Buckets:

hf-doc-build
/

doc

Files
xet
hf-doc-build/doc / diffusers /main /id /quantization /modelopt.html
mishig
3 days ago
download
raw
26.1 kB
<meta charset="utf-8" /><meta name="hf:doc:metadata" content="{&quot;title&quot;:&quot;NVIDIA ModelOpt&quot;,&quot;local&quot;:&quot;nvidia-modelopt&quot;,&quot;sections&quot;:[{&quot;title&quot;:&quot;NVIDIAModelOptConfig&quot;,&quot;local&quot;:&quot;nvidiamodeloptconfig&quot;,&quot;sections&quot;:[],&quot;depth&quot;:2},{&quot;title&quot;:&quot;Jenis kuantisasi yang didukung&quot;,&quot;local&quot;:&quot;supported-quantization-types&quot;,&quot;sections&quot;:[],&quot;depth&quot;:2},{&quot;title&quot;:&quot;Menserialisasi dan Mendeserialisasi model terkuantisasi&quot;,&quot;local&quot;:&quot;serializing-and-deserializing-quantized-models&quot;,&quot;sections&quot;:[],&quot;depth&quot;:2}],&quot;depth&quot;:1}"/>
		<link href="/docs/diffusers/main/en/_app/immutable/entry/start.0S0_iEoG.js" rel="modulepreload">
		<link href="/docs/diffusers/main/en/_app/immutable/chunks/I9eooN77.js" rel="modulepreload">
		<link href="/docs/diffusers/main/en/_app/immutable/chunks/Bupkjr9f.js" rel="modulepreload">
		<link href="/docs/diffusers/main/en/_app/immutable/entry/app.CPdOxCbw.js" rel="modulepreload">
		<link href="/docs/diffusers/main/en/_app/immutable/chunks/CyvZ3FS5.js" rel="modulepreload">
		<link href="/docs/diffusers/main/en/_app/immutable/chunks/l0QnnOED.js" rel="modulepreload">
		<link href="/docs/diffusers/main/en/_app/immutable/chunks/DnVIJfC5.js" rel="modulepreload">
		<link href="/docs/diffusers/main/en/_app/immutable/chunks/DsnmJJEf.js" rel="modulepreload">
		<link href="/docs/diffusers/main/en/_app/immutable/nodes/0.DteTvmQZ.js" rel="modulepreload">
		<link href="/docs/diffusers/main/en/_app/immutable/chunks/CoSrdQax.js" rel="modulepreload">
		<link href="/docs/diffusers/main/en/_app/immutable/nodes/354.Da0b8rJZ.js" rel="modulepreload">
		<link href="/docs/diffusers/main/en/_app/immutable/chunks/DSPDl1gz.js" rel="modulepreload">
		<!--v66wdz--><meta name="hf:doc:metadata" content="{&quot;title&quot;:&quot;NVIDIA ModelOpt&quot;,&quot;local&quot;:&quot;nvidia-modelopt&quot;,&quot;sections&quot;:[{&quot;title&quot;:&quot;NVIDIAModelOptConfig&quot;,&quot;local&quot;:&quot;nvidiamodeloptconfig&quot;,&quot;sections&quot;:[],&quot;depth&quot;:2},{&quot;title&quot;:&quot;Supported quantization types&quot;,&quot;local&quot;:&quot;supported-quantization-types&quot;,&quot;sections&quot;:[],&quot;depth&quot;:2},{&quot;title&quot;:&quot;Serializing and Deserializing quantized models&quot;,&quot;local&quot;:&quot;serializing-and-deserializing-quantized-models&quot;,&quot;sections&quot;:[],&quot;depth&quot;:2}],&quot;depth&quot;:1}"/><!---->
		<link href="/docs/diffusers/main/en/_app/immutable/assets/0.tn0RQdqM.css" rel="modulepreload"> <!--[--><!--[0--><!--[--><!--[0--><!--[--><p></p> <div class="items-center shrink-0 min-w-[100px] max-sm:min-w-[50px] justify-end ml-auto flex" style="float: inline-end; margin-inline-start: 10px; display: inline-flex; position: relative; z-index: 10;"><div dir="ltr" class="inline-flex rounded-md max-sm:rounded-sm"><button class="inline-flex items-center gap-1 h-7 max-sm:h-7 px-2 max-sm:px-1.5 text-sm font-medium text-gray-800 border border-r-0 rounded-l-md max-sm:rounded-l-sm border-gray-200 bg-white hover:shadow-inner dark:border-gray-850 dark:bg-gray-950 dark:text-gray-200 dark:hover:bg-gray-800" aria-live="polite"><span class="inline-flex items-center justify-center rounded-md p-0.5 max-sm:p-0 hover:text-gray-800 dark:hover:text-gray-200"><svg class="sm:size-3.5 size-3" xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg><!