diffusers/pr_14180/en/_app/immutable/nodes/269.CdQY9OYQ.js

import"../chunks/DsnmJJEf.js";import{i as L,h as D,C as K,H as e,a as l,b as N,E as P,s as $}from"../chunks/CmJXCtRL.js";import{p as O,o as ss,s,f as p,a as n,b as ls,d as y,n as b}from"../chunks/DK803DsY.js";import{H as m}from"../chunks/BtTdhXOX.js";const es='{"title":"bitsandbytes","local":"bitsandbytes","sections":[{"title":"8-bit (LLM.int8() algorithm)","local":"8-bit-llmint8-algorithm","sections":[{"title":"Outlier threshold","local":"outlier-threshold","sections":[],"depth":3},{"title":"Skip module conversion","local":"skip-module-conversion","sections":[],"depth":3}],"depth":2},{"title":"4-bit (QLoRA algorithm)","local":"4-bit-qlora-algorithm","sections":[{"title":"Compute data type","local":"compute-data-type","sections":[],"depth":3},{"title":"Normal Float 4 (NF4)","local":"normal-float-4-nf4","sections":[],"depth":3},{"title":"Nested quantization","local":"nested-quantization","sections":[],"depth":3}],"depth":2},{"title":"Dequantizing bitsandbytes models","local":"dequantizing-bitsandbytes-models","sections":[],"depth":2},{"title":"torch.compile","local":"torchcompile","sections":[],"depth":2},{"title":"Resources","local":"resources","sections":[],"depth":2}],"depth":1}';var as=y('<meta name="hf:doc:metadata"/>'),ts=y(`<p>Quantizing a model in 8-bit halves the memory-usage:</p> <p>bitsandbytes is supported in both Transformers and Diffusers, so you can quantize both the <a href="/docs/diffusers/pr_14180/en/api/models/flux_transformer#diffusers.FluxTransformer2DModel">FluxTransformer2DModel</a> and <code>T5EncoderModel</code>.</p> <p>For Ada and higher-series GPUs. we recommend changing <code>torch_dtype</code> to <code>torch.bfloat16</code>.</p> <blockquote class="tip"><p>The <code>CLIPTextModel</code> and <a href="/docs/diffusers/pr_14180/en/api/models/autoencoderkl#diffusers.AutoencoderKL">AutoencoderKL</a> aren’t quantized because they’re already small in size and because <a href="/docs/diffusers/pr_14180/en/api/models/autoencoderkl#diffusers.AutoencoderKL">AutoencoderKL</a> only has a few <code>torch.nn.Linear</code> layers.</p></blockquote> <!> <p>By default, all the other modules such as <code>torch.nn.LayerNorm</code> are converted to <code>torch.float16</code>. You can change the data type of these modules with the <code>torch_dtype</code> parameter.</p> <!> <p>Let’s generate an image using our quantized models.</p> <p>Setting <code>device_map="auto"</code> automatically fills all available space on the GPU(s) first, then the
CPU, and finally, the hard drive (the absolute slowest option) if there is still not enough memory.</p> <!> <div class="flex justify-center"><img src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/diffusers/quant-bnb/8bit.png"/></div> <p>When there is enough memory, you can also directly move the pipeline to the GPU with <code>.to("cuda")</code> and apply <a href="/docs/diffusers/pr_14180/en/api/pipelines/overview#diffusers.DiffusionPipeline.enable_model_cpu_offload">enable_model_cpu_offload()</a> to optimize GPU memory usage.</p> <p>Once a model is quantized, you can push the model to the Hub with the <a href="/docs/diffusers/pr_14180/en/api/schedulers/overview#diffusers.utils.PushToHubMixin.push_to_hub">push_to_hub()</a> method. The quantization <code>config.json</code> file is pushed first, followed by the quantized model weights. You can also save the serialized 8-bit models locally with <a href="/docs/diffusers/pr_14180/en/api/models/overview#diffusers.ModelMixin.save_pretrained">save_pretrained()</a>.</p>`,1),ns=y('<p>Quantizing a model in 4-bit reduces your memory-usage by 4x:</p> <p>bitsandbytes is supported in both Transformers and Diffusers, so you can can quantize both the <a href="/docs/diffusers/pr_14180/en/api/models/flux_transformer#diffusers.FluxTransformer2DModel">FluxTransformer2DModel</a> and <code>T5EncoderModel</code>.</p> <p>For Ada and higher-series GPUs. we recommend changing <code>torch_dtype</code> to <code>torch.bfloat16</code>.</p> <blockquote class="tip"><p>The <code>CLIPTextModel</code> and <a href="/docs/diffusers/pr_14180/en/api/models/autoencoderkl#diffusers.AutoencoderKL">AutoencoderKL</a> aren’t quantized because they’re already small in size and because <a href="/docs/diffusers/pr_14180/en/api/models/autoencoderkl#diffusers.AutoencoderKL">AutoencoderKL</a> only has a few <code>torch.nn.Linear</code> layers.</p></blockquote> <!> <p>By default, all the other modules such as <code>torch.nn.LayerNorm</code> are converted to <code>torch.float16</code>. You can change the data type of these modules with the <code>torch_dtype</code> parameter.