Getting UnwrapDatasetVariant Exception after learning model

Deep Speech
#1

Hi guys,
Sorry for my bad English!:slight_smile:

I’m not professional with deep learning and DeepSpeech. I just wanna learn my Persian(Farsi) model for my internship project.

I have Geforce 1060 GTX, AMD Ryzen 5 2600 Six-Core Processor and 12 GB Ram.

I learned my model using the documentation with Cuda. I installed Cuda and the cudnn in my system.
So, now I wanna test my model and check model with a .wav file. after I run my model I get this error:

Screenshot%20from%202019-08-15%2013-58-31
Screenshot from 2019-08-15 13-58-31.png1346×768 246 KB

I leaned my model with this command:

python ./DeepSpeech.py --train_files /home/mohammad/Documents/DeepSpeech/Farsi/clips/train.csv --epochs 1 --export_dir /home/mohammad/Documents/DeepSpeech/Farsi --checkpoint_dir /home/mohammad/Documents/DeepSpeech/Farsi/checkpoints --alphabet_config_path /home/mohammad/Documents/DeepSpeech/Farsi/alphabet.txt

and test my model with this command:

deepspeech --model /home/mohammad/Documents/DeepSpeech/Farsi/output_graph.pb --alphabet /home/mohammad/Documents/DeepSpeech/Farsi/alphabet.txt --audio /home/mohammad/Documents/DeepSpeech/Farsi/clips/common_voice_fa_18202356.wav

So, what am I doing wrong?

#2

Looks like you trained on the master branch but you’re trying to use a v0.5.1 client. You should build the client from source or use the latest alpha release.

#3

Thanks, I downloaded deepspeech-0.6.0a4-cp36-cp36m-manylinux1_x86_64.whl and installed it.
Now, when running the model, I don’t get any response. Just a clear enter. Is that because the model didn’t learn well?? or there is another problem?

Screenshot%20from%202019-08-17%2019-55-47
Screenshot from 2019-08-17 19-55-47.png1274×858 79.4 KB

#4

That’s the likely answer, yes.

#5

Thank you very much!

#6

Please avoid sharing screenshots.