----></span> <span dir="auto">Copy page</span></button> <button class="inline-flex items-center justify-center w-6 max-sm:w-5 h-7 max-sm:h-7 disabled:pointer-events-none text-sm text-gray-500 hover:text-gray-700 dark:hover:text-white rounded-r-md max-sm:rounded-r-sm border border-l transition border-gray-200 bg-white hover:shadow-inner dark:border-gray-850 dark:bg-gray-950 dark:text-gray-200 dark:hover:bg-gray-800" aria-haspopup="menu" aria-expanded="false" aria-label="Open copy menu"><svg class="transition-transform text-gray-400 overflow-visible sm:size-3.5 size-3 rotate-0" width="1em" height="1em" viewBox="0 0 12 7" fill="none" xmlns="http://www.w3.org/2000/svg"><path d="M1 1L6 6L11 1" stroke="currentColor"></path></svg><!----></button></div> <!--[-1--><!--]--></div><!----> <!--[0--><h1 class="relative group"><a id="nvidia-modelopt" class="header-link block pr-1.5 text-lg no-hover:hidden with-hover:absolute with-hover:p-1.5 with-hover:opacity-0 with-hover:group-hover:opacity-100 with-hover:right-full" href="#nvidia-modelopt"><span><svg xmlns="http://www.w3.org/2000/svg" xmlns:xlink="http://www.w3.org/1999/xlink" aria-hidden="true" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 256 256"><path d="M167.594 88.393a8.001 8.001 0 0 1 0 11.314l-67.882 67.882a8 8 0 1 1-11.314-11.315l67.882-67.881a8.003 8.003 0 0 1 11.314 0zm-28.287 84.86l-28.284 28.284a40 40 0 0 1-56.567-56.567l28.284-28.284a8 8 0 0 0-11.315-11.315l-28.284 28.284a56 56 0 0 0 79.196 79.197l28.285-28.285a8 8 0 1 0-11.315-11.314zM212.852 43.14a56.002 56.002 0 0 0-79.196 0l-28.284 28.284a8 8 0 1 0 11.314 11.314l28.284-28.284a40 40 0 0 1 56.568 56.567l-28.285 28.285a8 8 0 0 0 11.315 11.314l28.284-28.284a56.065 56.065 0 0 0 0-79.196z" fill="currentColor"></path></svg><!----></span></a> <span><!---->NVIDIA ModelOpt<!----></span></h1><!--]--><!----> <p><a href="https://github.com/NVIDIA/Model-Optimizer" rel="nofollow">NVIDIA-ModelOpt</a> adalah pustaka terpadu yang berisi teknik optimasi model terkini seperti kuantisasi, pemangkasan, distilasi, dekode spekulatif, dan lain-lain. Pustaka ini mengompres model pembelajaran mendalam untuk kerangka kerja penerapan hilir seperti TensorRT-LLM atau TensorRT guna mengoptimalkan kecepatan inferensi.</p> <p>Sebelum Anda memulai, pastikan Anda telah menginstal nvidia_modelopt.</p> <div class="code-block relative "><div class="absolute top-2.5 right-4"><button class="inline-flex items-center relative text-sm focus:text-green-500 cursor-pointer focus:outline-none transition duration-200 ease-in-out opacity-0 mx-0.5   text-gray-600  " title="code excerpt" type="button"><svg xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg><!----> <div class=" absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0 "><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent;"></div> Disalin</div><!----></button><!----></div> <pre class="language-bash "><!---->pip install -U <span class="hljs-string">&quot;nvidia_modelopt[hf]&quot;</span><!----></pre></div><!----> <p>Kuantisasi model dengan meneruskan <code>NVIDIAModelOptConfig</code> ke <a href="/docs/diffusers/main/id/api/models/overview#diffusers.ModelMixin.from_pretrained">from_pretrained()</a> (Anda juga dapat memuat model yang sudah dikuantisasi). Ini berlaku untuk model apa pun dalam modalitas apa pun, selama model tersebut mendukung pemuatan dengan <a href="https://hf.co/docs/accelerate/index" rel="nofollow">Accelerate</a> dan berisi lapisan <code>torch.nn.Linear</code>.</p> <p>Contoh di bawah ini hanya mengkuantisasi bobot ke FP8.</p> <div class="code-block relative "><div class="absolute top-2.5 right-4"><button class="inline-flex items-center relative text-sm focus:text-green-500 cursor-pointer focus:outline-none transition duration-200 ease-in-out opacity-0 mx-0.5   text-gray-600  " title="code excerpt" type="button"><svg xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg><!----> <div class=" absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0 "><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent;"></div> Disalin</div><!----></button><!----></div> <pre class="language-python "><!----><span class="hljs-keyword">import</span> torch
<span class="hljs-keyword">from</span> diffusers <span class="hljs-keyword">import</span> AutoModel, SanaPipeline, NVIDIAModelOptConfig

