I\'m experiencing a weird behaviour when converting a scalar or array to numpy.float32. I suspect it\'s linked to the precision limit but can\'t exactly pinpoint where it\'s com
