diff --git a/funasr/models/extract_tokens/model.py b/funasr/models/extract_tokens/model.py
index e383f0d44..faf088bad 100644
--- a/funasr/models/extract_tokens/model.py
+++ b/funasr/models/extract_tokens/model.py
@@ -1082,7 +1082,7 @@ class SenseVoiceQuantizedEncoderPitch(nn.Module):
         self.normalized_quant_input = normalized_quant_input
         self.quantizer = self.build_quantizer(quantizer_config)
         
-        self.pitch_predictor = torch.Linear(units, 1)
+        self.pitch_predictor = torch.nn.Linear(units, 1)
         self.pitch_act = torch.nn.ReLU()
 
     def build_quantizer(self, vq_config):