diff options
| author |  A. Unique TensorFlower <gardener@tensorflow.org> | 2018-05-15 13:38:34 -0700 |
|---|---|---|
| committer | TensorFlower Gardener <gardener@tensorflow.org> | 2018-05-15 13:40:56 -0700 |
| commit | 12ea999991f53a4d1cecdd1816bfceff6da13ae4 (patch) | |
| tree | c98ca1d5b40f0404b595558e59791a33931bac53 /tensorflow/core/kernels/depthwise_conv_op.cc | |
| parent | 4d91734ca1e400a7fef012e75d2590377583261c (diff) | |
Don't ever use cuDNN to perform depthwise convolutions on CPU.
PiperOrigin-RevId: 196721302
Diffstat (limited to 'tensorflow/core/kernels/depthwise_conv_op.cc')
| -rw-r--r-- | tensorflow/core/kernels/depthwise_conv_op.cc | 2 |
1 files changed, 1 insertions, 1 deletions
diff --git a/tensorflow/core/kernels/depthwise_conv_op.cc b/tensorflow/core/kernels/depthwise_conv_op.cc index d5f4a68120..f0902fdba6 100644 --- a/tensorflow/core/kernels/depthwise_conv_op.cc +++ b/tensorflow/core/kernels/depthwise_conv_op.cc @@ -290,7 +290,7 @@ class DepthwiseConv2dNativeOp : public BinaryOp<T> { OP_REQUIRES_OK(context, context->GetAttr("padding", &padding_)); // For in_depth == 1 and grouped convolutions. - use_cudnn_ = CanUseCudnn(); + use_cudnn_ = CanUseCudnn() && std::is_same<Device, GPUDevice>::value; cudnn_use_autotune_ = CudnnUseAutotune(); use_cudnn_grouped_conv_ = false; dtype_ = DataTypeToEnum<T>::value; |