Fix Model with nn.Module

Signed-off-by: Jael Gu <mengjia.gu@zilliz.com>

nn_fingerprint.py

@@ -20,6 +20,7 @@ from pathlib import Path
 from typing import List, Union
 
 import torch
+from torch import nn
 import torchaudio
 import numpy
 
@@ -35,8 +36,9 @@ warnings.filterwarnings('ignore')
 log = logging.getLogger('nnfp_op')
 
 
-class Model:
+class Model(nn.Module):
     def __init__(self, model_name, device='cpu', model_path=None):
+        super().__init__()
         self.device = device
         if model_name == 'nnfp_default':
             self.params = default_params
@@ -76,8 +78,8 @@ class Model:
         self.model.eval()
         log.info('Model is loaded.')
 
-    def __call__(self, data: 'Tensor'):
-        return self.model(data)
+    def __call__(self, *args, **kwargs):
+        return self.model(*args, **kwargs)
 
 
 @register(output_schema=['vecs'])
@@ -205,7 +207,7 @@ class NNFingerprint(NNOperator):
         elif format == 'onnx':
             path = path + '.onnx'
             try:
-                torch.onnx.export(self.model.model,
+                torch.onnx.export(self.model,
                                   dummy_input,
                                   path,
                                   input_names=['input'],