LoRA fine-tuning of OpenAI's GLIDE diffusion model on CelebA for compositional face generation. Combine and negate attributes, e.g. "a woman AND smiling AND NOT glasses", with composable diffusion guidance.
machine-learning computer-vision deep-learning pytorch glide image-generation lora text-to-image celeba face-generation fine-tuning peft diffusion-models generative-ai classifier-free-guidance composable-diffusion compositional-generation
-
Updated
Oct 11, 2026 - Jupyter Notebook