How to arrange dataset for optimal m_in and m_out values?
utkutpcgl opened this issue · comments
Thanks for the great contribution!
To optimize m_in and m_out how did you arrange the validation dataset. I see that you have used 10% from the cifar10 dataset here:
Line 93 in 77f3c09
Did you use the entire auxiliary dataset (tiny images) to measure the OOD performance of the chosen parameters (m_in and m_out)?