Matrix Multiplication - Matrix dot Matrix

[D1Asif]

From maths, we know that to multiply a matrix with another matrix the order of the matrix has to be like this:
1st matrix: m x n
2nd matrix: n x p
that is, 1st matrix's number of column has to be equal to the number of rows of the second matrix. But in the video, we can see that [1, 2, 3] dot [1, 2, 3] is 14. How is it possible?

[Perian-Yan]

Hi,
do you mean pytorch.dot()? If so, it's used for computing the inner product of two 1D vectors.

There are two types of multiplication: one is matrix multiplication, as you said. We can use torch.matmul(a,b) or @. The other one is element-wise multiplication and we use * or torch.mul(a,b)

Hope this helps.

[D1Asif]

I was asking why the torch.matmul(a, b) = 14 if a = b = [1, 2, 3]. Shouldn't it be an error? Since the inner dimensions don't match?

[Perian-Yan]

In the torch.matmul(a,b) document: If both tensors are 1-dimensional, the dot product (scalar) is returned.

[D1Asif]

looks like the sum of all the elements of scalar product is returned.