Deepspeech: error: argument --lm_beta: invalid float value: 'models/lm.binary'

Hi, I am a researcher and am trying to install DeepSpeech to simplify and speed up transcription of audio recordings of interviews. I have managed to install deepspeech based on some guidelines I found on the web.

When I try to run deepspeech, I get some confusing errors:
Command:
deepspeech --model models/output_graph.pbmm --alphabet models/alphabt.txt --lm models/lm.binary --trie models/trie --audio interviewCloete.wav

Output: usage: deepspeech [-h] --model MODEL [–scorer SCORER] --audio AUDIO
[–beam_width BEAM_WIDTH] [–lm_alpha LM_ALPHA]
[–lm_beta LM_BETA] [–version] [–extended] [–json]
[–candidate_transcripts CANDIDATE_TRANSCRIPTS]
deepspeech: error: ambiguous option: --lm could match --lm_alpha, --lm_beta

I then specified lm_alpha:
deepspeech --model models/output_graph.pbmm --alphabet models/alphabt.txt --lm_alpha models/lm.binary --trie models/trie --audio interviewCloete.wav

I then get the following output:
usage: deepspeech [-h] --model MODEL [–scorer SCORER] --audio AUDIO
[–beam_width BEAM_WIDTH] [–lm_alpha LM_ALPHA]
[–lm_beta LM_BETA] [–version] [–extended] [–json]
[–candidate_transcripts CANDIDATE_TRANSCRIPTS]
deepspeech: error: argument --lm_alpha: invalid float value: ‘models/lm.binary’

I am not a coder or a hacker, so I don’t know how to resolve this. Can anyone assist?
Many thanks in advance!

Have you read the output ? There is no --lm option, please use the documented --scorer.

It really looks like you are applying instructions from a different release …

ok thanks for the rapid response.

this is the version I installed: 0.7.1-cp36-cp36m-manylinux1_x86_64

where do I find the correct commands to use for this version?

when I run the commands modified:
deepspeech --model models/output_graph.pbmm --alphabet models/alphabet.txt --scorer models/lm.binary --trie models/trie --audio interviewCloete.wav

I get: error: unrecognized arguments: --alphabet models/alphabet.txt --trie models/trie

You have it in the error message that you pasted as well as in the doc https://deepspeech.readthedocs.io/en/v0.7.1/USING.html

Yes, as documented no more trie nor alphabet parameters are required, they are included in the model / scorer.

1 Like

ok thanks am downloading the relevant pre-trained models
will give it a go when that is done, using the commands in the link you provided.
as I am running v 0.7.1, do I have to download models only for v0.7.1 or can I use v0.7.0 as reflected in the documentation?

0.7.0 models are fine

1 Like

ok so I have downleaded these model files and dropped them into the directory with the other deepspeech files. when I run the command:
./deepspeech --model deepspeech-0.7.0-models.pbmm --scorer deepspeech-0.7.0-models.scorer --audio interviewCloete.wav
I get:
bash: ./deepspeech: No such file or directory
I’m stumped…not sure how to proceed

ok ignore that - resolved.