Kaggle CLI

June 18, 2018 · View on GitHub

(CLI = Command Line Interface)

Resource

Kaggle CLI Wiki

Installation

Check to see if kaggle-cli is installed:
kaggle-cli --version

Install kaggle-cli:
pip install kaggle-cli
or pip3 install kaggle-cli

May need to update package if you run into errors:
pip install kaggle-cli --upgrade
or pip3 install kaggle-cli --upgrade

Kaggle Competition Datasets

Note 1: You must have a Kaggle user ID and password. If you logged in to Kaggle using FB or LI, you'll have to reset your password, as that is needed for command line access to the data.

Note 2: Pick a competition, and ensure you have accepted the rules of that competition. Otherwise, you will not be able to download the data using the CLI.

Step 1: Identify the competition I will use

https://www.kaggle.com/c/dogs-vs-cats

Note: the competition name can be found in the url; here it is dogs-vs-cats

Step 2: Accept competition rules

https://www.kaggle.com/c/dogs-vs-cats/rules

Step 3: Set up data directory

ls
mkdir data
cd data

my example

ubuntu@ip-10-0-0-13:~$ ls
anaconda2  anaconda3  downloads  git  nbs  temp
ubuntu@ip-10-0-0-13:~$ mkdir data
ubuntu@ip-10-0-0-13:~$ cd data

Step 4a: Download data (try 1)

Syntax:
kg config -g -u 'username' -p 'password' -c 'competition'
kg download

Note: Here's an example of warning message I receive when I tried to download data before accepting the rules of the competition:

my example

ubuntu@ip-10-0-0-13:~/data$ kg config -g -u 'reshamashaikh' -p 'xxx' -c dogs-vs-cats
ubuntu@ip-10-0-0-13:~/data$ kg download
Starting new HTTPS connection (1): www.kaggle.com
downloading https://www.kaggle.com/c/dogs-vs-cats/download/sampleSubmission.csv

sampleSubmission.csv N/A% |                                                                                                                   | ETA:  --:--:--   0.0 s/B

Warning: download url for file sampleSubmission.csv resolves to an html document rather than a downloadable file. 
Is it possible you have not accepted the competition's rules on the kaggle website?

Step 4b: Dowload data (try 2)

Note 1: I have accepted the competition rules; will try downloading again
config -g -u 'username' -p 'password' -c 'competition'
kg download

my example

ubuntu@ip-10-0-0-13:~/data$ kg config -g -u 'reshamashaikh' -p 'xxx' -c dogs-vs-cats
ubuntu@ip-10-0-0-13:~/data$ kg download
Starting new HTTPS connection (1): www.kaggle.com
downloading https://www.kaggle.com/c/dogs-vs-cats/download/sampleSubmission.csv

Starting new HTTPS connection (1): storage.googleapis.com
sampleSubmission.csv 100% |##################################################################################################################| Time: 0:00:00 320.2 KiB/s

downloading https://www.kaggle.com/c/dogs-vs-cats/download/test1.zip

test1.zip 100% |#############################################################################################################################| Time: 0:00:08  32.5 MiB/s

downloading https://www.kaggle.com/c/dogs-vs-cats/download/train.zip

train.zip 100% |#############################################################################################################################| Time: 0:00:17  31.4 MiB/s

Download Kaggle Data (another way)

Note: sometimes setting up the configuration results in an error the next time you try to download another competition. You may want to bypass configuration and directly include your user ID, password and competition name in one command line.

kg download -u 'reshamashaikh' -p 'xxx' -c statoil-iceberg-classifier-challenge

Step 5: Look at data that was downloaded

ls -alt 

ubuntu@ip-10-0-0-13:~/data$ ls -alt
total 833964
-rw-rw-r--  1 ubuntu ubuntu 569546721 Nov  4 18:24 train.zip
drwxrwxr-x  2 ubuntu ubuntu      4096 Nov  4 18:24 .
-rw-rw-r--  1 ubuntu ubuntu 284321224 Nov  4 18:24 test1.zip
-rw-rw-r--  1 ubuntu ubuntu     88903 Nov  4 18:23 sampleSubmission.csv
drwxr-xr-x 22 ubuntu ubuntu      4096 Nov  4 18:23 ..
ubuntu@ip-10-0-0-13:~/data$

Step 6: Unzip Files

Note 1: You will need to install and use unzip to unzip files.

For Window users:

  1. First Download ubuntu from Window Microsoft Store

  2. Open PowerShell as Administrator and run:Enable-WindowsOptionalFeature -Online -FeatureName Microsoft-Windows-Subsystem-Linux

  3. Once the download has completed, select "Launch".This will open a console window. Wait for installation to complete then you will be prompted to create your LINUX user account.

  4. Create your LINUX username and password.

  5. Go to Control Panel and Turn on Developer Mode .

  6. Run bash from command-prompt. After that you can follow same as Linux users guide.

For Linux Users:

sudo apt install unzip
unzip train.zip
unzip -q test.zip (Note: -q means to unzip quietly, suppressing the printing)

ubuntu@ip-10-0-0-13:~/nbs/data$ ls train/dogs/dog.1.jpg
train/dogs/dog.1.jpg
ubuntu@ip-10-0-0-13:~/nbs/data$ ls -l train/dogs/ | wc -l
12501
ubuntu@ip-10-0-0-13:~/nbs/data$ 


ubuntu@ip-10-0-0-13:~/nbs/data$ ls -l train/cats/ | wc -l
12501
ubuntu@ip-10-0-0-13:~/nbs/data$
ubuntu@ip-10-0-0-13:~/nbs/data$ ls test1 | wc -l
12500
ubuntu@ip-10-0-0-13:~/nbs/data$

Kaggle - Submit Results

kg submit <submission-file> -u <username> -p <password> -c <competition> -m "<message>"

my example

/home/ubuntu/data/iceberg/sub
(fastai) ubuntu@ip-172-31-2-59:~/data/iceberg/sub$ 

kg submit resnext50_sz150_zm13.csv -u 'reshamashaikh' -p 'xxx' -c statoil-iceberg-classifier-challenge

Jeremy’s Setup

Good to copy 100 or so the sample directory; enough to check that the scripts are working

Advice 1: Separate TEST data into VALIDATION TASK: move 1000 each dogs / cats into valid

> ls valid/cats/ | wc -l
1000
> ls valid/dogs/ | wc -l
1000

Advice 2:  Do all of your work on sample data
> ls sample/train

> ls sample/valid

> ls sample/train/cats | wc -l
8
> ls sample/valid/cats | wc -l
4

Kaggle API

Another option is to use the Kaggle API https://github.com/Kaggle/kaggle-api