Hi, great thanks for your inspiring work! I am trying to reproduce the data-free KD experiment on CIFAR10. The descriptions in the paper and supplementary material help a lot, but I cannot find the temperature of KD you used. Could you help with this?
The code in the cifar10 directory can only generate one batch of data, there is no KD training there. If possible, could you complete the left training part? I am afraid unofficial implementations may not reproduce your results (as noted in other issues). Really appreciate it!
Regards,