shixian.shi
2023-06-27 6ad3397b716a3736510ad07170b4b8c3e3270857
debug
1个文件已修改
9 ■■■■ 已修改文件
funasr/datasets/large_datasets/utils/tokenize.py 9 ●●●● 补丁 | 查看 | 原始文档 | blame | 历史
funasr/datasets/large_datasets/utils/tokenize.py
@@ -57,7 +57,14 @@
        if hw_config['pre_hwlist'] is not None and hw_config['pre_prob'] > 0:
            # enable preset hotword detect in sampling
            import pdb; pdb.set_trace()
        hotword_indxs = sample_hotword(length, **hw_config)
            pre_index = None
            for hw in hw_config['pre_hwlist']:
                _find = text.find(hw)
                if _find != -1:
                    _find = text[:_find].count(" ")  # bpe sometimes
                    pre_index = [_find, _find + max(hw.count(" "), 1)]
                    break
        hotword_indxs = sample_hotword(length, **hw_config, pre_index=pre_index)
        data['hotword_indxs'] = hotword_indxs
        del data['hw_tag']
    for i in range(length):