kaggle-template
local env
Use Docker in your local environment to access both Python and Jupyter environments.
For the Docker image, pull the image provided by kaggle and use it.
There are two types, CPU and GPU, but we'll start with the CPU version.
Here is the github repository provided by kaggle.
https://github.com/Kaggle/docker-python
About kaggle api
Download competitions data and dataset
kaggle competitions download {competitions name}
Displayed in the URL of the competition page with the name of the competition specified in kaggle-api.
Example:https://www.kaggle.com/c/{competition name}
The same is true for datasets: https://www.kaggle.com/{user name}/{dataset name}, where {user name} and {dataset name} are used to download the dataset.
kaggle dataset download {user name}/{dataset name}
Upload file to kaggle
Submit submission.csv to competition.
kaggle competitions submissions {competition name} -f submission.csv -m "messasge."
Show same lists
Show lists competitions
kaggle competitions list
Add -h options after list you can see list help.
Sort example:
kaggle competitions list --sort-by 'prize'
Show kernels list.
kaggle kernels list -s titanic
Push Notebook and code
kaggle kernel init -p {path}- setting
kernel-metadata.json kaggle kernel push -p {path}
If you want to change the output destination, change kernel-metadata.json.
Note: If you change the output destination, please change the title as well. If you do not change the title, the files on kaggle will conflict.
Link jupyternotebook to Colaboratory
Install a chrome extentions that will allow you to open github hosted notebooks in colab.
https://chrome.google.com/webstore/detail/open-in-colab/iogfkhleblhcpcekbiedikdehleodpjo/related
And put the following html in a markdown cell at the top of the notebooks.
<a href="https://colab.research.google.com/{your notebook url}" target="_parent"><img src="https://colab.research.google.com/assets/colab-badge.svg" alt="Open In Colab"/></a>