NVIDIA / OpenSeq2Seq

Toolkit for efficient experimentation with Speech Recognition, Text2Speech and NLP
https://nvidia.github.io/OpenSeq2Seq
Apache License 2.0
1.54k stars 372 forks source link

Cannot parse trie file for speech recognition #502

Open hao-olli-ai opened 4 years ago

hao-olli-ai commented 4 years ago

When I use language model for decoder, the system shows this error:

Error: Can't parse trie file, invalid header. Try updating your trie file.

I followed this link https://discourse.mozilla.org/t/tutorial-how-i-trained-a-specific-french-model-to-control-my-robot/22830 to create language model files including *.binary and trie.

lmplz and build_binary are from kenlm, and generate_trie is from Deepspeech. All of them are the newest version.

I tried several times to generate trie file but at the end the system shows the error below.