Google launches the high - speed text - to - diffusion model Gemini Diffusion

Google Launches High-Speed Text Diffusion Model Gemini Diffusion

At today's I/O conference, Google unveiled its latest research model, Gemini Diffusion, a state-of-the-art text diffusion model that learns to generate outputs by transforming random noise into coherent text or code, much like the current state-of-the-art image and video generation models. The experimental demo version of Gemini Diffusion released today generates content significantly faster than the currently known fastest models, while having comparable coding performance. The sampling speed without request overhead is 1479 tokens per second, with an additional overhead of approximately 0.84 seconds per request. The model now has an open waitlist for registration.

—— Google Blog

via Wind Vane Reference Express - Telegram Channel