funasr/models/lcbnet/model.py
@@ -444,7 +444,7 @@ encoder_out, encoder_out_lens = self.encode(speech, speech_lengths) if isinstance(encoder_out, tuple): encoder_out = encoder_out[0] pdb.set_trace() ocr_list_new = [[x + 1 if x != 0 else x for x in sublist] for sublist in ocr_sample_list] ocr = torch.tensor(ocr_list_new).to(device=kwargs["device"]) ocr_lengths = ocr.new_full([1], dtype=torch.long, fill_value=ocr.size(1)).to(device=kwargs["device"])