</p> <!> <p>Let’s generate an image using our quantized models.</p> <p>Setting <code>device_map="auto"</code> automatically fills all available space on the GPU(s) first, then the CPU, and finally, the hard drive (the absolute slowest option) if there is still not enough memory.</p> <!> <div class="flex justify-center"><img src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/diffusers/quant-bnb/4bit.png"/></div> <p>When there is enough memory, you can also directly move the pipeline to the GPU with <code>.to("cuda")</code> and apply <a href="/docs/diffusers/pr_14180/en/api/pipelines/overview#diffusers.DiffusionPipeline.enable_model_cpu_offload">enable_model_cpu_offload()</a> to optimize GPU memory usage.</p> <p>Once a model is quantized, you can push the model to the Hub with the <a href="/docs/diffusers/pr_14180/en/api/schedulers/overview#diffusers.utils.PushToHubMixin.push_to_hub">push_to_hub()</a> method. The quantization <code>config.json</code> file is pushed first, followed by the quantized model weights. You can also save the serialized 4-bit models locally with <a href="/docs/diffusers/pr_14180/en/api/models/overview#diffusers.ModelMixin.save_pretrained">save_pretrained()</a>.</p>',1),Y=y("<!> <!>",1),os=y(`<p></p> <!> <!> <p><a href="https://huggingface.co/docs/bitsandbytes/index" rel="nofollow">bitsandbytes</a> is the easiest option for quantizing a model to 8 and 4-bit. 8-bit quantization multiplies outliers in fp16 with non-outliers in int8, converts the non-outlier values back to fp16, and then adds them together to return the weights in fp16. This reduces the degradative effect outlier values have on a model’s performance.</p> <p>4-bit quantization compresses a model even further, and it is commonly used with <a href="https://hf.co/papers/2305.14314" rel="nofollow">QLoRA</a> to finetune quantized LLMs.</p> <p>This guide demonstrates how quantization can enable running <a href="https://huggingface.co/black-forest-labs/FLUX.1-dev" rel="nofollow">FLUX.1-dev</a> on less than 16GB of VRAM and even on a free Google
Colab instance.</p> <p><img src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/diffusers/quant-bnb/comparison.png" alt="comparison image"/></p> <p>To use bitsandbytes, make sure you have the following libraries installed:</p> <!> <p>Now you can quantize a model by passing a <a href="/docs/diffusers/pr_14180/en/api/quantization#diffusers.BitsAndBytesConfig">BitsAndBytesConfig</a> to <a href="/docs/diffusers/pr_14180/en/api/models/overview#diffusers.ModelMixin.from_pretrained">from_pretrained()</a>. This works for any model in any modality, as long as it supports loading with <a href="https://hf.co/docs/accelerate/index" rel="nofollow">Accelerate</a> and contains <code>torch.nn.Linear</code> layers.</p> <!> <blockquote class="warning"><p>Training with 8-bit and 4-bit weights are only supported for training <em>extra</em> parameters.</p></blockquote> <p>Check your memory footprint with the <code>get_memory_footprint</code> method:</p> <!> <p>Note that this only tells you the memory footprint of the model params and does <em>not</em> estimate the inference memory requirements.</p> <p>Quantized models can be loaded from the <a href="/docs/diffusers/pr_14180/en/api/models/overview#diffusers.ModelMixin.from_pretrained">from_pretrained()</a> method without needing to specify the <code>quantization_config</code> parameters:</p> <!> <!> <blockquote class="tip"><p>Learn more about the details of 8-bit quantization in this <a href="https://huggingface.co/blog/hf-bitsandbytes-integration" rel="nofollow">blog post</a>!</p></blockquote> <p>This section explores some of the specific features of 8-bit models, such as outlier thresholds and skipping module conversion.</p> <!> <p>An “outlier” is a hidden state value greater than a certain threshold, and these values are computed in fp16. While the values are usually normally distributed ([-3.5, 3.5]), this distribution can be very different for large models ([-60, 6] or [6, 60]). 8-bit quantization works well for values ~5, but beyond that, there is a significant performance penalty. A good default threshold value is 6, but a lower threshold may be needed for more unstable models (small models or finetuning).</p> <p>To find the best threshold for your model, we recommend experimenting with the <code>llm_int8_threshold</code> parameter in <a href="/docs/diffusers/pr_14180/en/api/quantization#diffusers.BitsAndBytesConfig">BitsAndBytesConfig</a>:</p> <!