Fix test-runner-aot-cuda

Author: larryliu0820

.github/workflows/runner-cuda-dtype.yml

@@ -56,7 +56,7 @@ jobs:
         for DTYPE in bfloat16; do
             python torchchat.py generate --dtype ${DTYPE} --checkpoint-path ${MODEL_DIR}/stories15M.pt --temperature 0 --prompt "${PROMPT}" --device cuda
 
-            python torchchat.py export --checkpoint-path ${MODEL_DIR}/stories15M.pt --output-pt2-path /tmp/model.pt2
+            python torchchat.py export --checkpoint-path ${MODEL_DIR}/stories15M.pt --output-aoti-package-path /tmp/model.pt2
 
             ./cmake-out/aoti_run /tmp/model.pt2 -d CUDA -z ${MODEL_DIR}/tokenizer.model -i "${PROMPT}"