Beginner

Installing FastAI

Get FastAI running on your machine in minutes. We cover pip and conda installation, GPU configuration, and cloud notebook options.

Install with pip

Bash

# Install fastai (includes PyTorch)
pip install fastai

# Or with specific PyTorch version
pip install torch torchvision torchaudio
pip install fastai

Install with conda

Bash

# Using conda (recommended for managing CUDA dependencies)
conda install -c fastchan fastai

Easiest Option: Use Google Colab or Kaggle Notebooks. FastAI comes pre-installed with GPU access — no setup required. Just open a notebook and start with from fastai.vision.all import *.

GPU Setup

FastAI uses PyTorch's CUDA support. If you have an NVIDIA GPU:

Bash

# Verify CUDA is available
python -c "import torch; print(torch.cuda.is_available())"

# Check GPU details
python -c "import torch; print(torch.cuda.get_device_name(0))"

Verify Your Installation

Python

from fastai.vision.all import *

# Check version
import fastai
print(f"fastai version: {fastai.__version__}")

# Quick test: download and train a model
path = untar_data(URLs.PETS)
print(f"Dataset downloaded to: {path}")
print("FastAI is working!")

Cloud Notebook Options

Platform	Free GPU	Setup Required
Google Colab	Yes (T4/V100)	None — fastai pre-installed
Kaggle Notebooks	Yes (P100/T4)	None — fastai pre-installed
Paperspace Gradient	Free tier available	Minimal — fastai template available
AWS SageMaker	No (paid)	Some setup required

Next Up: Vision

Now that FastAI is installed, let's build your first image classifier with just a few lines of code.

Next: Vision →

← Introduction Vision →