“Qixin Wang 2016-04-25”版本间的差异
来自cslt Wiki
（以“ * Work done last week: ** New topic-many model with sequence LV model and constrain out YUN model ** Refactoring the code, pre-build model ** Accelerating the code,...”为内容创建页面） |
|||
| 第8行： | 第8行： | ||
** Made code run in WeChat (bugs and YUN LV problem solved) | ** Made code run in WeChat (bugs and YUN LV problem solved) | ||
| − | + | * some results (you can try it on WeChat now): | |
作诗5：花前留下 | 作诗5：花前留下 | ||
2016年4月25日 (一) 01:01的版本
- Work done last week:
- New topic-many model with sequence LV model and constrain out YUN model
- Refactoring the code, pre-build model
- Accelerating the code, return hidden layer, pre-calculate encoder, accelerate up to 10 times faster
- Random sample function
- Solved poly-phone problem, rebuilt the YUNLV list
- Made code run in WeChat (bugs and YUN LV problem solved)
- some results (you can try it on WeChat now):
作诗5：花前留下
红裙带翠低，露湿锦帘西。长步摇条柳，垂杨巧向堤。
玉梁映霓裳，美叶弄黄妆。翠盖飘香雾，披襟荡落光。
- Training the topic-many model for Song Iambic 15k
