How long does model training take

How long did it take to model training for version 0.7? (the model that is provided here is https://github.com/mozilla/DeepSpeech/releases/tag/v0.7.0)

That highly depends on your hardware. Please be more specific in what you intend to train on.

Iā€™m interested in how much time you spent on the pre-trained model?

a server with 8 Quadro RTX 6000 GPUs each with 24GB of VRAM

The various phases of training were described here all made use of 8 Quadro RTX 6000 GPUs

  • The first phase took approx 6 days
  • The second phase took approx 5 days
  • The third phase took approx 4 days

For optimization of the alpha + beta parameters it was about 6 days of compute, but only using a single RTX 6000.

So in sum total about 21 days.

2 Likes

Thanks for the information provided.