffusion-xl-base-1.0&quot;</span>,
  torch_dtype=torch.bfloat16,
  device_map=<span class="hljs-string">&quot;cuda&quot;</span>
)`,lang:"py",wrap:!1});var b=a(M,4);l(b,{code:"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",highlighted:`pipeline.scheduler = DPMSolverMultistepScheduler.from_config(pipeline.scheduler.config)

prompt = <span class="hljs-string">&quot;&quot;&quot;
cinematic film still of a cat sipping a margarita in a pool in Palm Springs, California
highly detailed, high budget hollywood movie, cinemascope, moody, epic, gorgeous, film grain
&quot;&quot;&quot;</span>

start_time = time.perf_counter()
image = pipeline(prompt).images[<span class="hljs-number">0</span>]
end_time = time.perf_counter()

<span class="hljs-built_in">print</span>(<span class="hljs-string">f&quot;Geração de imagem levou <span class="hljs-subst">{end_time - start_time:<span class="hljs-number">.3</span>f}</span> segundos&quot;</span>)`,lang:"py",wrap:!1});var J=a(b,2);e(J,{title:"Qualidade de geração",local:"qualidade-de-geração",headingTag:"h2"});var o=a(J,4),s=i(o),U=a(i(s),2);l(U,{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> DiffusionPipeline

pipeline = DiffusionPipeline.from_pretrained(
    <span class="hljs-string">&quot;stabilityai/stable-diffusion-xl-base-1.0&quot;</span>,
    torch_dtype=torch.bfloat16,
    device_map=<span class="hljs-string">&quot;cuda&quot;</span>
)

prompt = <span class="hljs-string">&quot;&quot;&quot;
cinematic film still of a cat sipping a margarita in a pool in Palm Springs, California
highly detailed, high budget hollywood movie, cinemascope, moody, epic, gorgeous, film grain
&quot;&quot;&quot;</span>
negative_prompt = <span class="hljs-string">&quot;low quality, blurry, ugly, poor details&quot;</span>
pipeline(prompt, negative_prompt=negative_prompt).images[<span class="hljs-number">0</span>]`,lang:"py",wrap:!1}),w(2),t(s);var h=a(s,2),W=a(i(h),2);l(W,{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> DiffusionPipeline, HeunDiscreteScheduler

pipeline = DiffusionPipeline.from_pretrained(
    <span class="hljs-string">&quot;stabilityai/stable-diffusion-xl-base-1.0&quot;</span>,
    torch_dtype=torch.bfloat16,
    device_map=<span class="hljs-string">&quot;cuda&quot;</span>
)
pipeline.scheduler = HeunDiscreteScheduler.from_config(pipeline.scheduler.config)

prompt = <span class="hljs-string">&quot;&quot;&quot;
cinematic film still of a cat sipping a margarita in a pool in Palm Springs, California
highly detailed, high budget hollywood movie, cinemascope, moody, epic, gorgeous, film grain
&quot;&quot;&quot;</span>
negative_prompt = <span class="hljs-string">&quot;low quality, blurry, ugly, poor details&quot;</span>
pipeline(prompt, negative_prompt=negative_prompt).images[<span class="hljs-number">0</span>]`,lang:"py",wrap:!1}),t(h),t(o);var y=a(o,2);e(y,{title:"Próximos passos",local:"próximos-passos",headingTag:"h2"});var G=a(y,4);X(G,{source:"https://github.com/huggingface/diffusers/blob/main/docs/source/pt/stable_diffusion.md"}),w(2),j(Z,n),R()}export{F as component};