Model weights

[ZonglinL]

Are the model weights the same as those reported in the paper? I tried to use the hed model but found the results differ a lot.

[lzm2275965881]

Are the model weights the same as those reported in the paper? I tried to use the hed model but found the results differ a lot.

The ssim and fid score of hed.safetensors are consistent with those in the paper. What are the ssim and fid you calculated?

[ZonglinL]

I got 0.71 for SSIM and 35 for FID. I use the provided inference code instead of the evaluation. I will try the evaluation code to see if this is the problem. Thank you!

[ZonglinL]

Are the model weights the same as those reported in the paper? I tried to use the hed model but found the results differ a lot.

The ssim and fid score of hed.safetensors are consistent with those in the paper. What are the ssim and fid you calculated?

Thank you for your response. I just figured out that when I run sample_t2i, the generated images are out of the range [-1,1] too much, and I am still trying to figure out the reason. After changing to the code of evluation it's good. Maybe there are some bugs on my added codes. Thank you!