no code implementations • 28 May 2024 • Seokil Ham, Sangmin Woo, Jin-Young Kim, Hyojun Go, Byeongjun Park, Changick Kim
We present Diffusion Model Patching (DMP), a simple method to boost the performance of pre-trained diffusion models that have already reached convergence, with a negligible increase in parameters.
1 code implementation • 14 Mar 2024 • Byeongjun Park, Hyojun Go, Jin-Young Kim, Sangmin Woo, Seokil Ham, Changick Kim
To achieve this, we employ a sparse mixture-of-experts within each transformer block to utilize semantic information and facilitate handling conflicts in tasks through parameter isolation.