“Zhiyuan Tang 2015-12-14”版本间的差异
来自cslt Wiki
（以“ Last week: 1. launched experiments on 1400h-Chinese plus 100h English with CTC/nnet3/Kaldi; 2. got refered ruselts of Kaldi/CTC on 1400h-Chinese[http://192.168.0....”为内容创建页面） |
|||
| 第2行： | 第2行： | ||
Last week: | Last week: | ||
| − | 1. launched experiments on 1400h-Chinese plus 100h English with CTC/nnet3/Kaldi; | + | 1. launched experiments on 1400h-Chinese plus 100h English with CTC/nnet3/Kaldi, it's much faster than those on train97 as it's on train88 with 8 jobs parallel; |
2. got refered ruselts of Kaldi/CTC on 1400h-Chinese[http://192.168.0.51:5555/cgi-bin/cvss/cvss_request.pl?account=tangzy&step=view_request&cvssid=492]; | 2. got refered ruselts of Kaldi/CTC on 1400h-Chinese[http://192.168.0.51:5555/cgi-bin/cvss/cvss_request.pl?account=tangzy&step=view_request&cvssid=492]; | ||
2015年12月14日 (一) 00:56的版本
Last week:
1. launched experiments on 1400h-Chinese plus 100h English with CTC/nnet3/Kaldi, it's much faster than those on train97 as it's on train88 with 8 jobs parallel;
2. got refered ruselts of Kaldi/CTC on 1400h-Chinese[1];
3. turned to SPEAKER.
This week:
1. more research about SPEAKER;
2. supervise the experiments;
3. reading.
TASKS
End-to-End
- monophone ASR --Zhiyuan
- MPE
- CTC/nnet3/Kaldi
conditioning learning
- language vector into multiple layers --Zhiyuan
- a Chinese paper
- speech rate into multiple layers --Zhiyuan
- verify the code for extra input(s) into DNN