> <!> <p>For some models, you don’t need to quantize every module to 8-bit which can actually cause instability. For example, for diffusion models like <a href="../api/pipelines/stable_diffusion/stable_diffusion_3">Stable Diffusion 3</a>, the <code>proj_out</code> module can be skipped using the <code>llm_int8_skip_modules</code> parameter in <a href="/docs/diffusers/pr_14180/en/api/quantization#diffusers.BitsAndBytesConfig">BitsAndBytesConfig</a>:</p> <!> <!> <blockquote class="tip"><p>Learn more about its details in this <a href="https://huggingface.co/blog/4bit-transformers-bitsandbytes" rel="nofollow">blog post</a>.</p></blockquote> <p>This section explores some of the specific features of 4-bit models, such as changing the compute data type, using the Normal Float 4 (NF4) data type, and using nested quantization.</p> <!> <p>To speedup computation, you can change the data type from float32 (the default value) to bf16 using the <code>bnb_4bit_compute_dtype</code> parameter in <a href="/docs/diffusers/pr_14180/en/api/quantization#diffusers.BitsAndBytesConfig">BitsAndBytesConfig</a>:</p> <!> <!> <p>NF4 is a 4-bit data type from the <a href="https://hf.co/papers/2305.14314" rel="nofollow">QLoRA</a> paper, adapted for weights initialized from a normal distribution. You should use NF4 for training 4-bit base models. This can be configured with the <code>bnb_4bit_quant_type</code> parameter in the <a href="/docs/diffusers/pr_14180/en/api/quantization#diffusers.BitsAndBytesConfig">BitsAndBytesConfig</a>:</p> <!> <p>For inference, the <code>bnb_4bit_quant_type</code> does not have a huge impact on performance. However, to remain consistent with the model weights, you should use the <code>bnb_4bit_compute_dtype</code> and <code>torch_dtype</code> values.</p> <!> <p>Nested quantization is a technique that can save additional memory at no additional performance cost. This feature performs a second quantization of the already quantized weights to save an additional 0.4 bits/parameter.</p> <!> <!> <p>Once quantized, you can dequantize a model to its original precision, but this might result in a small loss of quality. Make sure you have enough GPU RAM to fit the dequantized model.</p> <!> <!> <p>Speed up inference with <code>torch.compile</code>. Make sure you have the latest <code>bitsandbytes</code> installed and we also recommend installing <a href="https://pytorch.org/get-started/locally/" rel="nofollow">PyTorch nightly</a>.</p> <!> <p>On an RTX 4090 with compilation, 4-bit Flux generation completed in 25.809 seconds versus 32.570 seconds without.</p> <p>Check out the <a href="https://gist.github.com/sayakpaul/0db9d8eeeb3d2a0e5ed7cf0d9ca19b7d" rel="nofollow">benchmarking script</a> for more details.</p> <!> <ul><li><a href="https://gist.github.com/sayakpaul/c76bd845b48759e11687ac550b99d8b4" rel="nofollow">End-to-end notebook showing Flux.1 Dev inference in a free-tier Colab</a></li> <li><a href="https://github.com/huggingface/diffusers/blob/8c661ea586bf11cb2440da740dd3c4cf84679b85/examples/dreambooth/README_hidream.md#using-quantization" rel="nofollow">Training</a></li></ul> <!> <p></p>`,1);function ps(H,x){O(x,!1),ss(()=>{new URLSearchParams(window.location.search).get("fw")}),L();var J=os();D("169h6sl",t=>{var M=as();$(M,"content",es),n(t,M)});var U=s(p(J),2);K(U,{containerStyle:"float: right; margin-left: 10px; display: inline-flex; position: relative; z-index: 10;"});var T=s(U,2);e(T,{title:"bitsandbytes",local:"bitsandbytes",headingTag:"h1"});var Z=s(T,12);l(Z,{code:"cGlwJTIwaW5zdGFsbCUyMGRpZmZ1c2VycyUyMHRyYW5zZm9ybWVycyUyMGFjY2VsZXJhdGUlMjBiaXRzYW5kYnl0ZXMlMjAtVQ==",highlighted:"pip install diffusers transformers accelerate bitsandbytes -U",lang:"bash",wrap:!1});var j=s(Z,4);N(j,{id:"bnb",options:["8-bit","4-bit"],children:(t,M)=>{var o=Y(),i=p(o);m(i,{id:"bnb",option:"8-bit",children:(a,h)=>{var r=ts(),d=s(p(r),8);l(d,{code:"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",highlighted:`<span class="hljs-keyword">from</span> diffusers <span class="hljs-keyword">import</span> BitsAndBytesConfig <span class="hljs-keyword">as</span> DiffusersBitsAndBytesConfig
<span class="hljs-keyword">from</span> transformers <span class="hljs-keyword">import</span> BitsAndBytesConfig <span class="hljs-keyword">as</span> TransformersBitsAndBytesConfig
<span class="hljs-keyword">import</span> torch
<span class="hljs-keyword">from</span> diffusers <span class="hljs-keyword">import</span> AutoModel
<span class="hljs-keyword">from</span> transformers <span class="hljs-keyword">import</span> T5EncoderModel

