Image generation with Stable Diffusion v3 and OpenVINO¶

Stable Diffusion V3 is next generation of latent diffusion image Stable Diffusion models family that outperforms state-of-the-art text-to-image generation systems in typography and prompt adherence, based on human preference evaluations. In comparison with previous versions, it based on Multimodal Diffusion Transformer (MMDiT) text-to-image model that features greatly improved performance in image quality, typography, complex prompt understanding, and resource-efficiency.

More details about model can be found in model card, research paper and Stability.AI blog post. In this tutorial, we will consider how to convert Stable Diffusion v3 for running with OpenVINO. An additional part demonstrates how to run optimization with NNCF to speed up pipeline. If you want to run previous Stable Diffusion versions, please check our other notebooks:

• Stable Diffusion
• Stable Diffusion v2
• Stable Diffusion XL
• LCM Stable Diffusion
• Turbo SDXL
• Turbo SD

Table of contents:¶

• Prerequisites
• Build PyTorch pipeline
• Convert models with OpenVINO
  • Transformer
  • T5 Text Encoder
  • Clip text encoders
  • VAE
• Prepare OpenVINO inference pipeline
• Run OpenVINO model
• Quantization
  • Prepare calibration dataset
  • Run Quantization
  • Run Weights Compression
  • Compare model file sizes
  • Compare inference time of the FP16 and optimized pipelines
• Interactive demo

Installation Instructions¶

This is a self-contained example that relies solely on its own code.

We recommend running the notebook in a virtual environment. You only need a Jupyter server to start. For details, please refer to Installation Guide.

Prerequisites¶

back to top ⬆️

Build PyTorch pipeline¶

back to top ⬆️

Note: run model with notebook, you will need to accept license agreement. You must be a registered user in 🤗 Hugging Face Hub. Please visit HuggingFace model card, carefully read terms of usage and click accept button. You will need to use an access token for the code below to run. For more information on access tokens, refer to this section of the documentation. You can login on Hugging Face Hub in notebook environment, using following code:

We will use Diffusers library integration for running Stable Diffusion v3 model. You can find more details in Diffusers documentation. Additionally, we can apply optimization for pipeline performance and memory consumption:

• Use flash SD3. Flash Diffusion is a diffusion distillation method proposed in Flash Diffusion: Accelerating Any Conditional Diffusion Model for Few Steps Image Generation. The model represented as a 90.4M LoRA distilled version of SD3 model that is able to generate 1024x1024 images in 4 steps. If you want disable it, you can unset checkbox Use flash SD3
• Remove T5 text encoder. Removing the memory-intensive 4.7B parameter T5-XXL text encoder during inference can significantly decrease the memory requirements for SD3 with only a slight loss in performance. If you want to use this model in pipeline, please set use t5 text encoder checkbox.

Convert models with OpenVINO¶

back to top ⬆️

Starting from 2023.0 release, OpenVINO supports PyTorch models directly via Model Conversion API. ov.convert_model function accepts instance of PyTorch model and example inputs for tracing and returns object of ov.Model class, ready to use or save on disk using ov.save_model function.

The pipeline consists of four important parts:

• Clip and T5 Text Encoders to create condition to generate an image from a text prompt.
• Transformer for step-by-step denoising latent image representation.
• Autoencoder (VAE) for decoding latent space to image.

We will use convert_sd3 helper function defined in sd3_helper.py that create original PyTorch model and convert each part of pipeline using ov.convert_model.

Prepare OpenVINO inference pipeline¶

back to top ⬆️

Run OpenVINO model¶

back to top ⬆️

Quantization¶

back to top ⬆️

NNCF enables post-training quantization by adding quantization layers into model graph and then using a subset of the training dataset to initialize the parameters of these additional quantization layers. Quantized operations are executed in INT8 instead of FP32/FP16 making model inference faster.

According to OVStableDiffusion3Pipeline structure, the transformer model takes up significant portion of the overall pipeline execution time. Now we will show you how to optimize the UNet part using NNCF to reduce computation cost and speed up the pipeline. Quantizing the rest of the pipeline does not significantly improve inference performance but can lead to a substantial degradation of accuracy. That's why we use 4-bit weight compression for the rest of the pipeline to reduce memory footprint.

Please select below whether you would like to run quantization to improve model inference speed.

NOTE: Quantization is time and memory consuming operation. Running quantization code below may take some time.

Let's load skip magic extension to skip quantization if to_quantize is not selected

Prepare calibration dataset¶

back to top ⬆️

We use a portion of google-research-datasets/conceptual_captions dataset from Hugging Face as calibration data. We use prompts below to guide image generation and to determine what not to include in the resulting image.

To collect intermediate model inputs for calibration we should customize CompiledModel. We should set the height and width of the image to 512 to reduce memory consumption during quantization.

Run Quantization¶

back to top ⬆️

Quantization of the first Convolution layer impacts the generation results. We recommend using IgnoredScope to keep accuracy sensitive layers in FP16 precision.

Run Weights Compression¶

back to top ⬆️

Quantizing of the Text Encoders and Autoencoder does not significantly improve inference performance but can lead to a substantial degradation of accuracy.

For reducing model memory consumption we will use weights compression. The Weights Compression algorithm is aimed at compressing the weights of the models and can be used to optimize the model footprint and performance of large models where the size of weights is relatively larger than the size of activations, for example, Large Language Models (LLM). Compared to INT8 compression, INT4 compression improves performance even more, but introduces a minor drop in prediction quality.

Let's compare the images generated by the original and optimized pipelines.

Compare model file sizes¶

back to top ⬆️

Compare inference time of the FP16 and optimized pipelines¶

back to top ⬆️

To measure the inference performance of the FP16 and optimized pipelines, we use mean inference time on 5 samples.

NOTE: For the most accurate performance estimation, it is recommended to run benchmark_app in a terminal/command prompt after closing other applications.

Interactive demo¶

back to top ⬆️

Please select below whether you would like to use the quantized models to launch the interactive demo.