add text to pinyin model as tts frontend #3506
Conversation
|
Thanks for your contribution! |
|
|
||
| Predict: | ||
| batch_size: 1 | ||
| input: "欢迎使用飞桨" |
There was a problem hiding this comment.
这里建议使用 txt文件路径，文字放到txt文件中。这样和其他模型保持统一。
| Yields: | ||
| list: list of file path. | ||
| """ | ||
| if isinstance(inputs, str): |
There was a problem hiding this comment.
建议在现有基础上，额外支持 str 为本地txt文件路径。
| assert len(simplified_charcters) == len(simplified_charcters) | ||
|
|
||
| s2t_dict = {} | ||
| t2s_dict = {} | ||
| for i, item in enumerate(simplified_charcters): | ||
| s2t_dict[item] = traditional_characters[i] | ||
| t2s_dict[traditional_characters[i]] = item |
There was a problem hiding this comment.
这段代码放到tranditional_to_simplified()中吧
| opencc | ||
| onnxruntime | ||
| pypinyin |
There was a problem hiding this comment.
按照最新的paddlex第三方依赖规范，这三个包似乎属于【三类特殊依赖】，应该放到setup.py中的EXTRAS中。且最新规范中，requirements.txt已经废弃不再使用。具体请参考：
https://ku.baidu-int.com/knowledge/HFVrC7hq1Q/yKeL8Lljko/YkH5mORwJ3/X1tDrwm1qSK2Zx
| from typing import Tuple | ||
|
|
||
| import numpy as np | ||
| import onnxruntime |
There was a problem hiding this comment.
依据最新的paddlex第三方依赖管理规范，onnxruntime 应该不是默认安装的依赖库，不应在顶层import，并考虑使用function_requires_deps修饰相关调用函数。具体请参考：https://ku.baidu-int.com/knowledge/HFVrC7hq1Q/yKeL8Lljko/YkH5mORwJ3/X1tDrwm1qSK2Zx
| "fix_mode": None, | ||
| "count_json": "train.count.json", | ||
| }, | ||
| "lr": 5e-5, |
There was a problem hiding this comment.
这个配置文件是做什么的呢？为什么还有学习率配置
| ) | ||
| elif char in self.char_bopomofo_dict: | ||
| partial_result[i] = pypinyin_result[i][0] | ||
| # partial_result[i] = self.style_convert_func(self.char_bopomofo_dict[char][0]) |
There was a problem hiding this comment.
这个tokenizer的改动也是和这个模型相关的吗？
No description provided.