Fixing Memory Free Bug

llvm/projects/hpvm-tensor-rt/tensor_runtime/include/tensor_utils.cu

@@ -410,7 +410,8 @@ void convertToFP16(struct Tensor* tensor){
   
   size_t size_in_bytes = tensor->size_in_bytes / 2;  
   checkCudaErrors(cudaMalloc(&tensor->gpu_half_data, size_in_bytes)); // Allocate memory on GPU
-
+  tensors_ptr.push_back(tensor->gpu_half_data);
+  
   f2h((float*) tensor->gpu_data, tensor->num_elems, (half*) tensor->gpu_half_data);
 
 }