Fast Inference Through The Reuse Of Attention Maps In Diffusion Models