Range mismatch in RealismLoss?

[beomjin-ahn]

Hi, thank you for sharing the code, and I have a question about your implementation.
I found that the both of the m_r and m_low are generated by tanh function (=[-1, 1]).
But the RealismLoss seems to calculate the L1 loss directly with (m_tar, m_r) and (m_tar, m_low) which is ranged in [0, 1]. I'm a bit curious about this is the correct one.
Thanks! 🙂