model_id = <span class="hljs-string">&quot;Efficient-Large-Model/Sana_600M_1024px_diffusers&quot;</span>
dtype = torch.bfloat16

quantization_config = NVIDIAModelOptConfig(quant_type=<span class="hljs-string">&quot;FP8&quot;</span>, quant_method=<span class="hljs-string">&quot;modelopt&quot;</span>)
transformer = AutoModel.from_pretrained(
    model_id,
    subfolder=<span class="hljs-string">&quot;transformer&quot;</span>,
    quantization_config=quantization_config,
    dtype=dtype,
)
pipe = SanaPipeline.from_pretrained(
    model_id,
    transformer=transformer,
    dtype=dtype,
)
pipe.to(<span class="hljs-string">&quot;cuda&quot;</span>)  <span class="hljs-comment"># or &quot;mps&quot;, &quot;xpu&quot;, &quot;cpu&quot;</span>

<span class="hljs-built_in">print</span>(<span class="hljs-string">f&quot;Pipeline memory usage: <span class="hljs-subst">{torch.cuda.max_memory_reserved() / <span class="hljs-number">1024</span>**<span class="hljs-number">3</span>:<span class="hljs-number">.3</span>f}</span> GB&quot;</span>)

prompt = <span class="hljs-string">&quot;A cat holding a sign that says hello world&quot;</span>
image = pipe(
    prompt, num_inference_steps=<span class="hljs-number">50</span>, guidance_scale=<span class="hljs-number">4.5</span>, max_sequence_length=<span class="hljs-number">512</span>
).images[<span class="hljs-number">0</span>]
image.save(<span class="hljs-string">&quot;output.png&quot;</span>)<!----></pre></div><!----> <blockquote><p><strong>Catatan:</strong></p> <p>Metode kuantisasi di NVIDIA-ModelOpt dirancang untuk mengurangi jejak memori bobot model menggunakan berbagai teknik QAT (Quantization-Aware Training) dan PTQ (Post-Training Quantization) sambil mempertahankan kinerja model. Namun, peningkatan kinerja aktual selama inferensi bergantung pada framework deployment (misalnya, TRT-LLM, TensorRT) dan konfigurasi hardware yang spesifik.</p> <p>Detail lebih lanjut dapat ditemukan <a href="https://github.com/NVIDIA/Model-Optimizer/tree/main/examples" rel="nofollow">di sini</a>.</p></blockquote> <!--[1--><h2 class="relative group"><a id="nvidiamodeloptconfig" class="header-link block pr-1.5 text-lg no-hover:hidden with-hover:absolute with-hover:p-1.5 with-hover:opacity-0 with-hover:group-hover:opacity-100 with-hover:right-full" href="#nvidiamodeloptconfig"><span><svg xmlns="http://www.w3.org/2000/svg" xmlns:xlink="http://www.w3.org/1999/xlink" aria-hidden="true" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 256 256"><path d="M167.594 88.393a8.001 8.001 0 0 1 0 11.314l-67.882 67.882a8 8 0 1 1-11.314-11.315l67.882-67.881a8.003 8.003 0 0 1 11.314 0zm-28.287 84.86l-28.284 28.284a40 40 0 0 1-56.567-56.567l28.284-28.284a8 8 0 0 0-11.315-11.315l-28.284 28.284a56 56 0 0 0 79.196 79.197l28.285-28.285a8 8 0 1 0-11.315-11.314zM212.852 43.14a56.002 56.002 0 0 0-79.196 0l-28.284 28.284a8 8 0 1 0 11.314 11.314l28.284-28.284a40 40 0 0 1 56.568 56.567l-28.285 28.285a8 8 0 0 0 11.315 11.314l28.284-28.284a56.065 56.065 0 0 0 0-79.196z" fill="currentColor"></path></svg><!----></span></a> <span><!---->NVIDIAModelOptConfig<!----></span></h2><!--]--><!----> <p>Kelas <code>NVIDIAModelOptConfig</code> menerima tiga parameter:</p> <ul><li><code>quant_type</code>: Nilai string yang menyebutkan salah satu tipe kuantisasi di bawah ini.</li> <li><code>modules_to_not_convert</code>: Daftar nama modul penuh/sebagian di mana kuantisasi tidak boleh dilakukan. Misalnya, untuk tidak melakukan kuantisasi pada blok proyeksi pos_embed dari <a href="/docs/diffusers/main/id/api/models/sd3_transformer2d#diffusers.SD3Transformer2DModel">SD3Transformer2DModel</a>, Anda dapat menentukan: <code>modules_to_not_convert=["pos_embed.proj.weight"]</code>.</li> <li><code>disable_conv_quantization</code>: Nilai boolean yang jika diatur ke <code>True</code> akan menonaktifkan kuantisasi untuk semua lapisan konvolusional dalam model. Ini berguna karena kuantisasi channel dan block umumnya tidak bekerja dengan baik pada lapisan konvolusional (dipakai dengan INT4, NF4, NVFP4). Jika Anda ingin menonaktifkan kuantisasi untuk lapisan konvolusional tertentu, gunakan <code>modules_to_not_convert</code> sebagai gantinya.</li> <li><code>algorithm</code>: Algoritma yang digunakan untuk menentukan skala, secara default <code>"max"</code>. Anda dapat memeriksa dokumentasi modelopt untuk algoritma dan detail lainnya.</li> <li><code>forward_loop</code>: Fungsi loop forward yang digunakan untuk mengkalibrasi aktivasi selama kuantisasi. Jika tidak disediakan, ia mengandalkan nilai skala statis yang dihitung menggunakan hanya bobot.</li> <li><code>kwargs</code>: Dict berisi argumen kata kunci untuk diteruskan ke metode kuantisasi dasar yang akan dipanggil berdasarkan <code>quant_type</code>.</li></ul> <!--[1--><h2 class="relative group"><a id="supported-quantization-types" class="header-link block pr-1.5 text-lg no-hover:hidden with-hover:absolute with-hover:p-1.5 with-hover:opacity-0 with-hover:group-hover:opacity-100 with-hover:right-full" href="#supported-quantization-types"><span><svg xmlns="http://www.w3.org/2000/svg" xmlns:xlink="http://www.w3.org/1999/xlink" aria-hidden="true" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 256 256"><path d="M167.594 88.393a8.001 8.001 0 0 1 0 11.314l-67.882 67.882a8 8 0 1 1-11.314-11.315l67.882-67.881a8.003 8.003 0 0 1 11.314 0zm-28.287 84.86l-28.284 28.284a40 40 0 0 1-56.567-56.567l28.284-28.284a8 8 0 0 0-11.315-11.315l-28.284 28.284a56 56 0 0 0 79.196 79.197l28.285-28.285a8 8 0 1 0-11.315-11.314zM212.852 43.14a56.002 56.002 0 0 0-79.196 0l-28.284 28.284a8 8 0 1 0 11.314 11.314l28.284-28.284a40 40 0 0 1 56.568 56.567l-28.285 28.285a8 8 0 0 0 11.315 11.314l28.284-28.284a56.065 56.065 0 0 0 0-79.196z" fill="currentColor"></path></svg><!----></span></a> <span><!---->Jenis kuantisasi yang didukung<!----></span></h2><!--]--><!----> <p>ModelOpt mendukung kuantisasi int8, fp8, int4, nf4, dan nvfp4 untuk weight-only, channel, dan block. Metode kuantisasi ini dirancang untuk mengurangi jejak memori dari bobot model sambil mempertahankan kinerja model selama inferensi.</p> <p>Kuantisasi hanya-berat menyimpan bobot model dalam tipe data bit rendah tertentu tetapi melakukan komputasi dengan tipe data presisi yang lebih tinggi, seperti <code>bfloat16</code>. Hal ini menurunkan kebutuhan memori dari bobot model tetapi mempertahankan puncak memori untuk komputasi aktivasi.</p> <p>Metode kuantisasi yang didukung adalah sebagai berikut:</p> <table><thead><tr><th><strong>Jenis Kuantisasi</strong></th><th><strong>Skema yang Didukung</strong></th><th><strong>Kwargs yang Diperlukan</strong></th><th><strong>Catatan Tambahan</strong></th></tr></thead><tbody><tr><td><strong>INT8</strong></td><td><code>int8 weight only</code>, <code>int8 channel quantization</code>, <code>int8 block quantization</code></td><td><code>quant_type</code>, <code>quant_type + channel_quantize</code>, <code>quant_type + channel_quantize + block_quantize</code></td><td></td></tr><tr><td><strong>FP8</strong></td><td><code>fp8 weight only</code>, <code>fp8 channel quantization</code>, <code>fp8 block quantization</code></td><td><code>quant_type</code>, <code>quant_type + channel_quantize</code>, <code>quant_type + channel_quantize + block_quantize</code></td><td></td></tr><tr><td><strong>INT4</strong></td><td><code>int4 weight only</code>, <code>int4 block quantization</code></td><td><code>quant_type</code>, <code>quant_type + channel_quantize + block_quantize</code></td><td><code>channel_quantize = -1 is only supported for now</code></td></tr><tr><td><strong>NF4</strong></td><td><code>nf4 weight only</code>, <code>nf4 double block quantization</code></td><td><code>quant_type</code>, <code>quant_type + channel_quantize + block_quantize + scale_channel_quantize</code> + <code>scale_block_quantize</code></td><td><code>channel_quantize = -1 and scale_channel_quantize = -1 are only supported for now</code></td></tr><tr><td><strong>NVFP4</strong></td><td><code>nvfp4 weight only</code>, <code>nvfp4 block quantization</code></td><td><code>quant_type</code>, <code>quant_type + channel_quantize + block_quantize</code></td><td><code>channel_quantize = -1 is only supported for now</code></td></tr></tbody></table> <p>Lihat <a href="https://nvidia.github.io/Model-Optimizer/" rel="nofollow">dokumentasi modelopt resmi</a> untuk pemahaman yang lebih baik tentang metode kuantisasi yang tersedia dan daftar lengkap opsi konfigurasi yang tersedia.</p> <!--[1--><h2 class="relative group"><a id="serializing-and-deserializing-quantized-models" class="header-link block pr-1.5 text-lg no-hover:hidden with-hover:absolute with-hover:p-1.5 with-hover:opacity-0 with-hover:group-hover:opacity-100 with-hover:right-full" href="#serializing-and-deserializing-quantized-models"><span><svg xmlns="http://www.w3.org/2000/svg" xmlns:xlink="http://www.w3.org/1999/xlink" aria-hidden="true" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 256 256"><path d="M167.594 88.393a8.001 8.001 0 0 1 0 11.314l-67.882 67.882a8 8 0 1 1-11.314-11.315l67.882-67.881a8.003 8.003 0 0 1 11.314 0zm-28.287 84.86l-28.284 28.284a40 40 0 0 1-56.567-56.567l28.284-28.284a8 8 0 0 0-11.315-11.315l-28.284 28.284a56 56 0 0 0 79.196 79.197l28.285-28.285a8 8 0 1 0-11.315-11.314zM212.852 43.14a56.002 56.002 0 0 0-79.196 0l-28.284 28.284a8 8 0 1 0 11.314 11.314l28.284-28.284a40 40 0 0 1 56.568 56.567l-28.285 28.285a8 8 0 0 0 11.315 11.314l28.284-28.284a56.065 56.065 0 0 0 0-79.196z" fill="currentColor"></path></svg><!----></span></a> <span><!---->Menserialisasi dan Mendeserialisasi model terkuantisasi<!----></span></h2><!--]--><!----> <p>Untuk mengserialisasi model terkuantisasi dalam dtype tertentu, pertama muat model dengan dtype kuantisasi yang diinginkan, lalu simpan menggunakan metode <a href="/docs/diffusers/main/id/api/models/overview#diffusers.ModelMixin.save_pretrained">save_pretrained()</a>.</p> <div class="code-block relative "><div class="absolute top-2.5 right-4"><button class="inline-flex items-center relative text-sm focus:text-green-500 cursor-pointer focus:outline-none transition duration-200 ease-in-out opacity-0 mx-0.5   text-gray-600  " title="code excerpt" type="button"><svg xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg><!----> <div class=" absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0 "><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent;"></div> Disalin</div><!----></button><!----></div> <pre class="language-python "><!----><span class="hljs-keyword">import</span> torch
<span class="hljs-keyword">from</span> diffusers <span class="hljs-keyword">import</span> AutoModel, NVIDIAModelOptConfig
<span class="hljs-keyword">from</span> modelopt.torch.opt <span class="hljs-keyword">import</span> enable_huggingface_checkpointing

