- torch.linalg.lu_factor(A, *, bool pivot=True, out=None) -> (Tensor, Tensor)¶
Computes a compact representation of the LU factorization with partial pivoting of a matrix.
This function computes a compact representation of the decomposition given by
torch.linalg.lu(). If the matrix is square, this representation may be used in
torch.linalg.lu_solve()to solve system of linear equations that share the matrix
The returned decomposition is represented as a named tuple (LU, pivots). The
LUmatrix has the same shape as the input matrix
A. Its upper and lower triangular parts encode the non-constant elements of
Uof the LU decomposition of
The returned permutation matrix is represented by a 1-indexed vector. pivots[i] == j represents that in the i-th step of the algorithm, the i-th row was permuted with the j-1-th row.
On CUDA, one may use
pivot= False. In this case, this function returns the LU decomposition without pivoting if it exists.
Supports inputs of float, double, cfloat and cdouble dtypes. Also supports batches of matrices, and if the inputs are batches of matrices then the output has the same batch dimensions.
When inputs are on a CUDA device, this function synchronizes that device with the CPU. For a version of this function that does not synchronize, see
The LU decomposition is almost never unique, as often there are different permutation matrices that can yield different LU decompositions. As such, different platforms, like SciPy, or inputs on different devices, may produce different valid decompositions.
Gradient computations are only supported if the input matrix is full-rank. If this condition is not met, no error will be thrown, but the gradient may not be finite. This is because the LU decomposition with pivoting is not differentiable at these points.
torch.linalg.lu_solve()solves a system of linear equations given the output of this function provided the input matrix was square and invertible.
A (Tensor) – tensor of shape (*, m, n) where * is zero or more batch dimensions.
- Keyword Arguments:
A named tuple (LU, pivots).
RuntimeError – if the
Amatrix is not invertible or any matrix in a batched
Ais not invertible.
>>> A = torch.randn(2, 3, 3) >>> B1 = torch.randn(2, 3, 4) >>> B2 = torch.randn(2, 3, 7) >>> A_factor = torch.linalg.lu_factor(A) >>> X1 = torch.linalg.lu_solve(A_factor, B1) >>> X2 = torch.linalg.lu_solve(A_factor, B2) >>> torch.allclose(A @ X1, B1) True >>> torch.allclose(A @ X2, B2) True