Equivalent to np.sum apart the fact that numpy upcast uint8 and int32 to int64 while tensorflow returns the same dtype as the input.
確定! 回上一頁