diff --git a/script/app-mlperf-inference-nvidia/customize.py b/script/app-mlperf-inference-nvidia/customize.py
index da3abae38..bbb9cb703 100644
--- a/script/app-mlperf-inference-nvidia/customize.py
+++ b/script/app-mlperf-inference-nvidia/customize.py
@@ -166,7 +166,8 @@ def preprocess(i):
 
         if not os.path.exists(fp32_model_path):
             env['CM_REQUIRE_GPTJ_MODEL_DOWNLOAD'] = 'yes' # download via prehook_deps
-            cmds.append(f"cp -r $CM_ML_MODEL_FILE_WITH_PATH {fp32_model_path}")
+            if make_command == "build_engine":
+                cmds.append(f"cp -r $CM_ML_MODEL_FILE_WITH_PATH {fp32_model_path}")
 
         model_name = "gptj"
         model_path = fp8_model_path