quant_config = TransformersBitsAndBytesConfig(load_in_8bit=<span class="hljs-literal">True</span>,)

text_encoder_2_8bit = T5EncoderModel.from_pretrained(
    <span class="hljs-string">&quot;black-forest-labs/FLUX.1-dev&quot;</span>,
    subfolder=<span class="hljs-string">&quot;text_encoder_2&quot;</span>,
    quantization_config=quant_config,
    torch_dtype=torch.float16,
)

quant_config = DiffusersBitsAndBytesConfig(load_in_8bit=<span class="hljs-literal">True</span>,)

transformer_8bit = AutoModel.from_pretrained(
    <span class="hljs-string">&quot;black-forest-labs/FLUX.1-dev&quot;</span>,
    subfolder=<span class="hljs-string">&quot;transformer&quot;</span>,
    quantization_config=quant_config,
    torch_dtype=torch.float16,
)`,lang:"py",wrap:!1});var c=s(d,4);l(c,{code:"dHJhbnNmb3JtZXJfOGJpdCUyMCUzRCUyMEF1dG9Nb2RlbC5mcm9tX3ByZXRyYWluZWQoJTBBJTIwJTIwJTIwJTIwJTIyYmxhY2stZm9yZXN0LWxhYnMlMkZGTFVYLjEtZGV2JTIyJTJDJTBBJTIwJTIwJTIwJTIwc3ViZm9sZGVyJTNEJTIydHJhbnNmb3JtZXIlMjIlMkMlMEElMjAlMjAlMjAlMjBxdWFudGl6YXRpb25fY29uZmlnJTNEcXVhbnRfY29uZmlnJTJDJTBBJTJCJTIwJTIwJTIwdG9yY2hfZHR5cGUlM0R0b3JjaC5mbG9hdDMyJTJDJTBBKQ==",highlighted:`transformer_8bit = AutoModel.from_pretrained(
    &quot;black-forest-labs/FLUX.1-dev&quot;,
    subfolder=&quot;transformer&quot;,
    quantization_config=quant_config,
<span class="hljs-addition">+   torch_dtype=torch.float32,</span>
)`,lang:"diff",wrap:!1});var f=s(c,6);l(f,{code:"ZnJvbSUyMGRpZmZ1c2VycyUyMGltcG9ydCUyMEZsdXhQaXBlbGluZSUwQSUwQXBpcGUlMjAlM0QlMjBGbHV4UGlwZWxpbmUuZnJvbV9wcmV0cmFpbmVkKCUwQSUyMCUyMCUyMCUyMCUyMmJsYWNrLWZvcmVzdC1sYWJzJTJGRkxVWC4xLWRldiUyMiUyQyUwQSUyMCUyMCUyMCUyMHRyYW5zZm9ybWVyJTNEdHJhbnNmb3JtZXJfOGJpdCUyQyUwQSUyMCUyMCUyMCUyMHRleHRfZW5jb2Rlcl8yJTNEdGV4dF9lbmNvZGVyXzJfOGJpdCUyQyUwQSUyMCUyMCUyMCUyMHRvcmNoX2R0eXBlJTNEdG9yY2guZmxvYXQxNiUyQyUwQSUyMCUyMCUyMCUyMGRldmljZV9tYXAlM0QlMjJhdXRvJTIyJTJDJTBBKSUwQSUwQXBpcGVfa3dhcmdzJTIwJTNEJTIwJTdCJTBBJTIwJTIwJTIwJTIwJTIycHJvbXB0JTIyJTNBJTIwJTIyQSUyMGNhdCUyMGhvbGRpbmclMjBhJTIwc2lnbiUyMHRoYXQlMjBzYXlzJTIwaGVsbG8lMjB3b3JsZCUyMiUyQyUwQSUyMCUyMCUyMCUyMCUyMmhlaWdodCUyMiUzQSUyMDEwMjQlMkMlMEElMjAlMjAlMjAlMjAlMjJ3aWR0aCUyMiUzQSUyMDEwMjQlMkMlMEElMjAlMjAlMjAlMjAlMjJndWlkYW5jZV9zY2FsZSUyMiUzQSUyMDMuNSUyQyUwQSUyMCUyMCUyMCUyMCUyMm51bV9pbmZlcmVuY2Vfc3RlcHMlMjIlM0ElMjA1MCUyQyUwQSUyMCUyMCUyMCUyMCUyMm1heF9zZXF1ZW5jZV9sZW5ndGglMjIlM0ElMjA1MTIlMkMlMEElN0QlMEElMEFpbWFnZSUyMCUzRCUyMHBpcGUoKipwaXBlX2t3YXJncyUyQyUyMGdlbmVyYXRvciUzRHRvcmNoLm1hbnVhbF9zZWVkKDApJTJDKS5pbWFnZXMlNUIwJTVE",highlighted:`<span class="hljs-keyword">from</span> diffusers <span class="hljs-keyword">import</span> FluxPipeline

pipe = FluxPipeline.from_pretrained(
    <span class="hljs-string">&quot;black-forest-labs/FLUX.1-dev&quot;</span>,
    transformer=transformer_8bit,
    text_encoder_2=text_encoder_2_8bit,
    torch_dtype=torch.float16,
    device_map=<span class="hljs-string">&quot;auto&quot;</span>,
)

pipe_kwargs = {
    <span class="hljs-string">&quot;prompt&quot;</span>: <span class="hljs-string">&quot;A cat holding a sign that says hello world&quot;</span>,
    <span class="hljs-string">&quot;height&quot;</span>: <span class="hljs-number">1024</span>,
    <span class="hljs-string">&quot;width&quot;</span>: <span class="hljs-number">1024</span>,
    <span class="hljs-string">&quot;guidance_scale&quot;</span>: <span class="hljs-number">3.5</span>,
    <span class="hljs-string">&quot;num_inference_steps&quot;</span>: <span class="hljs-number">50</span>,
    <span class="hljs-string">&quot;max_sequence_length&quot;</span>: <span class="hljs-number">512</span>,
}

