paraformer-speaker inference pipeline

shixian.shi

2023-10-10 8a0930d682fe3206e0b41c694fc03d7d10c7eed2

 funasr/export/models/encoder/conformer_encoder.py

@@ -61,7 +61,6 @@
                speech: torch.Tensor,
                speech_lengths: torch.Tensor,
                ):
        speech = speech * self._output_size ** 0.5
        mask = self.make_pad_mask(speech_lengths)
        mask = self.prepare_mask(mask)
        if self.embed is None: