Closed aisagarw closed 1 year ago
I'm not sure but I guess that excessive gamma value could be the reason. The best hyperparameter can be different depending on what pretrained model is in use.
Can you please tell me which variable in your code refers to gamma? I can see only _schgamma
I meant "hs_coeff"!
Probably 0.3~0.4 would be the best.
Thanks
Thanks! That helped in improving the results. Another doubt I have is - which argument in main.py refers to t_boost?
Sorry for the confusion.
't_boost' is 't_noise' in main.py!
sure, thanks!
Hi, I have been trying to generate results (diff_style) using the test images available for afhq dog dataset.. I downloaded the afhqdog_pt2 checkpoint from the provided link. Attaching the results I am getting for these test images for the diffstyle case. I wanted to understand what could be the possible reasons for this. Maybe I am missing out on setting certain hyperparameters? Would appreciate any help in fixing the issue. Thanks in advance!