image = pipe(**pipe_kwargs, generator=torch.manual_seed(<span class="hljs-number">0</span>),).images[<span class="hljs-number">0</span>]`,lang:"py",wrap:!1}),b(6),n(a,r)},$$slots:{default:!0}});var u=s(i,2);m(u,{id:"bnb",option:"4-bit",children:(a,h)=>{var r=ns(),d=s(p(r),8);l(d,{code:"ZnJvbSUyMGRpZmZ1c2VycyUyMGltcG9ydCUyMEJpdHNBbmRCeXRlc0NvbmZpZyUyMGFzJTIwRGlmZnVzZXJzQml0c0FuZEJ5dGVzQ29uZmlnJTBBZnJvbSUyMHRyYW5zZm9ybWVycyUyMGltcG9ydCUyMEJpdHNBbmRCeXRlc0NvbmZpZyUyMGFzJTIwVHJhbnNmb3JtZXJzQml0c0FuZEJ5dGVzQ29uZmlnJTBBaW1wb3J0JTIwdG9yY2glMEFmcm9tJTIwZGlmZnVzZXJzJTIwaW1wb3J0JTIwQXV0b01vZGVsJTBBZnJvbSUyMHRyYW5zZm9ybWVycyUyMGltcG9ydCUyMFQ1RW5jb2Rlck1vZGVsJTBBJTBBcXVhbnRfY29uZmlnJTIwJTNEJTIwVHJhbnNmb3JtZXJzQml0c0FuZEJ5dGVzQ29uZmlnKGxvYWRfaW5fNGJpdCUzRFRydWUlMkMpJTBBJTBBdGV4dF9lbmNvZGVyXzJfNGJpdCUyMCUzRCUyMFQ1RW5jb2Rlck1vZGVsLmZyb21fcHJldHJhaW5lZCglMEElMjAlMjAlMjAlMjAlMjJibGFjay1mb3Jlc3QtbGFicyUyRkZMVVguMS1kZXYlMjIlMkMlMEElMjAlMjAlMjAlMjBzdWJmb2xkZXIlM0QlMjJ0ZXh0X2VuY29kZXJfMiUyMiUyQyUwQSUyMCUyMCUyMCUyMHF1YW50aXphdGlvbl9jb25maWclM0RxdWFudF9jb25maWclMkMlMEElMjAlMjAlMjAlMjB0b3JjaF9kdHlwZSUzRHRvcmNoLmZsb2F0MTYlMkMlMEEpJTBBJTBBcXVhbnRfY29uZmlnJTIwJTNEJTIwRGlmZnVzZXJzQml0c0FuZEJ5dGVzQ29uZmlnKGxvYWRfaW5fNGJpdCUzRFRydWUlMkMpJTBBJTBBdHJhbnNmb3JtZXJfNGJpdCUyMCUzRCUyMEF1dG9Nb2RlbC5mcm9tX3ByZXRyYWluZWQoJTBBJTIwJTIwJTIwJTIwJTIyYmxhY2stZm9yZXN0LWxhYnMlMkZGTFVYLjEtZGV2JTIyJTJDJTBBJTIwJTIwJTIwJTIwc3ViZm9sZGVyJTNEJTIydHJhbnNmb3JtZXIlMjIlMkMlMEElMjAlMjAlMjAlMjBxdWFudGl6YXRpb25fY29uZmlnJTNEcXVhbnRfY29uZmlnJTJDJTBBJTIwJTIwJTIwJTIwdG9yY2hfZHR5cGUlM0R0b3JjaC5mbG9hdDE2JTJDJTBBKQ==",highlighted:`<span class="hljs-keyword">from</span> diffusers <span class="hljs-keyword">import</span> BitsAndBytesConfig <span class="hljs-keyword">as</span> DiffusersBitsAndBytesConfig
<span class="hljs-keyword">from</span> transformers <span class="hljs-keyword">import</span> BitsAndBytesConfig <span class="hljs-keyword">as</span> TransformersBitsAndBytesConfig
<span class="hljs-keyword">import</span> torch
<span class="hljs-keyword">from</span> diffusers <span class="hljs-keyword">import</span> AutoModel
<span class="hljs-keyword">from</span> transformers <span class="hljs-keyword">import</span> T5EncoderModel

quant_config = TransformersBitsAndBytesConfig(load_in_4bit=<span class="hljs-literal">True</span>,)

text_encoder_2_4bit = T5EncoderModel.from_pretrained(
    <span class="hljs-string">&quot;black-forest-labs/FLUX.1-dev&quot;</span>,
    subfolder=<span class="hljs-string">&quot;text_encoder_2&quot;</span>,
    quantization_config=quant_config,
    torch_dtype=torch.float16,
)

quant_config = DiffusersBitsAndBytesConfig(load_in_4bit=<span class="hljs-literal">True</span>,)

transformer_4bit = AutoModel.from_pretrained(
    <span class="hljs-string">&quot;black-forest-labs/FLUX.1-dev&quot;</span>,
    subfolder=<span class="hljs-string">&quot;transformer&quot;</span>,
    quantization_config=quant_config,
    torch_dtype=torch.float16,
)`,lang:"py",wrap:!1});var c=s(d,4);l(c,{code:"dHJhbnNmb3JtZXJfNGJpdCUyMCUzRCUyMEF1dG9Nb2RlbC5mcm9tX3ByZXRyYWluZWQoJTBBJTIwJTIwJTIwJTIwJTIyYmxhY2stZm9yZXN0LWxhYnMlMkZGTFVYLjEtZGV2JTIyJTJDJTBBJTIwJTIwJTIwJTIwc3ViZm9sZGVyJTNEJTIydHJhbnNmb3JtZXIlMjIlMkMlMEElMjAlMjAlMjAlMjBxdWFudGl6YXRpb25fY29uZmlnJTNEcXVhbnRfY29uZmlnJTJDJTBBJTJCJTIwJTIwJTIwdG9yY2hfZHR5cGUlM0R0b3JjaC5mbG9hdDMyJTJDJTBBKQ==",highlighted:`transformer_4bit = AutoModel.from_pretrained(
    &quot;black-forest-labs/FLUX.1-dev&quot;,
    subfolder=&quot;transformer&quot;,
    quantization_config=quant_config,
<span class="hljs-addition">+   torch_dtype=torch.float32,</span>
)`,lang:"diff",wrap:!1});var f=s(c,6);l(f,{code:"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",highlighted:`<span class="hljs-keyword">from</span> diffusers <span class="hljs-keyword">import</span> FluxPipeline

pipe = FluxPipeline.from_pretrained(
    <span class="hljs-string">&quot;black-forest-labs/FLUX.1-dev&quot;</span>,
    transformer=transformer_4bit,
    text_encoder_2=text_encoder_2_4bit,
    torch_dtype=torch.float16,
    device_map=<span class="hljs-string">&quot;auto&quot;</span>,
)

