mindspore.ops.MatMul

class mindspore.ops.MatMul(transpose_a=False, transpose_b=False)[source]

Multiplies matrix a and matrix b.

(O u t p u t)_{i j} = \sum_{k = 1}^{p} a_{i k} b_{k j} = a_{i 1} b_{1 j} + a_{i 2} b_{2 j} + \dots + a_{i p} b_{p j}, p \in N

where the $i, j$ indicates the output of the i-th row and j-th column element.

Note

If $N * M$ cannot be divided by 16, the performance will be poor in ascend environment.
The dtype of inputs must be same.
On Ascend, float64 doesn't be supported.

Parameters

transpose_a (bool) – If True , a is transposed before multiplication. Default: False .
transpose_b (bool) – If True , b is transposed before multiplication. Default: False .

Inputs:

a (Tensor) - The first tensor to be multiplied. The shape of the tensor is $(N, C)$ . If transpose_a is True , its shape must be $(C, N)$ after transpose.
b (Tensor) - The second tensor to be multiplied. The shape of the tensor is $(C, M)$ . If transpose_b is True , its shape must be $(M, C)$ after transpose.

Outputs:

Tensor, the shape of the output tensor is $(N, M)$ .

Raises

TypeError – If transpose_a or transpose_b is not a bool.
TypeError – If the dtype of a and the dtype of b are not the same.
ValueError – If the column of matrix dimensions of a is not equal to the row of matrix dimensions of b.
ValueError – If length of shape of a or b is not equal to 2.

Supported Platforms:: Ascend GPU CPU

Examples

>>> import mindspore
>>> import numpy as np
>>> from mindspore import Tensor, ops
>>> a = Tensor(np.ones(shape=[1, 3]), mindspore.float32)
>>> b = Tensor(np.ones(shape=[3, 4]), mindspore.float32)
>>> matmul = ops.MatMul()
>>> output = matmul(a, b)
>>> print(output)
[[3. 3. 3. 3.]]