enable_huggingface_checkpointing()

model_id = <span class="hljs-string">&quot;Efficient-Large-Model/Sana_600M_1024px_diffusers&quot;</span>
quant_config_fp8 = {<span class="hljs-string">&quot;quant_type&quot;</span>: <span class="hljs-string">&quot;FP8&quot;</span>, <span class="hljs-string">&quot;quant_method&quot;</span>: <span class="hljs-string">&quot;modelopt&quot;</span>}
quant_config_fp8 = NVIDIAModelOptConfig(**quant_config_fp8)
model = AutoModel.from_pretrained(
    model_id,
    subfolder=<span class="hljs-string">&quot;transformer&quot;</span>,
    quantization_config=quant_config_fp8,
    dtype=torch.bfloat16,
)
model.save_pretrained(<span class="hljs-string">&#x27;path/to/sana_fp8&#x27;</span>, safe_serialization=<span class="hljs-literal">False</span>)<!----></pre></div><!----> <p>Untuk memuat model kuantisasi yang telah diserialisasi, gunakan metode <a href="/docs/diffusers/main/id/api/models/overview#diffusers.ModelMixin.from_pretrained">from_pretrained()</a>.</p> <div class="code-block relative "><div class="absolute top-2.5 right-4"><button class="inline-flex items-center relative text-sm focus:text-green-500 cursor-pointer focus:outline-none transition duration-200 ease-in-out opacity-0 mx-0.5   text-gray-600  " title="code excerpt" type="button"><svg xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg><!----> <div class=" absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0 "><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent;"></div> Disalin</div><!----></button><!----></div> <pre class="language-python "><!----><span class="hljs-keyword">import</span> torch
<span class="hljs-keyword">from</span> diffusers <span class="hljs-keyword">import</span> AutoModel, NVIDIAModelOptConfig, SanaPipeline
<span class="hljs-keyword">from</span> modelopt.torch.opt <span class="hljs-keyword">import</span> enable_huggingface_checkpointing

