mindspore.ops.BatchMatMul

class mindspore.ops.BatchMatMul(transpose_a=False, transpose_b=False)

Computes matrix multiplication between two tensors by batch.

$$$$

text{output}[…, :, :] = text{matrix}(x[…, :, :]) * text{matrix}(y[…, :, :])

The rank of the two input tensors must be at least 2, and the two input tensors must have the same rank if the environment is GPU or CPU.

Parameters

• transpose_a (bool) – If True , the last two dimensions of x is transposed before multiplication. Default: False .

• transpose_b (bool) – If True , the last two dimensions of y is transposed before multiplication. Default: False .

Inputs:

• x (Tensor) - The first tensor to be multiplied. The shape of the tensor is $(\ast B,N,C)$, where $\ast B$ represents the batch size which can be multidimensional, $N$ and $C$ are the size of the last two dimensions. If transpose_a is True , its shape must be $(\ast B,C,N)$.

• y (Tensor) - The second tensor to be multiplied. The shape of the tensor is $(\ast B,C,M)$. If transpose_b is True , its shape must be $(\ast B,M,C)$.

Outputs:

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

Raises

• TypeError – If transpose_a or transpose_b is not a bool.

• ValueError – If length of shape of x is not equal to length of shape of y or length of shape of inputs is less than 2.

Supported Platforms:

Ascend GPU CPU

Examples

>>> import mindspore
>>> import numpy as np
>>> from mindspore import Tensor, ops
>>> x = Tensor(np.ones(shape=[2, 4, 1, 3]), mindspore.float32)
>>> y = Tensor(np.ones(shape=[2, 4, 3, 4]), mindspore.float32)
>>> batmatmul = ops.BatchMatMul()
>>> output = batmatmul(x, y)
>>> print(output.shape)
(2, 4, 1, 4)
>>> x = Tensor(np.ones(shape=[2, 4, 3, 1]), mindspore.float32)
>>> y = Tensor(np.ones(shape=[2, 4, 3, 4]), mindspore.float32)
>>> batmatmul = ops.BatchMatMul(transpose_a=True)
>>> output = batmatmul(x, y)
>>> print(output.shape)
(2, 4, 1, 4)