Really nice work! When reading through the paper, I have some questions about the proposed soft-min-snr loss. Would appreciate your feedback on this.
In eq (5) of the hourglass diffusion transformers, it's mentioned that c_out^{-2}(\sigma) is incorporated, however, based on the definition of c_out, eq (5) should be
In this code, gamma is hardcoded to depend on sigma_data, with gamma being chosen as gamma = sigma_data^-2. This, combined with the preconditioner compensation, leads to the formula you're seeing.
Really nice work! When reading through the paper, I have some questions about the proposed soft-min-snr loss. Would appreciate your feedback on this.
c_out^{-2}(\sigma)
is incorporated, however, based on the definition ofc_out
, eq (5) should beThe
\gamma=4 or 5
proposed in the paper doesn't seem to be used. Am I missing anything here?