Zhou Hao
Zhou Hao
新年快乐! 看到 #291 的issue,之前有过一个类似代码,于是今天参考GUI写了个流式输出的demo。 对`inference_webui.py`有2处改动: 1. `get_tts_wav` 增加一个stream参数(默认False不影响原逻辑),开启后按text推理bytes的结果; 2. 增加`if __name__ == '__main__':`避免从`inference_webui.py`引用函数时启动gradio app;
`hanlp-index`模式下,`全量池测试`得到一个很特殊的词组方式 例子(hanlp版本1.5.3): ```json POST _analyze { "analyzer" : "hanlp-index", "text": "搜索全量池测试" } { "tokens": [ { "token": "搜索", "start_offset": 0, "end_offset": 2, "type": "vn", "position": 0 }, { "token": "全量池测试",...
Fix following error on Windows: * [KeyError: ‘unexpected key “module.*](https://discuss.pytorch.org/t/solved-keyerror-unexpected-key-module-encoder-embedding-weight-in-state-dict/1686/3) ``` RuntimeError: Error(s) in loading state_dict for DSTformer: Missing key(s) in state_dict: "temp_embed", "pos_embed", "joints_embed.weight", ... Unexpected key(s) in state_dict:...