add nvcc to pytorch path

Author: whoisj

qa/common/gen_qa_model_repository

@@ -155,6 +155,7 @@ chmod -R 777 $TRITON_MDLS_QA_RESHAPE_MODEL
 # chmod -R 777 $SVOLUME_EQDESTDIR
 # python3 $TRITON_MDLS_SRC_DIR/gen_qa_dyna_sequence_models.py --openvino --models_dir=$TRITON_MDLS_QA_DYNA_SEQUENCE_MODEL
 # chmod -R 777 $TRITON_MDLS_QA_DYNA_SEQUENCE_MODEL
+exit 0
 EOF
 
 log_message.status "run: chmod a+x ${OPENVINOSCRIPT}"
@@ -218,6 +219,7 @@ python3 $TRITON_MDLS_SRC_DIR/gen_qa_identity_models.py --ensemble --models_dir=$
 python3 $TRITON_MDLS_SRC_DIR/gen_qa_sequence_models.py --ensemble --models_dir=$TRITON_MDLS_QA_ENSEMBLE_MODEL/qa_sequence_model_repository
 python3 $TRITON_MDLS_SRC_DIR/gen_qa_sequence_models.py --ensemble --variable --models_dir=$TRITON_MDLS_QA_ENSEMBLE_MODEL/qa_variable_sequence_model_repository
 chmod -R 777 $TRITON_MDLS_QA_ENSEMBLE_MODEL
+exit 0
 EOF
 
 log_message.status "run: chmod a+x ${ONNXSCRIPT}"
@@ -237,6 +239,7 @@ nvidia-smi || true
 pip3 install onnxscript
 set -e
 set -x
+PATH=$PATH:/usr/local/cuda-13.0/bin
 python3 $TRITON_MDLS_SRC_DIR/gen_qa_models.py --libtorch --models_dir=$TRITON_MDLS_QA_MODEL
 python3 $TRITON_MDLS_SRC_DIR/gen_qa_models.py --torch-aoti --models_dir=$TRITON_MDLS_QA_MODEL
 chmod -R 777 $TRITON_MDLS_QA_MODEL
@@ -272,6 +275,7 @@ python3 $TRITON_MDLS_SRC_DIR/gen_qa_custom_ops_models.py --libtorch --models_dir
 mkdir -p $TRITON_MDLS_QA_CUSTOM_OPS/libtorch_modulo/
 cp \${TORCH_EXTENSIONS_DIR}/custom_modulo/custom_modulo.so $TRITON_MDLS_QA_CUSTOM_OPS/libtorch_modulo/.
 chmod -R 777 $TRITON_MDLS_QA_CUSTOM_OPS
+exit 0
 EOF
 
 log_message.status "run: chmod a+x ${TORCHSCRIPT}"
@@ -337,6 +341,7 @@ rm -rf build && mkdir build && \
 cd build && cmake .. && make -j && cp libcustomHardmaxPlugin.so $TRITON_MDLS_QA_TRT_PLUGIN_MODEL/.
 LD_PRELOAD=$TRITON_MDLS_QA_TRT_PLUGIN_MODEL/libcustomHardmaxPlugin.so python3 $TRITON_MDLS_SRC_DIR/gen_qa_trt_plugin_models.py --models_dir=$TRITON_MDLS_QA_TRT_PLUGIN_MODEL
 chmod -R 777 $TRITON_MDLS_QA_TRT_PLUGIN_MODEL
+exit 0
 EOF
 
 log_message.status "run: chmod a+x ${TRTSCRIPT}"
@@ -428,6 +433,7 @@ if which docker ; then
             --label PROJECT_NAME=$PROJECT_NAME \
             $DOCKER_GPU_ARGS \
             -v $DOCKER_VOLUME:/mnt \
+            -t \
             $UBUNTU_IMAGE \
             bash -xe $TRITON_MDLS_SRC_DIR/$OPENVINOSCRIPT
 
@@ -444,6 +450,7 @@ if which docker ; then
         --label PROJECT_NAME=$PROJECT_NAME \
         $DOCKER_GPU_ARGS \
         -v $DOCKER_VOLUME:/mnt \
+        -t \
         $UBUNTU_IMAGE \
         bash -xe $TRITON_MDLS_SRC_DIR/$ONNXSCRIPT
 
@@ -462,6 +469,7 @@ if which docker ; then
         --label PROJECT_NAME=$PROJECT_NAME \
         $DOCKER_GPU_ARGS \
         -v $DOCKER_VOLUME:/mnt \
+        -t \
         $PYTORCH_IMAGE \
         bash -xe $TRITON_MDLS_SRC_DIR/$TORCHSCRIPT
 
@@ -481,6 +489,7 @@ if which docker ; then
             --label PROJECT_NAME=$PROJECT_NAME \
             $DOCKER_GPU_ARGS \
             -v $DOCKER_VOLUME:/mnt \
+            -t \
             -e TRT_VERBOSE \
             $TENSORRT_IMAGE \
             bash -xe $TRITON_MDLS_SRC_DIR/$TRTSCRIPT

qa/common/gen_qa_models.py

@@ -1336,6 +1336,7 @@ def forward(self, INPUT0, INPUT1):
         AddSubNet(swap), (torch.randn(*input_shape), torch.randn(*input_shape))
     )
     torch._inductor.aoti_compile_and_package(ep, package_path=model_version_dir + "/model.pt2")
+    torch._inductor.aoti_load_package
 
 
 def create_libtorch_modelconfig(