FFusionXL-BASE

ONNX VersionSUPORTING NVIDIA / AMD / INTEL and anything running on Direct X <a target="_blank" rel="ugc" href="https://github.com/microsoft/DirectML#dxdispatch-tool">https://github.com/microsoft/DirectML#dxdispatch-tool</a> <a target="_blank" rel="ugc" href="https://github.com/ssube/onnx-web">https://github.com/ssube/onnx-web</a>you can use pipeline = ORTStableDiffusionXLPipeline.from_pretrained FFusion/FFusionXL-BASE is availableor load from dir (when extracting the zip from here)<pre><code>pipeline = ORTStableDiffusionXLPipeline.from_pretrained(model_dir, provider="DmlExecutionProvider")</code></pre>We are proud to announce a fully optimized Microsoft ONNX Version exclusively compatible with the latest DirectML Execution Provider. All the ONNX files are optimized (Quantization) to fp16 for fast inference and training across all devices.The Vae_Decoder is kept at fp32 with settings:<pre><code>"float16": false,
"use_gpu": true,
"keep_io_types": true,
"force_fp32_ops": ["RandomNormalLike"]
</code></pre>to avoid black screens and broken renders. As soon as a proper solution for a full fp16 VAE decoder arrives, we will update it. VAE encoder and everything else is fully optimized 🤟.Our ONNX is OPTIMIZED using ONNX v8:<ul><li>producer: onnxruntime.transformers 1.15.1</li><li>imports: ai.onnx v18, <a target="_blank" rel="ugc" href="http://com.microsoft">com.microsoft</a>.nchwc v1, <a target="_blank" rel="ugc" href="http://ai.onnx.ml">ai.onnx.ml</a> v3, <a target="_blank" rel="ugc" href="http://com.ms">com.ms</a>.internal.nhwc v19, <a target="_blank" rel="ugc" href="http://ai.onnx.training">ai.onnx.training</a> v1, <a target="_blank" rel="ugc" href="http://ai.onnx.preview.training">ai.onnx.preview.training</a> v1, <a target="_blank" rel="ugc" href="http://com.microsoft">com.microsoft</a> v1, <a target="_blank" rel="ugc" href="http://com.microsoft">com.microsoft</a>.experimental v1, org.pytorch.aten v1, <a target="_blank" rel="ugc" href="http://com.microsoft">com.microsoft</a>.dml v1, graph: torch_jit</li></ul>🔖 ### 📌 ONNX DetailsNETRON Detrails:InstallmacOS: <a target="_blank" rel="ugc" href="https://github.com/lutzroeder/netron/releases/latest">Download</a> the <code>.dmg</code> file or run <code>brew install --cask netron</code>Linux: <a target="_blank" rel="ugc" href="https://github.com/lutzroeder/netron/releases/latest">Download</a> the <code>.AppImage</code> file or run <code>snap install netron</code>Windows: <a target="_blank" rel="ugc" href="https://github.com/lutzroeder/netron/releases/latest">Download</a> the <code>.exe</code> installer or run <code>winget install -s winget netron</code><a target="_blank" rel="ugc" href="https://netron.app/">https://netron.app/</a>-- NETRON browser version: <a target="_blank" rel="ugc" href="https://netron.app/?url=https://huggingface.co/FFusion/FFusionXL-BASE/blob/main/text_encoder/model.onnx">Start Text Encoder</a>--NETRON browser version: <a target="_blank" rel="ugc" href="https://netron.app/?url=https://huggingface.co/FFusion/FFusionXL-BASE/blob/main/text_encoder_2/model.onnx">Start Text Encoder 2</a>--NETRON browser version: <a target="_blank" rel="ugc" href="https://netron.app/?url=https://huggingface.co/FFusion/FFusionXL-BASE/blob/main/vae_decoder/model.onnx">Start VAE decoder</a>--NETRON browser version: <a target="_blank" rel="ugc" href="https://netron.app/?url=https://huggingface.co/FFusion/FFusionXL-BASE/blob/main/vae_encoder/model.onnx">Start VAE encoder</a>--NETRON browser version: <a target="_blank" rel="ugc" href="https://netron.app/?url=https://huggingface.co/stabilityai/FFusion/FFusionXL-BASE/blob/main/unet/model.onnx">Start UNET</a>🔖 ### 📌 AMD Support for Microsoft® DirectML Optimization of Stable DiffusionAMD has released support for Microsoft DirectML optimizations for Stable Diffusion, working closely with Microsoft for optimal performance on AMD devices.<a target="_blank" rel="ugc" href="https://microsoft.github.io/DirectML/">Microsoft DirectML</a> <a target="_blank" rel="ugc" href="https://gpuopen.com/amd-microsoft-directml-stable-diffusion/">AMD Microsoft DirectML Stable Diffusion</a>🔖 ### 📌 ONNX Inference Instructions🔖 ### 📌 Text-to-ImageHere is an example of how you can load an ONNX Stable Diffusion model and run inference using ONNX Runtime:<pre><code>from optimum.onnxruntime import ORTStableDiffusionPipeline

model_id = "FFusion/FFusionXL-BASE"
pipeline = ORTStableDiffusionPipeline.from_pretrained(model_id)
prompt = "sailing ship in storm by Leonardo da Vinci"
images = pipeline(prompt).images</code></pre>

Files with SHA256: 2683382eb0880175a4e0047b2e62239a23c41a8b4206ed7ae4688fe56acc9149