feature add Neo image uri config for Pytorch 1.12

Author: HappyAmazonian

tests/data/pytorch_neo/code/inference.py

@@ -71,8 +71,8 @@ def model_fn(model_dir):
     logger.info("model_fn")
     neopytorch.config(model_dir=model_dir, neo_runtime=True)
     device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
-    # The compiled model is saved as "model.pth"
-    model = torch.jit.load(os.path.join(model_dir, "model.pth"), map_location=device)
+    # The compiled model is saved as "model.pth" or "model.pt"
+    model = torch.jit.load(os.path.join(model_dir, "model.pt"), map_location=device) 
 
     # It is recommended to run warm-up inference during model load
     sample_input_path = os.path.join(model_dir, "sample_input.pkl")