The model weights and language model come from Baidu. I upload them in my repo for convenient purpose.
The, language model is too big to upload as one file. After download all files for language model, you can use '''cat english_lm.tar.gz.parta* > english_lm.tar.gz''' to reconstruct the complete file.