CLN: rename UnsafeDiv => DivNoNan

2 files changed, 11 insertions, 12 deletions

diff --git a/tensorflow/cc/gradients/math_grad.cc b/tensorflow/cc/gradients/math_grad.cc
index c6e60689fa..cd215f740d 100644
--- a/tensorflow/cc/gradients/math_grad.cc
+++ b/tensorflow/cc/gradients/math_grad.cc
@@ -441,21 +441,20 @@ Status RealDivGrad(const Scope& scope, const Operation& op,
 }
 REGISTER_GRADIENT_OP("RealDiv", RealDivGrad);
 
-Status UnsafeDivGrad(const Scope& scope, const Operation& op,
-                     const std::vector<Output>& grad_inputs,
-                     std::vector<Output>* grad_outputs) {
+Status DivNoNanGrad(const Scope& scope, const Operation& op,
+                    const std::vector<Output>& grad_inputs,
+                    std::vector<Output>* grad_outputs) {
   auto x_1 = ConjugateHelper(scope, op.input(0));
   auto x_2 = ConjugateHelper(scope, op.input(1));
   // y = x_1 / x_2
   // dy/dx_1 = 1/x_2
   // dy/dx_2 = -x_1/x_2^2
-  auto gx_1 = UnsafeDiv(scope, grad_inputs[0], x_2);
-  auto gx_2 =
-      Mul(scope, grad_inputs[0],
-          UnsafeDiv(scope, UnsafeDiv(scope, Neg(scope, x_1), x_2), x_2));
+  auto gx_1 = DivNoNan(scope, grad_inputs[0], x_2);
+  auto gx_2 = Mul(scope, grad_inputs[0],
+                  DivNoNan(scope, DivNoNan(scope, Neg(scope, x_1), x_2), x_2));
   return BinaryGradCommon(scope, op, grad_outputs, gx_1, gx_2);
 }
-REGISTER_GRADIENT_OP("UnsafeDiv", UnsafeDivGrad);
+REGISTER_GRADIENT_OP("DivNoNan", DivNoNanGrad);
 
 Status SquaredDifferenceGrad(const Scope& scope, const Operation& op,
                              const std::vector<Output>& grad_inputs,
diff --git a/tensorflow/cc/gradients/math_grad_test.cc b/tensorflow/cc/gradients/math_grad_test.cc
index 12a19bcf28..147428cc39 100644
--- a/tensorflow/cc/gradients/math_grad_test.cc
+++ b/tensorflow/cc/gradients/math_grad_test.cc
@@ -33,6 +33,7 @@ using ops::AddN;
 using ops::BatchMatMul;
 using ops::Const;
 using ops::Div;
+using ops::DivNoNan;
 using ops::MatMul;
 using ops::Max;
 using ops::Maximum;
@@ -47,7 +48,6 @@ using ops::RealDiv;
 using ops::SquaredDifference;
 using ops::Sub;
 using ops::Sum;
-using ops::UnsafeDiv;
 using ops::Where3;
 
 // TODO(andydavis) Test gradient function against numeric gradients output.
@@ -854,13 +854,13 @@ TEST_F(NaryGradTest, RealDiv) {
   RunTest({x}, {x_shape}, {y}, {x_shape});
 }
 
-TEST_F(NaryGradTest, UnsafeDiv) {
+TEST_F(NaryGradTest, DivNoNan) {
   {
     TensorShape x_shape({3, 2, 5});
     const auto x = Placeholder(scope_, DT_FLOAT, Placeholder::Shape(x_shape));
     // Test x / (1 + |x|) rather than x_1 / x_2 to avoid triggering large
     // division errors in the numeric estimator used by the gradient checker.
-    const auto y = UnsafeDiv(
+    const auto y = DivNoNan(
         scope_, x, Add(scope_, Const<float>(scope_, 1), Abs(scope_, x)));
     RunTest({x}, {x_shape}, {y}, {x_shape});
   }
@@ -868,7 +868,7 @@ TEST_F(NaryGradTest, UnsafeDiv) {
     // Return 0 gradient (rather than NaN) for division by zero.
     const auto x = Placeholder(scope_, DT_FLOAT);
     const auto zero = Const<float>(scope_, 0.0);
-    const auto y = UnsafeDiv(scope_, x, zero);
+    const auto y = DivNoNan(scope_, x, zero);
 
     std::vector<Output> grad_outputs;
     TF_EXPECT_OK(AddSymbolicGradients(scope_, {y}, {x}, &grad_outputs));