FastAI Docker Environment for Local GPU

This Docker container provides a minimal environment for running FastAI course notebooks with local NVIDIA GPU support. It's specifically tested with the NVIDIA RTX 5070 Ti (sm120).

Prerequisites

• Docker installed on your system
• NVIDIA GPU drivers installed
• NVIDIA Container Toolkit (nvidia-docker2)

Features

• Minimal base image to reduce container size
• CUDA support for GPU acceleration
• Jupyter Lab environment
• FastAI library and dependencies
• Python 3.x environment

Quick Start

1. Using the docker compose (recommended):

docker-compose up --build

2. Or manually build and run:

docker volume create fastai-notebooks

docker build -t fastai-local-v2 .

3. Run the container:

docker run --gpus all -p 8888:8888 -v fastai-notebooks:/home/jupyter/ fastai-local-v2

This will:

• Enable GPU access with --gpus all
• Map port 8888 for Jupyter Lab access
• Mount current directory to /workspace in container

Accessing Notebooks

Once the container is running:

1. Open your browser to http://localhost:8888
2. The Jupyter Lab interface will load with access to your notebooks
3. Any notebooks in your current directory will be available in the /workspace folder

GPU Verification

To verify GPU access inside the container, you can run:

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

Notes

• This container is optimized for the NVIDIA RTX 5070 Ti with Compute Capability 12.0
• Adjust CUDA version in Dockerfile if using different GPU models
• Container includes minimal dependencies to reduce size and build time

License

This project is open-source and available under the MIT License.