[1.15] Fix precision loss (int vs float mixup)

[1.15] Fix precision loss (int vs float mixup) #836