Choose an execution environment
The default executable reference is Linux CPU, Python 3.12. Core workflows also run on Windows in CI. Hosted account creation, provider quotas, source access and assistive-technology interfaces need separate verification; a configuration or launch link does not establish live service usability.
| Route | Core | Framework modules 8–12 and 14 | Saving and limits |
|---|---|---|---|
| Local Linux CPU | Automated execution reference | Separate environment tested | Persistent files; allow download space and memory |
| Local Windows | Core CI exercised | Installation recipe; no complete platform claim | Use explicit environment executable |
| Local macOS | Use the Linux reference | Use the Linux reference | Use instructor support if installation fails |
| Codespaces | Core devcontainer configured | Run framework bootstrap in a separate environment | Account/access/usage allowance required; download work before deleting workspace |
| Colab | Import complete source and core bootstrap | Requires a matching runtime/kernel; use the Linux reference when incompatible | Upload the complete source archive; download runs and notebooks |
| Binder | Core requirements configured | Heavy frameworks excluded | Public source required by ordinary launch; ephemeral and quota-dependent |
Codespaces
Signed in to GitHub, choose Code, Codespaces, Create codespace. The checked-in devcontainer installs the core dependencies and notebook kernel. Open the desired notebook and select the installed Python. Run python scripts/preflight.py --profile core. To use frameworks in the Linux terminal, run python scripts/bootstrap.py --profile frameworks, then select the ml-ai-frameworks kernel. This installs several large packages; review your provider's available resources first.
Colab
Obtain the complete source ZIP from the repository or supplied archive, upload it using Colab's Files panel, and extract it. Never embed an access token in a notebook. In a notebook cell, use pathlib to find the folder containing pyproject.toml, then os.chdir into it. In the runtime's Python, run !python scripts/bootstrap.py --profile core --current. Restart the session if existing imported packages were changed, then rerun setup and open the course notebook in that complete tree. Colab's default package/kernel versions change; validate preflight and the selected module before assigning this route. The local separate-venv framework recipe does not automatically switch Colab's kernel.
Binder
A Binder service must be able to retrieve the complete source repository. If it cannot access the source or build the environment, use the local reference environment or an instructor-provided workspace. The .binder requirements install only the core profile. Binder build success, capacity and interface accessibility remain service checks. Save notebooks and runs outside the session before it closes.
Local setup
Follow workspace acquisition, first run and framework setup. A report-reading route can support interpretation, but required implementation outcomes need actual execution through an effective interface. Preflight and troubleshooting support all routes.