enable_huggingface_checkpointing()

quantization_config = NVIDIAModelOptConfig(quant_type=<span class="hljs-string">&quot;FP8&quot;</span>, quant_method=<span class="hljs-string">&quot;modelopt&quot;</span>)
transformer = AutoModel.from_pretrained(
    <span class="hljs-string">&quot;path/to/sana_fp8&quot;</span>,
    subfolder=<span class="hljs-string">&quot;transformer&quot;</span>,
    quantization_config=quantization_config,
    dtype=torch.bfloat16,
)
pipe = SanaPipeline.from_pretrained(
    <span class="hljs-string">&quot;Efficient-Large-Model/Sana_600M_1024px_diffusers&quot;</span>,
    transformer=transformer,
    dtype=torch.bfloat16,
)
pipe.to(<span class="hljs-string">&quot;cuda&quot;</span>)  <span class="hljs-comment"># or &quot;mps&quot;, &quot;xpu&quot;, &quot;cpu&quot;</span>
prompt = <span class="hljs-string">&quot;A cat holding a sign that says hello world&quot;</span>
image = pipe(
    prompt, num_inference_steps=<span class="hljs-number">50</span>, guidance_scale=<span class="hljs-number">4.5</span>, max_sequence_length=<span class="hljs-number">512</span>
).images[<span class="hljs-number">0</span>]
image.save(<span class="hljs-string">&quot;output.png&quot;</span>)<!----></pre></div><!----> <a class="!text-gray-400 !no-underline text-sm flex items-center not-prose mt-4" href="https://github.com/huggingface/diffusers/blob/main/docs/source/en/quantization/modelopt.md" target="_blank"><svg class="mr-1" xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M31,16l-7,7l-1.41-1.41L28.17,16l-5.58-5.59L24,9l7,7z"></path><path d="M1,16l7-7l1.41,1.41L3.83,16l5.58,5.59L8,23l-7-7z"></path><path d="M12.419,25.484L17.639,6.552l1.932,0.518L14.351,26.002z"></path></svg><!----> <span><span class="underline">Update</span> on GitHub</span></a><!----> <p></p><!--]--><!----><!--]--><!--]--><!--]--> <!--[-1--><!--]--><!--]-->
			
			<script>
				{
					__sveltekit_1qsdbwf = {
						base: "/docs/diffusers/main/en",
						assets: "/docs/diffusers/main/en"
					};
					const element = document.currentScript.parentElement;
					Promise.all([
						import("/docs/diffusers/main/en/_app/immutable/entry/start.0S0_iEoG.js"),
						import("/docs/diffusers/main/en/_app/immutable/entry/app.CPdOxCbw.js")
					]).then(([kit, app]) => {
						kit.start(app, element, {
							node_ids: [0, 354],
							data: [null,null],
							form: null,
							error: null
						});
					});
				}
			</script>
		
Xet Storage Details

Size:: 26.1 kB
Xet hash:: 79ff86fae2a71481fdc1bf895e87ee418c64cc950d7ef87126b58f0ebd55b1e4
Xet efficiently stores files, intelligently splitting them into unique chunks and accelerating uploads and downloads. More info.