#include "tensorstore/util/bfloat16.h" template <typename T> requires std::is_convertible_v<T, float>explicit tensorstore::BFloat16::BFloat16(T x); Possibly lossy conversion from any type convertible to float.