op {
  graph_op_name: "MatrixTriangularSolve"
  in_arg {
    name: "matrix"
    description: <<END
Shape is `[..., M, M]`.
END
  }
  in_arg {
    name: "rhs"
    description: <<END
Shape is `[..., M, K]`.
END
  }
  out_arg {
    name: "output"
    description: <<END
Shape is `[..., M, K]`.
END
  }
  attr {
    name: "lower"
    description: <<END
Boolean indicating whether the innermost matrices in `matrix` are
lower or upper triangular.
END
  }
  attr {
    name: "adjoint"
    description: <<END
Boolean indicating whether to solve with `matrix` or its (block-wise)
         adjoint.

@compatibility(numpy)
Equivalent to scipy.linalg.solve_triangular
@end_compatibility
END
  }
  summary: "Solves systems of linear equations with upper or lower triangular matrices by"
  description: <<END
backsubstitution.

`matrix` is a tensor of shape `[..., M, M]` whose inner-most 2 dimensions form
square matrices. If `lower` is `True` then the strictly upper triangular part
of each inner-most matrix is assumed to be zero and not accessed.
If `lower` is False then the strictly lower triangular part of each inner-most
matrix is assumed to be zero and not accessed.
`rhs` is a tensor of shape `[..., M, K]`.

The output is a tensor of shape `[..., M, K]`. If `adjoint` is
`True` then the innermost matrices in `output` satisfy matrix equations
`matrix[..., :, :] * output[..., :, :] = rhs[..., :, :]`.
If `adjoint` is `False` then the strictly then the  innermost matrices in
`output` satisfy matrix equations
`adjoint(matrix[..., i, k]) * output[..., k, j] = rhs[..., i, j]`.
END
}