Hi
I have a question about deepspeech STT training. I use this final command to start training process:
python3 /home/files/DeepSpeech/DeepSpeech.py --alphabet_config_path /home/files/DeepSpeech/data/alphabet.txt --train_files /home/files/audiosamples/clips/train.csv --dev_files /home/files/audiosamples/clips/dev.csv --test_files /home/files/audiosamples/clips/test.csv --train_cudnn true --checkpoint_dir /home/files/checkpoints/ --export_dir /home/files/exported_model/
I didn’t use the scorer_path
flag in this command, does deepspeech use any scorer by default or any internal scorer???
othiele
(Olaf Thiele)
August 12, 2020, 1:31pm
2
Scorer is used for testing only, not for training. Please search before you post, this came up many times in the last months:
We are happy for every new user of DeepSpeech. As it is under active development, stuff changes quickly and you might have problems running it. We are here to help and might point to this thread if needed
Please do the following before posting:
Search this forum or Google for your problem
Read the documentation for your version here
Check the new Playbook on how to use DeepSpeech
Check Github issues for more technical problems
If you still struggle, feel free to open a new t…
I’m searching for about two months, but I didn’t see any special comment on “Scorer is used for testing only, not for training”, neither in the document no in other questions. So what does ‘scorer/scorer_path’ do in training command???
Thanks in advanced
baconator
(Bacon Ator)
August 15, 2020, 5:20am
4
Nothing. It’s only used for inference.
othiele
(Olaf Thiele)
August 17, 2020, 9:12am
5
Reading this forum will give you a lot knowledge about the process as you’ll probably want to build a custom scorer next …
Just for reference, not to sound like a clever one, the thread I meant, please read threads, you’ll learn a lot:
Hi,
I want to use the pre trained deepspeech 0.7.0 model as is. To use it for my use case, I saw that making an external scorer helps.
There are many hyperparameters that impact the scorer like “lm alpha”, “lm beta” along with a bunch of other parameters like: “beam width”
How will changing these parameters affect the transcription output?
Also is the scorer required to train the acoustic model??
Thanks!
1 Like