TypeError: __init__() missing 1 required positional argument: 'config_path' training language model

Archives DeepSpeech
#1

Hello everyone !

I’m trying to add more sentences to the language model. and to do this, i downloaded a wiki dump, preprocessed and cleaned it to be one sentence per line and also removed all sentences with numbers.

I am doing this as i find myself with a larger error for WER than CER in evaluation.

Here are some details;
DeepSpeech 0.7.4
Ubuntu 20.04
Python 3.7
tensorflow 1.15

Below is what i used to create binaries which were successfully created.

python3 generate_lm.py --input_txt /media/kamla/data/Voice_dataset/language_model_stuff/corpus/librispeech.txt --output_dir /media/kamla/847636CE7636C0AA/Users/offic/Documents/Kenlm --top_k 50000 --kenlm_bins /media/kamla/data/Voice_dataset/language_model_stuff/kenlm/build/bin --arpa_order 5 --max_arpa_memory "85%" --arpa_prune "0|0|1" --binary_a_bits 255 --binary_q_bits 8 --binary_type trie

Now when i try to use generate_package.py like below;
python3 generate_package.py --alphabet ../alphabet.txt --lm lm.binary --vocab vocab-50000.txt --package kenlm.scorer --default_alpha 0.931289039105002 --default_beta 1.1834137581510284

i get the following error;

    50000 unique words read from vocabulary file.
Doesn't look like a character based model.
Using detected UTF-8 mode: False
Traceback (most recent call last):
  File "generate_package.py", line 157, in <module>
    main()
  File "generate_package.py", line 152, in main
    args.default_beta,
  File "generate_package.py", line 48, in create_bundle
    alphabet = NativeAlphabet()
TypeError: __init__() missing 1 required positional argument: 'config_path'
#2

Please use 0.9.3, 0.7.4 is old and not supported anymore.

#3

Dear lissyx,

Thank you for the timely response.

Shall try what you suggest and update.

#4

Hello,what is librispeech.txt, Is it the same for all languages? Thank you so much

#5

I came across the same issue
had to downgrade ds-ctcdecoder
pip install ds-ctcdecoder==0.7.0