Kaggle Kernels OK?

I’m a bit confused. Is it OK to use Kaggle Kernels in place of setting up GPU?

Take look here for info on using Kaggle for the course notebooks. You can also create your own Kaggle kernals using the fastai library with a GPU.

Thank you Jon!

FYI – Just ran the Pets dataset through Resnet-34 using the Kaggle Kernals. Got the following results:

Total time: 10:32

epoch train_loss valid_loss error_rate
1 1.399896 0.350267 0.100135
2 0.551874 0.247893 0.071719
3 0.345219 0.221379 0.069012
4 0.264218 0.212395 0.068336

I can tell you that setting-up AWS EC2 and getting to this point when I did v1 of the course a couple years ago took days. Last year, because I already had an AWS instance, it took less time when I did v2 but tonight by using the Kaggle Kernal, this is by-far the fastest I’ve gotten to the point where I could train a model. I’m running Pets through Resnet50 next. Nice resource – thx!!

1 Like

yes, just make sure GPU is on and internet is Connected.

1 Like

Thank you!