I am interested in knowing the minimum hardware requirement for processing 50000 audio files each of length 1-2 min, within 1 minute or in minimum time possible using deepspeech ?
cpu? gpu? in parallel? in sequence? what’s the usecase?
1 Like
I can suggest you to prefer Google Colab. I tried previously it is very fine.
But you must need to save the checkpoint for every one hour or based on global steps like 1000, because Colab will disconnect when internet issues or 12 hours after it will stop. So you need to again restore last checkpoint.
Thanks @lissyx for the reply.
I am trying to convert speech from audio files to text in bulk. For now it will be helpful if you let me know for CPU, in parallel.
Just have a look at evaluate_tflite.py example if you need …