Closed bkiefer closed 7 months ago
Hi. Thanks for the question.
Please follow the data preparation instructions to generate dev_error.json: You need to run get_rarewords.py
with rareword_error.txt
and data for dev set to generate dev_error.json.
Hi, thanks for the instructions!
I hope this is not a completely stupid issue, but using LibriSpeech and train_large.sh as described in the ReadMe, it breaks with the error that it can not load data/LibriSpeech/dev_error.json, which is obviously not present, and there is no description what to use to replicate the results.