mindspore.ops.matrix_band_part
- mindspore.ops.matrix_band_part(x, lower, upper)[source]
Copy a tensor setting everything outside a central band in each innermost matrix to zero.
- Parameters
x (Tensor) – Input tensor. \((*, m, n)\) where \(*\) means, any number of additional dimensions. The data type must be float16, float32, float64, int32 or int64.
lower (Union[int, Tensor]) – Number of subdiagonals to keep. The data type must be int32 or int64. If negative, keep entire lower triangle.
upper (Union[int, Tensor]) – Number of superdiagonals to keep. The data type must be int32 or int64. If negative, keep entire upper triangle.
- Returns
Tensor, has the same type and shape as x.
- Raises
TypeError – If x is not a Tensor.
TypeError – If dtype of x is not one of float16, float32, float64, int32 or int64.
TypeError – If lower is neither a number nor a Tensor.
TypeError – If upper is neither a number nor a Tensor.
TypeError – If dtype of lower is neither int32 nor int64.
TypeError – If dtype of upper is neither int32 nor int64.
ValueError – If the shape of x is not greater than or equal to 2D.
ValueError – If the shape of lower is not equal to 0D.
ValueError – If the shape of upper is not equal to 0D.
- Supported Platforms:
GPU
CPU
Examples
>>> x = Tensor(np.ones([2, 4, 4]).astype(np.float32)) >>> output = ops.matrix_band_part(x, 2, 1) >>> print(output) [[[1. 1. 0. 0.] [1. 1. 1. 0.] [1. 1. 1. 1.] [0. 1. 1. 1.]] [[1. 1. 0. 0.] [1. 1. 1. 0.] [1. 1. 1. 1.] [0. 1. 1. 1.]]]