how long does the train take (section 6)? about tfodcourse HOT 5 OPEN

@nguaki how long did you let it train after that?

from tfodcourse.

I am training using Google Collab. Training in GPU failed miserably for me. Although training in TPU worked.
I followed Nick's project by letter by letter. Google Collab has some serious production flaws IMO. CUDA and cuDNN mismatch with Tensorflow. All these problems are out of my control.

from tfodcourse.

you can check the Cuda version by !nvcc --version and maybe downgrade the Tensorflow version accordingly. I made the mistake of not creating the test and train folders and it got stuck on the training on both local and on collab.

from tfodcourse.

Google colab recommends using the default TF. Does not recommend downgrading TF.
Which version did you use? What exact pip command did you use? Did you use any pip option?
I tried to downgrade TF, but that result in the same error.
I must be doing something wrong.

from tfodcourse.

@nguaki hey does this problem solved for you ?? Can you help me please

from tfodcourse.