hello, very good work!
I'm having some problems with the sampling process, the loss seems to be decreasing, but the sampled picture is not right, it's noisy and it's only the output of the cal, the sample is still noisy, has anyone come across a similar problem, thanks a lot!This result is not expected to improve with continued training, my lr is 1e-4,batchsize is 1, no dpmsolver
cal：
sample：
loss：
This looks bad!
hello, very good work!
I'm having some problems with the sampling process, the loss seems to be decreasing, but the sampled picture is not right, it's noisy and it's only the output of the cal, the sample is still noisy, has anyone come across a similar problem, thanks a lot!This result is not expected to improve with continued training, my lr is 1e-4,batchsize is 1, no dpmsolver
cal：