Hugging Face's logo

stabilityai
/
stable-diffusion-3-medium_amdgpu

Text-to-Image
Diffusers
ONNX
Model card Files
xet
Community
3
stable-diffusion-3-medium_amdgpu / README.md
entropi-sai's picture
entropi-sai
Update Model Card for NPUv0530 (#3)
da1c602 verified
preview code
|
raw
history blame contribute delete
5.89 kB
metadata
license: other
license_name: stabilityai-ai-community
license_link: LICENSE.md
base_model:
  - stabilityai/stable-diffusion-3-medium
pipeline_tag: text-to-image
library_name: diffusers

stabilityai/stable-diffusion-3-medium - AMD Optimized ONNX and Ryzen(TM) AI NPU

This repository hosts the AMD Optimized version of Stable Diffusion 3 Medium and the AMD Ryzen™ AI optimized version of SD 3 Medium created in collaboration with AMD.

The AMDGPU model is an AMD-optimized ONNX port of the Stable Diffusion 3 Medium model offering significantly higher inferencing speeds compared to the base model on compatible AMD hardware.

The ONNX-ported Ryzen™ AI model is the world’s first Block FP16 model with the UNET and VAE decoder completely in Block FP16. Built for the AMD XDNA™ 2 based NPU, this model combines the accuracy of FP16 with the performance of INT8.

Model Description

Refer to the Stable Diffusion 3 Medium Model card for more details.

_io32 vs. _io16

  • _io32: Model input is fp32, model will convert the input to fp16, perform ops in fp16 and write the final result in fp32

  • _io16: Model input is fp16, perform ops in fp16 and write the final result in fp16

Running

GPU: Use Amuse GUI application to run it: https://www.amuse-ai.com/. Use *_io32 version to run with the Amuse application

NPU: On a compatible XDNA(TM) 2 NPU device, Install and Open Amuse GUI, Click on Advanced Mode. Download SD 3 Medium AMDGPU from Model Manager. Click on Image Generation. Change variant to Ryzen AI. Load. 

Inference Result

Intro Image Prompt="a majestic Royal Bengal Tiger on the mountain top overlooking beautiful Lake Tahoe snowy mountains and deep blue lake, deep blue sky, ultra hd, 8k, photorealistic"

License

  • Community License: Free for research, non-commercial, and commercial use for organizations or individuals with less than $1M in total annual revenue. More details can be found in the Community License Agreement. Read more at https://stability.ai/license.
  • For individuals and organizations with annual revenue above $1M: please contact us to get an Enterprise License.

Model Sources

For research purposes, we recommend our Github repository (https://github.com/Stability-AI/sd3-ref), which provides a reference implementation for SD3 inference.

Repository: https://github.com/Stability-AI/sd3-ref

Stable Diffusion 3 Paper: https://arxiv.org/pdf/2403.03206

Training Data and Strategy

This model was trained on a wide variety of data, including synthetic data and filtered publicly available data.

Uses

Intended Uses

Intended uses include the following:

  • Generation of artworks and use in design and other artistic processes.
  • Applications in educational or creative tools.
  • Research on generative models, including understanding the limitations of generative models.

All uses of the model must be in accordance with our Acceptable Use Policy.

Out-of-Scope Uses

The model was not trained to be factual or true representations of people or events. As such, using the model to generate such content is out-of-scope of the abilities of this model.

Safety

As part of our safety-by-design and responsible AI development approach, we prioritize integrity from the earliest stages of model development. We implement safeguards throughout the development process to help reduce the risk of misuse. While we’ve built in protections to mitigate certain harms, we encourage developers to test responsibly based on their intended use cases and apply additional mitigations as needed.

For more about our approach to Safety, please visit our Safety page.

Integrity Evaluation

Our integrity evaluation methods include structured evaluations and red-teaming testing for certain harms. Testing was conducted primarily in English and may not cover all possible harms.

Risks identified and mitigations:

  • Harmful content: We have used filtered data sets when training our models and implemented safeguards that attempt to strike the right balance between usefulness and preventing harm. However, this does not guarantee that all possible harmful content has been removed. All developers and deployers should exercise caution and implement content safety guardrails based on their specific product policies and application use cases.
  • Misuse: Technical limitations and developer and end-user education can help mitigate against malicious applications of models. All users are required to adhere to our Acceptable Use Policy, including when applying fine-tuning and prompt engineering mechanisms. Please reference the Stability AI Acceptable Use Policy for information on violative uses of our products.
  • Privacy violations: Developers and deployers are encouraged to adhere to privacy regulations with techniques that respect data privacy.

Contact

Please report any issues with the model or contact us: