Kaggle Notebook Usage

Kaggle Notebooks provide an excellent cloud-based environment for data science and machine learning projects. This guide covers essential features, including GPU acceleration, data uploads, generating zip files, and downloading outputs via IPython.display.FileLink.

Setting Up a Kaggle Notebook

To start a new Kaggle Notebook:

• Go to Kaggle

• Click on New Notebook

• Choose the Notebook Settings (CPU/GPU/TPU as needed)

• Click Create

Enabling GPU/TPU

Kaggle provides free access to GPUs and TPUs. To enable GPU/TPU in your notebook:

• Click on Settings (gear icon on the right panel)

• Under Accelerator, choose GPU or TPU (If you creating a new account then you must verify phone number to access).

• Restart the notebook for changes to take effect

• To check GPU availability in your notebook, run:

import torch
print(torch.cuda.is_available())

Uploading Data to Kaggle Notebook

You can upload custom datasets by:

1. Clicking on Add Data (right panel)

2. Uploading files from your local system

3. Accessing the uploaded files in /kaggle/input/

To check uploaded files:

import os 
os.listdir("/kaggle/input/")

Creating a ZIP File from Output

If your notebook generates multiple files, you might want to compress them into a ZIP file for easy downloading.

Run the following command to create a ZIP file:

import shutil
shutil.make_archive("output", "zip", "/kaggle/working/")

This will create output.zip inside /kaggle/working/.

Downloading the ZIP File

Kaggle does not provide a direct file download button, but you can use IPython.display.FileLink to generate a downloadable link.

from IPython.display import FileLink
FileLink("output.zip")

Click on the generated link to download your ZIP file.