That's exactly what DiffusionGemma does. Instead of writing one word at a time (like GPT or Gemma), it generates 256 tokens simultaneously on a canvas and refines them through iterative denoising — borrowing ideas from image diffusion models like Stable Diffusion, but adapted for text.
Gaurav14cs17/DiffusionGemma: That's exactly what DiffusionGemma does. Instead of writing one word at a time (like GPT or Gemma), it generates 256 tokens simultaneously on a canvas and refines them through iterative denoising — borrowing ideas from image diffusion models like Stable Diffusion, but adapted for text.
