Update README.md

hnluo

2023-04-24 9b1cd2e89fb2471ba073b87ea7ce9dde2b6fc61d

Update README.md

 egs_modelscope/asr/TEMPLATE/README.md

@@ -176,6 +176,21 @@
    - `max_epoch`: number of training epoch
    - `lr`: learning rate

- Training data formats：
```sh
cat ./example_data/text
BAC009S0002W0122 而 对 楼 市 成 交 抑 制 作 用 最 大 的 限 购
BAC009S0002W0123 也 成 为 地 方 政 府 的 眼 中 钉
english_example_1 hello world
english_example_2 go swim 去 游 泳

cat ./example_data/wav.scp
BAC009S0002W0122 /mnt/data/wav/train/S0002/BAC009S0002W0122.wav
BAC009S0002W0123 /mnt/data/wav/train/S0002/BAC009S0002W0123.wav
english_example_1 /mnt/data/wav/train/S0002/english_example_1.wav
english_example_2 /mnt/data/wav/train/S0002/english_example_2.wav
```

- Then you can run the pipeline to finetune with:
```shell
python finetune.py
@@ -210,4 +225,4 @@
    --njob 64 \
    --checkpoint_dir "./checkpoint" \
    --checkpoint_name "valid.cer_ctc.ave.pb"
```
```