Performance in high scale
yuhuUSTC opened this issue · 0 comments
yuhuUSTC commented
Dear Zhao,
Thank you for sharing this great work and open source!
When following this work, I found that both UniPC and DPM-solver++ performs worse than DDIM under high classifier guidance scale, eg, scale=8 or 16. Concretely, under scale=8 or 16, the performance of DDIM > DPM-solver++ > UniPC.\
This contradicts the reported results in the paper.
Thank you for your explanation and help.