Run your full Python project in Google Colab without breaking paths

Most examples show small Colab notebooks with a few cells, but real ML work means full projects: multiple scripts, data folders, models, and utilities. This guide shows how to run an entire Python project on Colab without file path chaos – and how to keep it reproducible.

1. Put your project in a single folder

my-python-project/
  src/
    __init__.py
    train.py
    utils.py
  notebooks/
    colab_entrypoint.ipynb
  data/
    raw/
    processed/
  models/
  requirements.txt
  README.md

Push this to GitHub or store it in Google Drive under MyDrive/projects/.

2. Mount Drive and change working directory

from google.colab import drive
drive.mount('/content/drive')

PROJECT_ROOT = "/content/drive/MyDrive/projects/my-python-project"
%cd $PROJECT_ROOT

Now src/, data/, and notebooks/ exist under a stable root, and you can import modules from src.

3. Install dependencies in the Colab runtime

From the notebook:

!pip install -r requirements.txt

Colab uses a fresh environment, so this step must run whenever the runtime resets. Put it near the top of your notebook.

4. Run project scripts from the notebook

You can either:

• call Python scripts using !python, or
• import functions from src/ and call them directly.

# Option 1: run training script
!python src/train.py --config configs/baseline.yaml

# Option 2: import & call
import sys
sys.path.append("src")

from train import run_experiment

run_experiment(config_path="configs/baseline.yaml")

5. Keep data paths relative to PROJECT_ROOT

Inside your code, avoid absolute paths like C:\Users\.... Use:

from pathlib import Path

PROJECT_ROOT = Path(__file__).resolve().parents[1]
DATA_DIR = PROJECT_ROOT / "data" / "raw"

This works both locally and in Colab (as long as you cd into PROJECT_ROOT first).

6. Use NoteCapsule for project snapshots

Once the project runs end-to-end in Colab, create a Capsule so others can reproduce the run later without asking you for “that exact environment”.

from notebookcapsule import create_capsule

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

Share the Capsule with your repo so collaborators know exactly which data and environment were used for a given run.