From dc13e5ff6fcb8b27e0654b3e5e4c35083ac72456 Mon Sep 17 00:00:00 2001 From: Sami Kama Date: Wed, 15 May 2019 12:52:17 -0700 Subject: [PATCH] Fix ops_testutil GetOutput by correctly copying the GPU tensor from device to host --- tensorflow/core/kernels/ops_testutil.cc | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/tensorflow/core/kernels/ops_testutil.cc b/tensorflow/core/kernels/ops_testutil.cc index 7aa7d1a5861..42947190027 100644 --- a/tensorflow/core/kernels/ops_testutil.cc +++ b/tensorflow/core/kernels/ops_testutil.cc @@ -53,7 +53,7 @@ Tensor* OpsTestBase::GetOutput(int output_index) { new Tensor(allocator(), output->dtype(), output->shape()); auto src = output->tensor_data(); auto dst = managed_output->tensor_data(); - context_->eigen_gpu_device().memcpy(const_cast(dst.data()), + context_->eigen_gpu_device().memcpyDeviceToHost(const_cast(dst.data()), src.data(), src.size()); context_->eigen_gpu_device().synchronize(); managed_outputs_[output_index] = managed_output;