Bug fixed.

Fix the GPU bug in Clip function.

source/tensor/core/math/Clip.cu

@@ -64,7 +64,6 @@ This is for float16 computation
 __global__
 void KernelClip(__half * a, __half * b, DTYPE lower, DTYPE upper, int size)
 {
-    ShowNTErrors("TODO!");
     return;
 }
 
@@ -101,7 +100,7 @@ void _CudaClip(const XTensor * a, XTensor * b, DTYPE lower, DTYPE upper)
         KernelClip<int> << <blocks, threads >> >((int *)a->data, (int *)b->data, lower1, upper1, a->unitNum);
     }
     else if (a->dataType == X_FLOAT16) {
-        KernelClip << <blocks, threads >> >((__half*)a->data, (__half*)b->data, lower, upper, a->unitNum);
+        ShowNTErrors("TODO!");
     }
     else {
         ShowNTErrors("TODO!");