🎯

What is Fast.ai?

Fast.ai is a high-level deep learning library built on PyTorch that makes AI accessible to everyone. Designed with the «practice first, theory later» principle.

✨ For Beginners: You can train state-of-the-art models with just 5 lines of code.
⚡ Already Installed: Your system has PyTorch 2.9.1 + Fast.ai 2.8.5 working correctly.
💡 Your Current Setup: CPU Mode (GPU not detected). Perfect for learning, but for serious training, consider switching to a GPU instance.
📈

Your Learning Progress

✓

Installation Complete

Environment configured and validated

1

First Steps

Learn to activate environment and use Jupyter

2

Your First Model

Train an image classifier

3

Personal Project

Apply Fast.ai to your own dataset

4

AWS Production

Deploy models to the cloud

🚀

Getting Started – Start in 5 Minutes!

1

Activate the Environment

First, activate your Fast.ai virtual environment:

cd ~/fastai_project && source fastai_env/bin/activate

You’ll see (fastai_env) in your terminal when active.

2

Start Jupyter Lab

Launch the interactive development environment:

jupyter lab –ip=0.0.0.0 –port=8888 –no-browser &

Open your browser at: http://your-ip:8888

🌐 On AWS: If using EC2, find your public IP with:
curl -s ifconfig.me
3

Your First Code!

Create a new notebook and test this code:

Image Classification
Text Analysis
Tabular Data
vision.ipynb 
# Your first vision model in 5 lines
from fastai.vision.all import *

# 1. Download example data (pets dataset)
path = untar_data(URLs.PETS)/'images'

# 2. Create DataLoaders automatically
dls = ImageDataLoaders.from_name_re(
    path, get_image_files(path), 
    pat=r'(.+)_\d+.jpg$',
    item_tfms=Resize(460),
    batch_tfms=aug_transforms(size=224)
)

# 3. Create pre-trained model
learn = vision_learner(dls, resnet34, metrics=error_rate)

# 4. Find optimal learning rate
learn.lr_find()

# 5. Train the model (4 epochs)
learn.fine_tune(4)
text.ipynb 
# Sentiment analysis with text
from fastai.text.all import *

# 1. Movie reviews data
path = untar_data(URLs.IMDB)

# 2. Create DataLoaders for text
dls = TextDataLoaders.from_folder(
    path, valid='test',
    text_vocab=Vocab.create(size=10000),
    seq_len=72
)

# 3. Language model
learn = text_classifier_learner(
    dls, AWD_LSTM, 
    drop_mult=0.5,
    metrics=accuracy
)

# 4. Progressive training
learn.fit_one_cycle(1, 2e-2)
learn.freeze_to(-2)
learn.fit_one_cycle(1, slice(1e-2/(2.6**4), 1e-2))
tabular.ipynb 
# Prediction with tabular data
from fastai.tabular.all import *
import pandas as pd

# 1. Load data (Titanic example)
path = untar_data(URLs.TITANIC)
df = pd.read_csv(path/'train.csv')

# 2. Define categorical and continuous variables
cat_names = ['Sex', 'Embarked', 'Pclass']
cont_names = ['Age', 'Fare', 'SibSp', 'Parch']

# 3. Create DataLoaders
dls = TabularDataLoaders.from_df(
    df, path, y_names="Survived",
    cat_names=cat_names,
    cont_names=cont_names,
    procs=[Categorify, FillMissing, Normalize]
)

# 4. Model for tabular data
learn = tabular_learner(dls, metrics=accuracy)

# 5. Train
learn.fit_one_cycle(5)
💼

Beginner Use Cases

Fast.ai is perfect for these initial projects:

🖼️

Image Classification

Recognize objects, animals, products.

Example: Classify fruit photos

Easy 2 hours
📊

Sentiment Analysis

Analyze emotions in texts, reviews.

Example: Classify positive/negative tweets

Easy 3 hours
🏠

Price Prediction

Predict values based on features.

Example: House price prediction

Easy 2 hours
🎨

Style Transfer

Apply artistic styles to photos.

Example: Photo in Van Gogh style

Intermediate 4 hours
📝 Project Structure Example: 
project/
├── data/                    # Your dataset
│   ├── train/
│   └── valid/
├── notebooks/              # Jupyter notebooks
│   ├── 01_exploration.ipynb
│   └── 02_training.ipynb
├── models/                 # Saved models
├── src/                    # Source code
│   └── utils.py
└── requirements.txt
☁️

AWS Optimization Guide

Your Current Instance

You’re running in CPU mode. For real deep learning, use GPU instances:

Instance Type GPU Price/Hour Best For
g4dn.xlarge NVIDIA T4 ~$0.526 Beginners, Prototyping
g5.xlarge NVIDIA A10G ~$1.006 Production, Teams
p3.2xlarge NVIDIA V100 ~$3.06 Research, Heavy Training
Spot Instances All types 60-90% discount Learning, Batch Jobs
# Check current instance type

Essential AWS Commands

1

Configure AWS CLI

aws configure

You’ll need: Access Key ID, Secret Access Key, region (e.g., us-east-1)

2

Upload Data to S3

aws s3 cp my_dataset.zip s3://my-bucket/datasets/
3

Download from S3

aws s3 sync s3://my-bucket/datasets/ ./datasets/
4

Monitor GPU Usage

nvidia-smi –loop=1
🔧

Troubleshooting Guide

Problem: «Command not found: jupyter»

Solution: Activate virtual environment first

source ~/fastai_project/fastai_env/bin/activate
Problem: «CUDA: False» (no GPU detected)

Solution: For real training, switch to GPU instance or use CPU mode:

# Train on CPU (slower but works)
# Install GPU drivers if available: sudo apt-get install nvidia-driver-535
Problem: «Out of memory»

Solution: Reduce batch size:

dls = DataLoaders(…, bs=16) # Try 8, 4, 2
Problem: Slow training on CPU

Solution: Optimize CPU performance:

export OMP_NUM_THREADS=$(nproc)
export MKL_NUM_THREADS=$(nproc)
💡 Pro Tip: Use Google Colab for free GPU

Visit colab.research.google.com and run !pip install fastai

🗺️

30-Day Learning Path

W1

Week 1: Fundamentals

  • Day 1-2: Install and run first examples
  • Day 3-4: Understand DataBlocks and DataLoaders
  • Day 5-7: Train models on sample datasets
W2

Week 2: Your First Project

  • Day 8-10: Choose a personal project
  • Day 11-12: Collect and prepare your data
  • Day 13-14: Train and evaluate your model
W3

Week 3: Advanced Topics

  • Day 15-17: Hyperparameter tuning
  • Day 18-20: Model interpretation and debugging
  • Day 21-22: Transfer learning techniques
W4

Week 4: Deployment

  • Day 23-24: Export models for production
  • Day 25-26: Deploy to AWS/GCP
  • Day 27-28: Create API endpoints
  • Day 29-30: Monitor and improve

📚 Quick Index

🎯 Quick Actions

⏱️ Time Estimates

  • First model: 15 minutes
  • Basic project: 2 hours
  • Basic mastery: 1 week
  • Advanced project: 1 month

📊 System Status

✅ Installation: Verified
🤖 PyTorch: 2.9.1
🚀 Fast.ai: 2.8.5
🎮 GPU: Not detected
📍 Mode: CPU Learning