search

vagrawal / deepsphinx

MIT License
19 stars 29 forks source link

Examples of script invocations for training and inference? #5

Closed yuvalkry closed 7 years ago

yuvalkry commented 7 years ago

Hello!

Thanks for making the project compatible with python-3 !

In setup.py, change the script name from 'bin/deepsphinx_train' to 'bin/deepsphinx-train' .

Could you please provide basic examples of script invocations for training and inference?

I ran python bin\deepsphinx-train --trans-file data\ds-input.txt --nouse-train-lm --job-dir data\train5 and got a division by zero, something was probably wrong in the call.

Thanks, Yuval

INFO:tensorflow:Getting speaker stats INFO:tensorflow:Starting training INFO:tensorflow:Epoch completed, saving INFO:tensorflow:Evaluation started INFO:tensorflow:Restoring parameters from data\elta\train55\checkpoints/-0 Traceback (most recent call last): File "bin\deepsphinx-train", line 270, in tf.app.run(train) File "C:\Users\user\AppData\Local\Programs\Python\Python35\lib\sitepackages\tensorflow\python\platform\app.py", line 48, in run_sys.exit(main(_sys.argv[:1] + flags_passthrough)) File "bin\deepsphinx-train", line 153, in train lm_fst) File "bin\deepsphinx-train", line 79, in run_eval tot_wer / tot_ev, tot_cer / tot_ev)) ZeroDivisionError: float division by zero

vagrawal commented 7 years ago

Your training data likely has less examples than batch size.

Sorry for being late for the docs, but I will be surely updating the docs for inference and training and provide a pre-trained model by tomorrow :-) .

yuvalkry commented 7 years ago

That would be great!

Does the system support adaptation? can I add training data to a pre-trained model?

Thanks, Yuval

vagrawal commented 7 years ago

Yes, you can start the training from a pre-trained model, but the vocabulary needs to be the same.

vagrawal commented 7 years ago

I have made a tool for inference. Please have a look.

Any more suggestions are highly welcome.