dc_tts icon indicating copy to clipboard operation
dc_tts copied to clipboard

Running on GPU

Open OOps717 opened this issue 4 years ago • 2 comments

Hello, I am trying to run training on both my computer and google colab, however even if I am trying to change something it still trains on CPU. I am guessing if there is any solution ?

OOps717 avatar May 09 '20 13:05 OOps717

I did face that issue and found that you would have to remove tensorflow package and install tensorflow-gpu package for python and then the code will start utilizing GPU RAM. But my CPU is still getting utilized 100% not sure, why but I can confirm that the training rate is increased exponentially.

Note: tensorflow-gpu package will require some extra configuration to be done or else you will start getting errors, also you would have to you python 3.6 env or else this code breaks (learnt this the hard way )

kheniparth avatar May 11 '20 12:05 kheniparth

The solution i am going to share worked for me

first,

unistall tensorflow from environment **pip uninistall tensorflow**

then,

install tensorflow-gpu 1.14 **pip install tensorflow-gpu==1.14**

Pawandeep-prog avatar Oct 09 '20 09:10 Pawandeep-prog