diff --git a/src/operator/tensor/la_op-inl.h b/src/operator/tensor/la_op-inl.h
index 2f457abf7d7c..4dead87b3dce 100644
--- a/src/operator/tensor/la_op-inl.h
+++ b/src/operator/tensor/la_op-inl.h
@@ -813,7 +813,6 @@ struct syevd_backward {
 struct inverse_backward {
   template<typename xpu, typename DType>
   static void op(const Tensor<xpu, 3, DType>& dA,
-                 const Tensor<xpu, 3, DType>& B,
                  const Tensor<xpu, 3, DType>& A,
                  const Tensor<xpu, 3, DType>& dB,
                  const OpContext& ctx, const nnvm::NodeAttrs& attrs) {
diff --git a/src/operator/tensor/la_op.cc b/src/operator/tensor/la_op.cc
index 3d4e4f57d5a5..2fa1fd3a1cb2 100644
--- a/src/operator/tensor/la_op.cc
+++ b/src/operator/tensor/la_op.cc
@@ -926,18 +926,18 @@ Examples::
 .set_attr<FResourceRequest>("FResourceRequest", [](const NodeAttrs& attrs)
   { return std::vector<ResourceRequest>{ResourceRequest::kTempSpace}; })
 .set_attr<FCompute>("FCompute<cpu>", LaOpForward<cpu, 2, 2, 1, 1, inverse>)
-.set_attr<nnvm::FGradient>("FGradient", ElemwiseGradUseInOut{"_backward_linalg_inverse"})
+.set_attr<nnvm::FGradient>("FGradient", ElemwiseGradUseOut{"_backward_linalg_inverse"})
 .add_argument("A", "NDArray-or-Symbol", "Tensor of square matrix");
 
 NNVM_REGISTER_OP(_backward_linalg_inverse)
-.set_num_inputs(3)
+.set_num_inputs(2)
 .set_num_outputs(1)
 .set_attr<nnvm::FInplaceOption>("FInplaceOption", [](const NodeAttrs& attrs)
   { return std::vector<std::pair<int, int> >{{0, 0}}; })
 .set_attr<FResourceRequest>("FResourceRequest", [](const NodeAttrs& attrs)
   { return std::vector<ResourceRequest>{ResourceRequest::kTempSpace}; })
 .set_attr<nnvm::TIsBackward>("TIsBackward", true)
-.set_attr<FCompute>("FCompute<cpu>", LaOpBackward<cpu, 2, 2, 3, 1, inverse_backward>);
+.set_attr<FCompute>("FCompute<cpu>", LaOpBackward<cpu, 2, 2, 2, 1, inverse_backward>);
 
 }  // namespace op
 }  // namespace mxnet