max_vals, _ = a.max(axis=1, keepdim=True) additions ... row to the next is equivalent to the maximum, so you can take the maximums first, ...
確定! 回上一頁