Installation

Choose the installation path that best fits your setup and experience level.

Comparison of installation paths

Path

GPU required

Docker required

Installation time

Best for

Google Colab

No (free GPU provided)

No

2 minutes

Quick start, no local setup

Dev container

Recommended

Yes

10 minutes

Full development

Native Python

Optional

No

5 minutes

Direct control

Conda

Optional

No

5 minutes

Conda users

Local training

Optional

No

5 minutes

Single GPU, no database

Prerequisites by path

All paths require:

  • 10GB free disk space (for CelebA dataset)

  • Internet connection

  • Database credentials from instructor (for distributed training)

Path-specific requirements:

Google Colab:

  • Google account

  • Web browser

Dev container:

  • Docker with GPU support

  • VS Code with Dev Containers extension

  • NVIDIA drivers ≥545 (for GPU support)

Native Python:

  • Python 3.10 or later

  • pip package manager

  • NVIDIA drivers (for GPU, optional)

Conda:

  • Anaconda or Miniconda

  • NVIDIA drivers (for GPU, optional)

CPU vs GPU training

Most paths support both CPU and GPU:

  • Google Colab: Choose GPU or CPU runtime

  • Dev container: Choose GPU or CPU configuration when opening

  • Native Python / Conda: Install GPU or CPU requirements, auto-detects hardware

  • Local training: Works with either

GPU training: Faster, recommended for active participation
CPU training: Works but slower

The worker automatically detects your hardware. If you encounter out-of-memory errors, reduce batch_size in config.yaml.

Verification

After installation, verify your setup:

# Test PyTorch installation
python -c "import torch; print(f'PyTorch: {torch.__version__}')"
python -c "import torch; print(f'CUDA available: {torch.cuda.is_available()}')"

# Test project imports
cd src
python -c "from models.dcgan import Generator, Discriminator; print('Models OK')"
python -c "from database.db_manager import DatabaseManager; print('Database OK')"

Next steps

Once installed, proceed to: