diff options
| author |  A. Unique TensorFlower <gardener@tensorflow.org> | 2017-10-09 17:49:32 -0700 |
|---|---|---|
| committer | TensorFlower Gardener <gardener@tensorflow.org> | 2017-10-09 17:53:50 -0700 |
| commit | 2cdd0647e08c1dc7948f70416ee8311c09598e59 (patch) | |
| tree | 846d501f81afc77da2ee96b9c86eeeac5cddcf9d /tensorflow/contrib/fused_conv | |
| parent | 07d78ddeafe41bc0363ac92efd7ca8ea60478989 (diff) | |
Make error message more explicit when running FusedConv2DBiasActivationOp with type int8 on a GPU that doesn't support it.
Old error message: "No algorithm worked!"
New error message: "FusedConv2DBiasActivation is only supported on GPUs with compute capability 6.1 or later."
PiperOrigin-RevId: 171614032
Diffstat (limited to 'tensorflow/contrib/fused_conv')
| -rw-r--r-- | tensorflow/contrib/fused_conv/kernels/fused_conv2d_bias_activation_op.cc | 11 |
1 files changed, 11 insertions, 0 deletions
diff --git a/tensorflow/contrib/fused_conv/kernels/fused_conv2d_bias_activation_op.cc b/tensorflow/contrib/fused_conv/kernels/fused_conv2d_bias_activation_op.cc index 256f200868..e4c39739f7 100644 --- a/tensorflow/contrib/fused_conv/kernels/fused_conv2d_bias_activation_op.cc +++ b/tensorflow/contrib/fused_conv/kernels/fused_conv2d_bias_activation_op.cc @@ -298,6 +298,17 @@ void LaunchFusedConv2DBiasActivationOp<GPUDevice, T, BiasType, ScaleType>:: constexpr int rank = is_int8x4 ? 5 : 4; constexpr int vect = is_int8x4 ? 4 : 1; + if (is_int8x4) { + int cc_major, cc_minor; + stream->parent()->GetDeviceDescription().cuda_compute_capability(&cc_major, + &cc_minor); + OP_REQUIRES( + ctx, cc_major >= 6 && cc_minor >= 1, + errors::Unimplemented( + "FusedConv2DBiasActivation for int8 is only supported on GPUs with " + "compute capability 6.1 or later.")); + } + const int batch_size = GetTensorDim(conv_input_param, data_format, 'N'); int conv_input_rows = GetTensorDim(conv_input_param, data_format, 'H'); int conv_input_cols = GetTensorDim(conv_input_param, data_format, 'W'); |