Added loop-no-unroll tag to both mode of frontend

(we'll come back to solve the real problem later)

hpvm/projects/torch2hpvm/torch2hpvm/template_hpvm.cpp.in

@@ -89,7 +89,7 @@ int main(int argc, char *argv[]){
 
   __hpvm__init();
   startMemTracking();
-  #pragma clang loop unroll(disable)
+#pragma clang loop unroll(disable)
   for (int i = 0; i < batch_count; i++){
     int start = i * batch_size, end = start + batch_size;
     void *{{input_name}} = readInputBatch(input_path.c_str(), 0, start, end, {{input_shape|join(', ')}});

hpvm/projects/torch2hpvm/torch2hpvm/template_hpvm_inspect.cpp.in

@@ -140,6 +140,7 @@ int main(){
     // Keep this open so the other side knows we have more batches to write
     auto* fp = open_fifo("{{fifo_path_w}}", "wb");
     float total_accuracy = 0;
+#pragma clang loop unroll(disable)
     for (int i = 0; i < batch_count; i++){
       int start = i * batch_size, end = start + batch_size;
       void *{{input_name}} = readInputBatch(input_pth, 0, start, end, {{input_shape|join(', ')}});