3h ago

Grigory Bartosh presents Dual-Rate Diffusion from Google DeepMind to accelerate inference in standard and distilled diffusion models by interleaving a heavy context encoder with a lightweight denoiser

The method lowers computational costs during generation for tested models.

——0——
Original post
Jan-Willem van de Meent#1677@JWVDMOPGrigory BartoshGrigory Bartosh@GRIGORYBARTOSH

🚀 Excited to share my @GoogleDeepMind student researcher project: Dual-Rate Diffusion✨ ⚡ A simple construction that speeds up both regular diffusion and distilled models by interleaving a heavy context encoder with a light conditional denoiser. 🧵👇

6:27 AM · May 20, 2026 View on X
QUOTE POSTEmiel Hoogeboom#1920Emiel Hoogeboom@EMIEL_HOOGEBOOM

Super cool work by @GrigoryBartosh thoroughly exploring the options for cheaper generation with diffusion models ✨

Grigory BartoshGrigory Bartosh@GrigoryBartosh

🚀 Excited to share my @GoogleDeepMind student researcher project: Dual-Rate Diffusion✨ ⚡ A simple construction that speeds up both regular diffusion and distilled models by interleaving a heavy context encoder with a light conditional denoiser. 🧵👇

1:27 PM · May 20, 2026 · 4.5K Views
4:29 PM · May 20, 2026 · 50 Views
Grigory Bartosh presents Dual-Rate Diffusion from Google DeepMind to accelerate inference in standard and distilled diffusion models by interleaving a heavy context encoder with a lightweight denoiser · Digg