pipe_kwargs = {
    <span class="hljs-string">&quot;prompt&quot;</span>: <span class="hljs-string">&quot;A cat holding a sign that says hello world&quot;</span>,
    <span class="hljs-string">&quot;height&quot;</span>: <span class="hljs-number">1024</span>,
    <span class="hljs-string">&quot;width&quot;</span>: <span class="hljs-number">1024</span>,
    <span class="hljs-string">&quot;guidance_scale&quot;</span>: <span class="hljs-number">3.5</span>,
    <span class="hljs-string">&quot;num_inference_steps&quot;</span>: <span class="hljs-number">50</span>,
    <span class="hljs-string">&quot;max_sequence_length&quot;</span>: <span class="hljs-number">512</span>,
}

image = pipe(**pipe_kwargs, generator=torch.manual_seed(<span class="hljs-number">0</span>),).images[<span class="hljs-number">0</span>]`,lang:"py",wrap:!1}),b(6),n(a,r)},$$slots:{default:!0}}),n(t,o)},$$slots:{default:!0}});var w=s(j,6);l(w,{code:"cHJpbnQobW9kZWwuZ2V0X21lbW9yeV9mb290cHJpbnQoKSk=",highlighted:'<span class="hljs-built_in">print</span>(model.get_memory_footprint())',lang:"py",wrap:!1});var g=s(w,6);l(g,{code:"ZnJvbSUyMGRpZmZ1c2VycyUyMGltcG9ydCUyMEF1dG9Nb2RlbCUyQyUyMEJpdHNBbmRCeXRlc0NvbmZpZyUwQSUwQXF1YW50aXphdGlvbl9jb25maWclMjAlM0QlMjBCaXRzQW5kQnl0ZXNDb25maWcobG9hZF9pbl80Yml0JTNEVHJ1ZSklMEElMEFtb2RlbF80Yml0JTIwJTNEJTIwQXV0b01vZGVsLmZyb21fcHJldHJhaW5lZCglMEElMjAlMjAlMjAlMjAlMjJoZi1pbnRlcm5hbC10ZXN0aW5nJTJGZmx1eC4xLWRldi1uZjQtcGtnJTIyJTJDJTIwc3ViZm9sZGVyJTNEJTIydHJhbnNmb3JtZXIlMjIlMEEp",highlighted:`<span class="hljs-keyword">from</span> diffusers <span class="hljs-keyword">import</span> AutoModel, BitsAndBytesConfig

quantization_config = BitsAndBytesConfig(load_in_4bit=<span class="hljs-literal">True</span>)

model_4bit = AutoModel.from_pretrained(
    <span class="hljs-string">&quot;hf-internal-testing/flux.1-dev-nf4-pkg&quot;</span>, subfolder=<span class="hljs-string">&quot;transformer&quot;</span>
)`,lang:"py",wrap:!1});var B=s(g,2);e(B,{title:"8-bit (LLM.int8() algorithm)",local:"8-bit-llmint8-algorithm",headingTag:"h2"});var _=s(B,6);e(_,{title:"Outlier threshold",local:"outlier-threshold",headingTag:"h3"});var C=s(_,6);l(C,{code:"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",highlighted:`<span class="hljs-keyword">from</span> diffusers <span class="hljs-keyword">import</span> AutoModel, BitsAndBytesConfig

quantization_config = BitsAndBytesConfig(
    load_in_8bit=<span class="hljs-literal">True</span>, llm_int8_threshold=<span class="hljs-number">10</span>,
)

model_8bit = AutoModel.from_pretrained(
    <span class="hljs-string">&quot;black-forest-labs/FLUX.1-dev&quot;</span>,
    subfolder=<span class="hljs-string">&quot;transformer&quot;</span>,
    quantization_config=quantization_config,
)`,lang:"py",wrap:!1});var R=s(C,2);e(R,{title:"Skip module conversion",local:"skip-module-conversion",headingTag:"h3"});var G=s(R,4);l(G,{code:"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",highlighted:`<span class="hljs-keyword">from</span> diffusers <span class="hljs-keyword">import</span> SD3Transformer2DModel, BitsAndBytesConfig

quantization_config = BitsAndBytesConfig(
    load_in_8bit=<span class="hljs-literal">True</span>, llm_int8_skip_modules=[<span class="hljs-string">&quot;proj_out&quot;</span>],
)

model_8bit = SD3Transformer2DModel.from_pretrained(
    <span class="hljs-string">&quot;stabilityai/stable-diffusion-3-medium-diffusers&quot;</span>,
    subfolder=<span class="hljs-string">&quot;transformer&quot;</span>,
    quantization_config=quantization_config,
)`,lang:"py",wrap:!1});var X=s(G,2);e(X,{title:"4-bit (QLoRA algorithm)",local:"4-bit-qlora-algorithm",headingTag:"h2"});var v=s(X,6);e(v,{title:"Compute data type",local:"compute-data-type",headingTag:"h3"});var z=s(v,4);l(z,{code:"aW1wb3J0JTIwdG9yY2glMEFmcm9tJTIwZGlmZnVzZXJzJTIwaW1wb3J0JTIwQml0c0FuZEJ5dGVzQ29uZmlnJTBBJTBBcXVhbnRpemF0aW9uX2NvbmZpZyUyMCUzRCUyMEJpdHNBbmRCeXRlc0NvbmZpZyhsb2FkX2luXzRiaXQlM0RUcnVlJTJDJTIwYm5iXzRiaXRfY29tcHV0ZV9kdHlwZSUzRHRvcmNoLmJmbG9hdDE2KQ==",highlighted:`<span class="hljs-keyword">import</span> torch
<span class="hljs-keyword">from</span> diffusers <span class="hljs-keyword">import</span> BitsAndBytesConfig

quantization_config = BitsAndBytesConfig(load_in_4bit=<span class="hljs-literal">True</span>, bnb_4bit_compute_dtype=torch.bfloat16)`,lang:"py",wrap:!1});var E=s(z,2);e(E,{title:"Normal Float 4 (NF4)",local:"normal-float-4-nf4",headingTag:"h3"});var V=s(E,4);l(V,{code:"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",highlighted:`<span class="hljs-keyword">from</span> diffusers <span class="hljs-keyword">import</span> BitsAndBytesConfig <span class="hljs-keyword">as</span> DiffusersBitsAndBytesConfig
<span class="hljs-keyword">from</span> transformers <span class="hljs-keyword">import</span> BitsAndBytesConfig <span class="hljs-keyword">as</span> TransformersBitsAndBytesConfig

<span class="hljs-keyword">from</span> diffusers <span class="hljs-keyword">import</span> AutoModel
<span class="hljs-keyword">from</span> transformers <span class="hljs-keyword">import</span> T5EncoderModel

quant_config = TransformersBitsAndBytesConfig(
    load_in_4bit=<span class="hljs-literal">True</span>,
    bnb_4bit_quant_type=<span class="hljs-string">&quot;nf4&quot;</span>,
)

text_encoder_2_4bit = T5EncoderModel.from_pretrained(
    <span class="hljs-string">&quot;black-forest-labs/FLUX.1-dev&quot;</span>,
    subfolder=<span class="hljs-string">&quot;text_encoder_2&quot;</span>,
    quantization_config=quant_config,
    torch_dtype=torch.float16,
)

quant_config = DiffusersBitsAndBytesConfig(
    load_in_4bit=<span class="hljs-literal">True</span>,
    bnb_4bit_quant_type=<span class="hljs-string">&quot;nf4&quot;</span>,
)

transformer_4bit = AutoModel.from_pretrained(
    <span class="hljs-string">&quot;black-forest-labs/FLUX.1-dev&quot;</span>,
    subfolder=<span class="hljs-string">&quot;transformer&quot;</span>,
    quantization_config=quant_config,
    torch_dtype=torch.float16,
)`,lang:"py",wrap:!1});var I=s(V,4);e(I,{title:"Nested quantization",local:"nested-quantization",headingTag:"h3"});var k=s(I,4);l(k,{code:"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",highlighted:`<span class="hljs-keyword">from</span> diffusers <span class="hljs-keyword">import</span> BitsAndBytesConfig <span class="hljs-keyword">as</span> DiffusersBitsAndBytesConfig
<span class="hljs-keyword">from</span> transformers <span class="hljs-keyword">import</span> BitsAndBytesConfig <span class="hljs-keyword">as</span> TransformersBitsAndBytesConfig

<span class="hljs-keyword">from</span> diffusers <span class="hljs-keyword">import</span> AutoModel
<span class="hljs-keyword">from</span> transformers <span class="hljs-keyword">import</span> T5EncoderModel

quant_config = TransformersBitsAndBytesConfig(
    load_in_4bit=<span class="hljs-literal">True</span>,
    bnb_4bit_use_double_quant=<span class="hljs-literal">True</span>,
)

text_encoder_2_4bit = T5EncoderModel.from_pretrained(
    <span class="hljs-string">&quot;black-forest-labs/FLUX.1-dev&quot;</span>,
    subfolder=<span class="hljs-string">&quot;text_encoder_2&quot;</span>,
    quantization_config=quant_config,
    torch_dtype=torch.float16,
)

quant_config = DiffusersBitsAndBytesConfig(
    load_in_4bit=<span class="hljs-literal">True</span>,
    bnb_4bit_use_double_quant=<span class="hljs-literal">True</span>,
)

transformer_4bit = AutoModel.from_pretrained(
    <span class="hljs-string">&quot;black-forest-labs/FLUX.1-dev&quot;</span>,
    subfolder=<span class="hljs-string">&quot;transformer&quot;</span>,
    quantization_config=quant_config,
    torch_dtype=torch.float16,
)`,lang:"py",wrap:!1});var W=s(k,2);e(W,{title:"Dequantizing bitsandbytes models",local:"dequantizing-bitsandbytes-models",headingTag:"h2"});var Q=s(W,4);l(Q,{code:"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",highlighted:`<span class="hljs-keyword">from</span> diffusers <span class="hljs-keyword">import</span> BitsAndBytesConfig <span class="hljs-keyword">as</span> DiffusersBitsAndBytesConfig
<span class="hljs-keyword">from</span> transformers <span class="hljs-keyword">import</span> BitsAndBytesConfig <span class="hljs-keyword">as</span> TransformersBitsAndBytesConfig

