Function std::half_round
float half_round(float v, int dtype)Round 'v' to the nearest value representable in 'dtype' (returned as float32). Any dtype other than TENSOR_FP16/TENSOR_BF16 returns 'v'.
float half_round(float v, int dtype)Round 'v' to the nearest value representable in 'dtype' (returned as float32). Any dtype other than TENSOR_FP16/TENSOR_BF16 returns 'v'.