FUNASR训练
parent: 6e5f075b | 补丁 | 提交 | show whitespace
hnluo
2023-02-09 1bae17a7a9421342ead4eb2d5579d52ef69dbdbc 
Merge pull request #82 from alibaba-damo-academy/dev_lhn

fix uniasr training bug
1个文件已修改
2 ■■■■■ 已修改文件
funasr/models/e2e_uni_asr.py 2 ●●●●● 补丁 | 查看 | 原始文档 | blame | 历史 
 funasr/models/e2e_uni_asr.py


@@ -206,7 +206,7 @@


            with torch.no_grad():


                speech_raw, encoder_out, encoder_out_lens = self.encode(speech, speech_lengths, ind=ind)


        else:


            speech_raw, encoder_out_lens = self.encode(speech, speech_lengths, ind=ind)


            speech_raw, encoder_out, encoder_out_lens = self.encode(speech, speech_lengths, ind=ind)




        intermediate_outs = None


        if isinstance(encoder_out, tuple):