adding ROCm support for the softplus_op
This commit is contained in:
parent
13ac37f915
commit
017b6bb043
@ -87,7 +87,7 @@ void SoftplusGradOp<Device, T>::OperateNoTemplate(OpKernelContext* context,
|
||||
TF_CALL_FLOAT_TYPES(REGISTER_KERNELS);
|
||||
#undef REGISTER_KERNELS
|
||||
|
||||
#if GOOGLE_CUDA
|
||||
#if GOOGLE_CUDA || TENSORFLOW_USE_ROCM
|
||||
// Forward declarations of the functor specializations for GPU.
|
||||
namespace functor {
|
||||
#define DECLARE_GPU_SPEC(T) \
|
||||
@ -119,6 +119,6 @@ TF_CALL_GPU_NUMBER_TYPES(DECLARE_GPU_SPEC);
|
||||
TF_CALL_GPU_NUMBER_TYPES(REGISTER_GPU_KERNELS);
|
||||
#undef REGISTER_GPU_KERNELS
|
||||
|
||||
#endif // GOOGLE_CUDA
|
||||
#endif // GOOGLE_CUDA || TENSORFLOW_USE_ROCM
|
||||
|
||||
} // namespace tensorflow
|
||||
|
@ -13,7 +13,7 @@ See the License for the specific language governing permissions and
|
||||
limitations under the License.
|
||||
==============================================================================*/
|
||||
|
||||
#if GOOGLE_CUDA
|
||||
#if GOOGLE_CUDA || TENSORFLOW_USE_ROCM
|
||||
|
||||
#define EIGEN_USE_GPU
|
||||
|
||||
@ -37,4 +37,4 @@ TF_CALL_GPU_NUMBER_TYPES(DEFINE_GPU_KERNELS);
|
||||
|
||||
} // end namespace tensorflow
|
||||
|
||||
#endif // GOOGLE_CUDA
|
||||
#endif // GOOGLE_CUDA || TENSORFLOW_USE_ROCM
|
||||
|
Loading…
Reference in New Issue
Block a user