You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Thank you so much for your work on text-to-image generation!
I noticed that in your paper, you evaluated the reconstruction performance of BSQ under different codebook sizes. However, the utilization rate of BSQ with different codebook sizes was not mentioned. Could you kindly provide information on this aspect?
The text was updated successfully, but these errors were encountered:
@walking-shadow Hello, thanks for your appreciation of our work.
For model of d=16, the codebook utilization is almost 100%
For model of d=32 and d=64, evaluating utilization is extremely time-consuming. Therefore, we didn't get the quantitative result.
Thank you so much for your work on text-to-image generation!
I noticed that in your paper, you evaluated the reconstruction performance of BSQ under different codebook sizes. However, the utilization rate of BSQ with different codebook sizes was not mentioned. Could you kindly provide information on this aspect?
The text was updated successfully, but these errors were encountered: