There is a nice way to use your local IDE with all the best benefits and your local resources and the kaggle cloud 🙂 I think it calls hybrid cloud (at least, this is what ChatGPT told me). For all Information, you can take a look here at kaggle or github for help about that API.
I use Visual Studio Code on my local machine and I can do quick EDA or first model selection. But If I’m moving to the hyperparameter tuning, then I’m using the resources from kaggle. As a kaggler you have every week up to 30 hours GPU/TPU time. There is a convenient way to use both worlds.
- Create a kaggle notebook, for example from a competition and give it a proper name.
- Pull the notebook with an easy API command
kaggle kernels pull dantheshark/playground-series-e4s333. Paste the command in your terminal, but you have to make some changes to it. Please add -p for the folder and -m for the metadata. In this example, we don’t add another folder. The notebook will appear at the same directory where you are.
kaggle kernels pull dantheshark/playground-series-e4s33 -p ./ -m
4. You are ready to push your file to the kaggle server. You can use following command:
kaggle kernels push -p ./
5. Of course, you can also download all data from the competitions with this command:
kaggle competitions download -c playgroundseries-s4e3If you have any questions, let me know
regards daniel