MIIA Pothole Image Classification Challenge
Identify which images of roads in South Africa contain potholes
198 data scientists enrolled, 36 on the leaderboard
2 September 2019
Can't use Kaggle due to lack of writing permissions
published 4 Sep 2019, 15:20

Hi all!

So after attempting for ages to download, combine files, rezip and upload to Kaggle, I've discovered that we can't use Kaggle due to the kernels not providing write permissions. This leads to this error "Exception: [Errno 30] Read-only file system: '../input/AI-Expo/models' Can't write to '../input/AI-Expo/models', set `learn.model_dir` attribute in Learner to a full libpath path that is writable" when at the "learn.lr_find()" stage of Jan's notebook (../input/AI-Expo/models is just my directory for my Kaggle data).

Has anyone successfully run Jan's notebook on Google colab yet? Asking because it takes ages to get thes upoads completed and hoping to save myself the bother if it isn't possible.


PS: Please reccomend an alternative if you know one.

I haven't used Kernels in a very long time. Is there no directory there that you have write access to?

I do think that Colab should work. You can try some of these steps to get started: https://course.fast.ai/start_colab.html.

I managed to run and modify Jan's code on Google Colab, so can confirm that that works. No changes other than editing the path to where I uploaded the data.

I think Kaggle have extra limits to prevent people messing about in the filesystem too much - an issue in this case.

Used google colab as well and it worked fine

Thank you to all of you for your responses. I will attempt Google colab now that you've provided some Intel. Cheers! To get you your question Jan, you are only able to use the Pandas .to_csv method to save text files. There is no directory that a user can access with write permissions, unfortunately.

Thanks again for the help, it's very much appreciated.

am using kaggle kernel but not with Jan's notebook