diff --git a/GPT_SoVITS/text/g2pw/onnx_api.py b/GPT_SoVITS/text/g2pw/onnx_api.py
index 374c9a4..32fc2c0 100644
--- a/GPT_SoVITS/text/g2pw/onnx_api.py
+++ b/GPT_SoVITS/text/g2pw/onnx_api.py
@@ -86,10 +86,10 @@ class G2PWOnnxConverter:
         sess_options.graph_optimization_level = onnxruntime.GraphOptimizationLevel.ORT_ENABLE_ALL
         sess_options.execution_mode = onnxruntime.ExecutionMode.ORT_SEQUENTIAL
         sess_options.intra_op_num_threads = 2
-        self.session_g2pW = onnxruntime.InferenceSession(
-            os.path.join(uncompress_path, 'g2pW.onnx'),
-            sess_options=sess_options, providers=['CUDAExecutionProvider', 'CPUExecutionProvider'])
-            # sess_options=sess_options)
+        try:
+            self.session_g2pW = onnxruntime.InferenceSession(os.path.join(uncompress_path, 'g2pW.onnx'),sess_options=sess_options, providers=['CUDAExecutionProvider', 'CPUExecutionProvider'])
+        except:
+            self.session_g2pW = onnxruntime.InferenceSession(os.path.join(uncompress_path, 'g2pW.onnx'),sess_options=sess_options, providers=['CPUExecutionProvider'])
         self.config = load_config(
             config_path=os.path.join(uncompress_path, 'config.py'),
             use_default=True)