The issue of efficiency

#42
by Epiphany10086 - opened

This work is nearly perfect! However, during testing, I noticed an interesting fact: under the same image size, the inference speed of FLUX.1-Fill far exceeds that of the generative model FLUX.1. But after reviewing the source code, I found that FLUX.1-Fill still follows a complete diffusion process. This has confused me, and I’m very eager to hear an explanation. Thank you!

Sign up or log in to comment