Hello,
Let me share a project that ported “Teach StableDiffusion new concepts via Textual Inversion” on Keras.io example into TensorFlow Extended(TFX) pipeline. I have simply made the example codes into script and port it into TFX pipeline.
Furthermore, TFX pipeline includes a custom component (HFPusher
) which automatically pushes fine-tuned text_encoder
to a Hugging Face Model repository and publishes Gradio application on Hugging Face Space repository. In this case, the appliction automatically refers to the trained text_encoder
out of the box.
Full details and step-by-step guide is included. Please check out the project repository if you are interested!
: GitHub - deep-diver/textual-inversion-pipeline