`Nans' observed in one environment and not in other

vision
1

Hi,

I am training one of the models on a dataset. It turns out that in one setting, I observe nans in one environment and do not observe nans in another. How to resolve this? Mismatch of which library can cause this? Should I copy the list of libraries installed in both conda environments?

Thanks,

2

Can you please post the exact variable where you are seeing NaN’s and what operations you performed to the variable before you get the NaN’s? Essentially, I am asking you to post your code.