<span class="hljs-keyword">from</span> diffusers <span class="hljs-keyword">import</span> AutoModel
<span class="hljs-keyword">from</span> transformers <span class="hljs-keyword">import</span> T5EncoderModel

quant_config = TransformersBitsAndBytesConfig(
    load_in_4bit=<span class="hljs-literal">True</span>,
    bnb_4bit_use_double_quant=<span class="hljs-literal">True</span>,
)

text_encoder_2_4bit = T5EncoderModel.from_pretrained(
    <span class="hljs-string">&quot;black-forest-labs/FLUX.1-dev&quot;</span>,
    subfolder=<span class="hljs-string">&quot;text_encoder_2&quot;</span>,
    quantization_config=quant_config,
    torch_dtype=torch.float16,
)

quant_config = DiffusersBitsAndBytesConfig(
    load_in_4bit=<span class="hljs-literal">True</span>,
    bnb_4bit_use_double_quant=<span class="hljs-literal">True</span>,
)

transformer_4bit = AutoModel.from_pretrained(
    <span class="hljs-string">&quot;black-forest-labs/FLUX.1-dev&quot;</span>,
    subfolder=<span class="hljs-string">&quot;transformer&quot;</span>,
    quantization_config=quant_config,
    torch_dtype=torch.float16,
)

text_encoder_2_4bit.dequantize()
transformer_4bit.dequantize()`,lang:"python",wrap:!1});var F=s(Q,2);e(F,{title:"torch.compile",local:"torchcompile",headingTag:"h2"});var q=s(F,4);N(q,{id:"bnb",options:["8-bit","4-bit"],children:(t,M)=>{var o=Y(),i=p(o);m(i,{id:"bnb",option:"8-bit",children:(a,h)=>{l(a,{code:"dG9yY2guX2R5bmFtby5jb25maWcuY2FwdHVyZV9keW5hbWljX291dHB1dF9zaGFwZV9vcHMlMjAlM0QlMjBUcnVlJTBBJTBBcXVhbnRfY29uZmlnJTIwJTNEJTIwRGlmZnVzZXJzQml0c0FuZEJ5dGVzQ29uZmlnKGxvYWRfaW5fOGJpdCUzRFRydWUpJTBBdHJhbnNmb3JtZXJfNGJpdCUyMCUzRCUyMEF1dG9Nb2RlbC5mcm9tX3ByZXRyYWluZWQoJTBBJTIwJTIwJTIwJTIwJTIyYmxhY2stZm9yZXN0LWxhYnMlMkZGTFVYLjEtZGV2JTIyJTJDJTBBJTIwJTIwJTIwJTIwc3ViZm9sZGVyJTNEJTIydHJhbnNmb3JtZXIlMjIlMkMlMEElMjAlMjAlMjAlMjBxdWFudGl6YXRpb25fY29uZmlnJTNEcXVhbnRfY29uZmlnJTJDJTBBJTIwJTIwJTIwJTIwdG9yY2hfZHR5cGUlM0R0b3JjaC5mbG9hdDE2JTJDJTBBKSUwQXRyYW5zZm9ybWVyXzRiaXQuY29tcGlsZShmdWxsZ3JhcGglM0RUcnVlKQ==",highlighted:`torch._dynamo.config.capture_dynamic_output_shape_ops = <span class="hljs-literal">True</span>

quant_config = DiffusersBitsAndBytesConfig(load_in_8bit=<span class="hljs-literal">True</span>)
transformer_4bit = AutoModel.from_pretrained(
    <span class="hljs-string">&quot;black-forest-labs/FLUX.1-dev&quot;</span>,
    subfolder=<span class="hljs-string">&quot;transformer&quot;</span>,
    quantization_config=quant_config,
    torch_dtype=torch.float16,
)
transformer_4bit.<span class="hljs-built_in">compile</span>(fullgraph=<span class="hljs-literal">True</span>)`,lang:"py",wrap:!1})},$$slots:{default:!0}});var u=s(i,2);m(u,{id:"bnb",option:"4-bit",children:(a,h)=>{l(a,{code:"cXVhbnRfY29uZmlnJTIwJTNEJTIwRGlmZnVzZXJzQml0c0FuZEJ5dGVzQ29uZmlnKGxvYWRfaW5fNGJpdCUzRFRydWUpJTBBdHJhbnNmb3JtZXJfNGJpdCUyMCUzRCUyMEF1dG9Nb2RlbC5mcm9tX3ByZXRyYWluZWQoJTBBJTIwJTIwJTIwJTIwJTIyYmxhY2stZm9yZXN0LWxhYnMlMkZGTFVYLjEtZGV2JTIyJTJDJTBBJTIwJTIwJTIwJTIwc3ViZm9sZGVyJTNEJTIydHJhbnNmb3JtZXIlMjIlMkMlMEElMjAlMjAlMjAlMjBxdWFudGl6YXRpb25fY29uZmlnJTNEcXVhbnRfY29uZmlnJTJDJTBBJTIwJTIwJTIwJTIwdG9yY2hfZHR5cGUlM0R0b3JjaC5mbG9hdDE2JTJDJTBBKSUwQXRyYW5zZm9ybWVyXzRiaXQuY29tcGlsZShmdWxsZ3JhcGglM0RUcnVlKQ==",highlighted:`quant_config = DiffusersBitsAndBytesConfig(load_in_4bit=<span class="hljs-literal">True</span>)
transformer_4bit = AutoModel.from_pretrained(
    <span class="hljs-string">&quot;black-forest-labs/FLUX.1-dev&quot;</span>,
    subfolder=<span class="hljs-string">&quot;transformer&quot;</span>,
    quantization_config=quant_config,
    torch_dtype=torch.float16,
)
transformer_4bit.<span class="hljs-built_in">compile</span>(fullgraph=<span class="hljs-literal">True</span>)`,lang:"py",wrap:!1})},$$slots:{default:!0}}),n(t,o)},$$slots:{default:!0}});var A=s(q,6);e(A,{title:"Resources",local:"resources",headingTag:"h2"});var S=s(A,4);P(S,{source:"https://github.com/huggingface/diffusers/blob/main/docs/source/en/quantization/bitsandbytes.md"}),b(2),n(H,J),ls()}export{ps as component};