Which GPU architecture Jeremy uses for training?

Which GPU (and also CPU and memory) configuration Jeremy uses for showing results of training in this course? I wonder to know that, because when I am doing training of notebooks in Google colab, I constantly getting 3-10 times slower performance for NN training.

Colab is a free resource but if you want to reach good compute speed then you need to get a paid GPU machine on cloud using Google cloud, Azure or other services. You can look at setup instruction for google cloud here - https://course.fast.ai/start_gcp.html

1 Like

Try Kaggle kernels, their GPU is very descent now (P100)

1 Like

Thank you for useful info!