Google Colab — Introduction

Google Colab — Introduction

What is Google Colab?

Google Colaboratory (Colab) is a free, cloud-hosted Jupyter notebook environment provided by Google. It requires no setup, runs entirely in the browser, and provides free access to GPUs and TPUs — making it the most popular platform for ML experimentation and learning.

Key facts:

• Built on Jupyter Notebook (.ipynb compatible)
• Runs on Google’s infrastructure
• Free tier includes GPU/TPU access
• Notebooks stored in Google Drive
• Collaborative editing (like Google Docs)

Colab vs Local Jupyter

Feature	Google Colab	Local Jupyter
Setup	None	Install Python + packages
GPU/TPU	Free (with limits)	Requires hardware
Storage	Google Drive	Local disk
RAM	12–52 GB (tier dependent)	Your machine’s RAM
Session timeout	12h (free), longer (paid)	No timeout
Collaboration	Real-time	Manual sharing
Internet access	Yes	Yes
Custom packages	Yes (per session)	Persistent
Cost	Free / Pro / Pro+	Free (hardware cost)

Colab Tiers

Tier	GPU	RAM	Session Length	Price
Free	T4 (shared, limited)	~12 GB	12h max	Free
Colab Pro	T4/V100 (priority)	~25 GB	24h	~$10/mo
Colab Pro+	A100/V100 (priority)	~52 GB	24h+ background	~$50/mo

Getting Started

1. Go to colab.research.google.com
2. Sign in with Google account
3. File → New notebook or open from Drive
4. Start coding — Python environment is pre-configured

Pre-installed Libraries

Colab comes with most data science libraries pre-installed:

# These work out of the box
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
import seaborn as sns
import sklearn
import tensorflow as tf
import torch
import cv2
import PIL

Checking Runtime

import platform
print(platform.python_version())  # Python version

# Check GPU
import torch
print(torch.cuda.is_available())
print(torch.cuda.get_device_name(0))

# TensorFlow GPU
import tensorflow as tf
print(tf.config.list_physical_devices('GPU'))

# System info
!nvidia-smi          # GPU details
!cat /proc/cpuinfo   # CPU info
!cat /proc/meminfo   # Memory info
!df -h               # Disk space

Connecting to Google Drive

from google.colab import drive
drive.mount('/content/drive')

# Access files
import pandas as pd
df = pd.read_csv('/content/drive/MyDrive/data/myfile.csv')

# Save files
df.to_csv('/content/drive/MyDrive/output/results.csv', index=False)

Uploading Files

from google.colab import files

# Upload from local machine
uploaded = files.upload()
# Returns dict: {filename: bytes}

import io
df = pd.read_csv(io.BytesIO(uploaded['data.csv']))

# Download a file
files.download('output.csv')

Runtime Types

Change via Runtime → Change runtime type:

Runtime	Use Case
CPU	General Python, light ML
GPU (T4/V100/A100)	Deep learning, PyTorch, TensorFlow
TPU	Large-scale TensorFlow training

# Verify GPU is active
import tensorflow as tf
print("GPU available:", tf.test.is_gpu_available())

# PyTorch
device = 'cuda' if torch.cuda.is_available() else 'cpu'
model = model.to(device)