funasr/models/e2e_asr_contextual_paraformer.py
@@ -350,6 +350,7 @@ else: hw_embed = self.bias_embed(hw_list_pad) hw_embed, (h_n, _) = self.bias_encoder(hw_embed) hw_embed = h_n.repeat(encoder_out.shape[0], 1, 1) else: hw_lengths = [len(i) for i in hw_list] hw_list_pad = pad_list([torch.Tensor(i).long() for i in hw_list], 0).to(encoder_out.device)
