some questions about lm config file

[tianwengang]

I hope to reproduce the results in this great paper: “Improved training of end-to-end attention models for speech recognition”.

I need to train 2 models: 1)the seq2seq model 2)an external LM，and they use the same vocabulary file.

I want to know which corpus is used to generate the vocabulary file("trans.bpe.vocab.lm.txt")
Is it from the voice tagging corpus(only 40M)，
or from the corpus of external LM(eg:librispeech corpus file is 4G)

I hope I have expressed it clearly. Looking forward to your reply, thanks.

[albertz]

Thanks for the interest in our work!
For LibriSpeech, the full setup to download the data, prepare it, create BPE codes + vocab, train the seq2seq model is here (excluding the LM currently; although the LM training config is here). That should answer how to train the seq2seq model and everything about the BPE.
The LM data is the official data for the LM. The same BPE as in training the seq2seq model is used for the LM. (@kazuki-irie might be able to give additional details, as he trained the LM.)

[tianwengang]

Appreciate your help!
LM now can be trained normally, except that "multiprocessing = True" does not seem to take effect.