Hi, when I use the default config to generate p_ and score_ in SO2Schedule, I noticed that some 0 values occur in p_, which subsequently lead to NaN values in score_.
I am wondering if this will affect training?

Thank you for your help!