How to integrate fastai library on Kaggle kernels?

@jeremy How can we integrate jupyter notebooks with fastai library on kaggle kernels, since we need to ‘source activate fastai’ to work on that. How can this be done for kaggle kernel?

I’m not sure it can - but the stuff from fastai.structured doesn’t really have any dependencies AFAIK, so you can probably just paste the functions you need into the kernel (just be sure to credit their source, of course).

Ok. Thanks. I will try that.

For now, I just uploaded the notebook on github and shared github link on the kernal with summary about it. Found it a lot easier.

Yeah I saw that - since you’re using the full library, that’s probably all you can do for now.

1 Like

Hi Dalia, I managed to replicate this issue on a Windows 10 laptop with the newest Pytorch and Fastai libraries.

I managed to replicate this issue on a Windows 10 laptop with the newest Pytorch!

Is it possible to have both versions installed? Although I strongly agree with @lafoss that there is need to keep a 0.7 version of fastai to maintain old kernels compatibility, it is also good to have the new fastai which has much better API and documentation… myccpay

The kaggle github page has instructions for if you want to add a library here to the Kaggle kernel. I imagine it would be safest to wait until an official.

Hi all,
If you’re looking to get started with the 2019 version of (William Horton | Expert | Kaggle) and I have created the new lesson kernels here: Kernel links for Fastai (2019): Practical Deep Learning for coders | Data Science and Machine Learning | Kaggle
If you need help with anything, please feel free to leave a comment in the thread linked above.