mutli train

[Yymax-max]

if i want to use open data and own data to train a zh model ,how i can do to generate a mainfest for train.can get me some example or some tips ? thank you

[yfyeung]

First, you need to add some recipes in lhotse to support your own corpus. You can refer the following PR in lhotse:
lhotse-speech/lhotse#996

Second, you need to add a recipe prepare.sh, using lhotse to prepare manifests for you data, just like https://github.com/k2-fsa/icefall/blob/master/egs/librispeech/ASR/prepare.sh

Third, you need to write a recipe asr_datamodule.py to generate dataloader for train & dev & test, just like:
https://github.com/k2-fsa/icefall/blob/master/egs/librispeech/ASR/tdnn_lstm_ctc/asr_datamodule.py