aby2s/google_qa

google_qa

A short description of the project.

Instructions

Download Kaggle data to input and unzip:

$ make download

Important

Kaggle kernels expect to find the dataset in directory ../input. In contrast, for a notebook in path kernels/exploration, the dataset is found in ../../input. To be able to use the same code both locally and in Kaggle, add a symbolic link in kernels/ pointing to input/:

ln -sf ../input ./kernels/input

Notebook organization

All kernels should have their own folder in kernels/.

Before a kernel with a given folder can be pushed to Kaggle from command-line, it needs the metadata file kernel-metadata.json in the same folder (see the documentation). You can create the file either by running

$ kaggle kernels init -p /path/to/kernel/directory

to initialize the file, or check kernel-metadata.json for reference and copy it (with appropriate changes) to the folder with your kernel.

Once you're happy with the kernel and metadata has been setup, push it to Kaggle for execution:

$ kaggle kernels push -p /path/to/kernel/directory

Note that all kernels are private by default.

Detailed organization

    ├── LICENSE
    ├── Makefile           <- Makefile with commands like `make download` or `make train`
    ├── README.md          <- The top-level README for developers using this project.
    ├── unzip_input.sh     <- Bash script for unzipping all archives in `input`
    ├── data
    │   ├── external       <- Data from third party sources.
    │   ├── interim        <- Intermediate data that has been transformed.
    │   ├── processed      <- The final, canonical data sets for modeling.
    │   └── raw            <- The original, immutable data dump.
    |
    |── input              <- Raw data downloaded from Kaggle with `make download`
    |
    ├── models             <- Trained and serialized models, model predictions, or model summaries
    │
    ├── kernels                 <- Jupyter notebooks. Naming convention is a number (for ordering),
    │   └ 1.0-initial-exploration  the creator's initials, and a short `-` delimited description, e.g.
    │                              `1.0-jqp-initial-data-exploration`.
    │
    ├── references         <- Data dictionaries, manuals, and all other explanatory materials.
    │
    └── requirements.txt   <- The requirements file for reproducing the analysis environment, e.g.
                              generated with `pip freeze > requirements.txt`

Project created with the cookiecutter template for Kaggle competitions.