Created by: misbullah

Hi,

Today, I did evaluation using different models (AM and LM). Plese see the following table.

I reduce sampling rate to be 8000 Hz when doing test for BaiduCN1.2k model because I think the model was trained using 8000 Hz dataset.

Question:

1. Why I get very worse result for BaiduCN1.2K? (As we know the model was trained using larger training data than AISHELL and the model parameter also larger than AISHELL model?
2. How to use tune.py to find good tuning parameter for alpha and beta? I cannot find any examples for the python code?

Thanks, Alim