Doesn’t look like a character based (Bytes Are All You Need) model (DeepSpeech)

Viewed 277

I have been following DeepSpeech documentation in order to build my own scorer. After implementing this blocks of code

cd data/lm python3 generate_lm.py --input_txt vocabulary.txt --output_dir . –top_k 1500 --kenlm_bins path/to/kenlm/build/bin/ –arpa_order 3 --max_arpa_memory “50%” --arpa_prune “0|0|1” –binary_a_bits 255 --binary_q_bits 8 --binary_type trie

curl -LO http://github.com/mozilla/DeepSpeech/releases/… tar xvf native_client.*.tar.xz ./generate_scorer_package --alphabet …/alphabet.txt --lm lm.binary --vocab vocab-1500.txt –package kenlm.scorer --default_alpha 0.931289039105002 --default_beta 1.1834137581510284

I get the following errors:

Doesn’t look like a character based (Bytes Are All You Need) model. –force_bytes_output_mode was not specified, using value infered from vocabulary contents: false Error: Can’t parse scorer file, invalid header. Try updating your scorer file. Error loading language model file: Invalid magic in trie header.

I want to mention that I use a different alphabet that also contains other characters besides english characters.

1 Answers

This error is usually encountered with non-English alphabets. You can overcome the error by using the --force_bytes_output_mode parameter when calling generate_scorer_package. I pushed a change to the PlayBook this morning which covers this error.

Related