Binbin Zhang

what's the error message?

any update now?

Any result about the accuracy?

> any WER comparison about the AC and WFST?

how many context phrases in your testing?

2.44? our best result shown in README is 2.66.

We just fixed a bug. please see https://github.com/wenet-e2e/wenet/pull/847. BPE is not used to tokenize the anotation due to the bug.

fixed. please see https://github.com/wenet-e2e/wenet/pull/848.

language_type is just for postprocessing of blank. For European language, you should set it to 1.

We are not sure. I think you can just ignore the warning and continue the training. The final WER should be comparable to the old code.