Reproduction for distribution matching with IPC=50

[gihyunkim]

Hi, I'm gihyun kim interested to your research and codes. I read some papers and the DM method shows 63% accuracy on CIFAR10 with ipc=50. However, I can not reproduce the performance by your code somehow, only got 57% accuracy.

I use default setting of argument except the number of experiments and number of evaluation, which I set 1 and 5 each.

Can you let me know how to reproduce your performance?

Thank you.