funasr/models/paraformer/model.py
@@ -475,6 +475,8 @@ speech = speech.to(device=kwargs["device"]) speech_lengths = speech_lengths.to(device=kwargs["device"]) # Encoder if kwargs.get("fp16", False): speech = speech.half() encoder_out, encoder_out_lens = self.encode(speech, speech_lengths) if isinstance(encoder_out, tuple): encoder_out = encoder_out[0]
