diff options
| author |  RJ Ryan <rjryan@google.com> | 2017-10-05 10:03:12 -0700 |
|---|---|---|
| committer |  TensorFlower Gardener <gardener@tensorflow.org> | 2017-10-05 10:14:21 -0700 |
| commit | 376147cd71d1a240dad428c3ff82ca4ea5f4e88e (patch) | |
| tree | b3cedfceb8cec23d1a32a5521fd576f0e5c3580d /tensorflow/cc/gradients | |
| parent | c49eeeee5463aff02b4bafbd1596288ba4b27739 (diff) | |
Save an unnecessary logical_not in the maximum/minimum gradient.
PiperOrigin-RevId: 171167415
Diffstat (limited to 'tensorflow/cc/gradients')
| -rw-r--r-- | tensorflow/cc/gradients/math_grad.cc | 2 |
1 files changed, 1 insertions, 1 deletions
diff --git a/tensorflow/cc/gradients/math_grad.cc b/tensorflow/cc/gradients/math_grad.cc index ac288b1d83..2417bf18a9 100644 --- a/tensorflow/cc/gradients/math_grad.cc +++ b/tensorflow/cc/gradients/math_grad.cc @@ -484,7 +484,7 @@ Status MaximumMinimumGradCommon(const Scope& scope, const Operation& op, auto grad = grad_inputs[0]; auto zeros = ZerosLike(scope, grad); auto gx_1 = Where3(scope, comparator, grad, zeros); - auto gx_2 = Where3(scope, LogicalNot(scope, comparator), grad, zeros); + auto gx_2 = Where3(scope, comparator, zeros, grad); return BinaryGradCommon(scope, op, grad_outputs, gx_1, gx_2); } |