Is related to a problem in numpy where subtraction of float64 numbers from float32 numbers somehow results in truncated values. Fix is easy, as we only need to ensure we only subtract float64 from float64 and only after that convert into float32 if necessary.
Is related to a problem in
numpy
where subtraction of float64 numbers from float32 numbers somehow results in truncated values. Fix is easy, as we only need to ensure we only subtract float64 from float64 and only after that convert into float32 if necessary.