Final-year ML project reproducibility checklist (Jupyter & Colab)

You’ve trained the model, tuned the hyperparameters, and got good accuracy. But will your guide, external examiner, or recruiter be able to run your notebook? This checklist helps you turn a fragile Jupyter/Colab project into a clean, reproducible final-year ML submission.

1. Clean up your project structure

final-year-ml-project/
  notebooks/
    exploration.ipynb
    training.ipynb
  data/
    raw/
    processed/
  models/
  reports/
    final_report.pdf
    slides.pptx
  capsules/
  requirements.txt
  README.md

Make sure all the files your project needs live under this folder (in Drive or locally).

2. Make your notebook runnable from top to bottom

• Restart kernel + run all cells – no errors.
• No manual tweaking of paths midway.
• All installs & imports are at the top.

# installs (for Colab)
!pip install -r ../requirements.txt

# imports
import numpy as np
import pandas as pd
...

3. Capture your environment

From the environment you used to run the notebook:

pip freeze > requirements.txt

Commit or include this file with your submission so others can recreate the same environment.

4. Make data references clear

• Use relative paths like ../data/processed/train.csv.
• If data is too large, include a sample and explain where to get the full dataset.
• Document any data pre-processing steps in README.md or the report.

5. Save your best model & metrics

• Save the final model under models/.
• Save evaluation metrics and key plots under reports/figures/.
• Reference these in your final report and slides.

6. Create a NoteCapsule Capsule for your submission

As a final step, create a Capsule that bundles your working notebook, environment snapshot, data manifest, and metadata. This is what you’ll share with your guide/examiner.

from notebookcapsule import create_capsule

create_capsule(
    name="final_year_submission",
    notebook_path="notebooks/training.ipynb",
    data_dirs=["./data", "./models"],
    base_dir=".",   # project root
)

Check that the Capsule folder contains everything needed to rerun the project.

7. Final-year reproducibility checklist