Housing Your Data in Google Drive

Mount Drive from Colab, save a gold set, load it back

Why this guide

In the worked tutorials, the course hands you data over a URL — e.g. Day 2’s day2_gold_standards_and_evaluation.ipynb loads a committed CEFR gold set straight from GitHub, so it just runs. That is the instructor’s distribution channel.

For the mini-project you flip roles: you build your own gold set (download → reshape → sample a subset) and you need somewhere to keep it. That place is your Google Drive — not GitHub.

NoteGitHub is ours, Drive is yours

GitHub is how we ship read-only course data to everyone. Your own rebuilt/sampled data lives in your Drive because it:

  • persists across Colab sessions — a Colab VM wipes /content/ when it disconnects, but your Drive files survive;
  • needs no git push — save and it’s there;
  • stays private and license-clean — some source datasets can’t be redistributed, so committing them to a repo would be a licensing problem. Keeping your copy in your Drive avoids that.

We’ll learn the pattern on the CEFR data you already know from Day 2, then reuse the exact same four steps for your mini-project dataset.

The pattern, in four steps

1. Mount your Drive

Run this in a Colab cell. It pops up a Google auth window the first time — approve it, and your Drive appears under /content/drive/MyDrive/.

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

2. Make a project folder

Keep everything for this course in one place so paths stay predictable.

import os

DATA_DIR = "/content/drive/MyDrive/lda2/gold"
os.makedirs(DATA_DIR, exist_ok=True)
print("Your data folder:", DATA_DIR)

3. Save your gold set into Drive

After you’ve built a gold set — e.g. by running the CEFR download notebook download_cefr_sp.ipynb, which produces a list of {"id", "text", "label"} items — write it to Drive:

import json

# `gold` is the list your download/build notebook produced
out_path = f"{DATA_DIR}/cefr_sentences.json"
with open(out_path, "w", encoding="utf-8") as f:
    json.dump(gold, f, ensure_ascii=False, indent=2)

print(f"Saved {len(gold)} items to {out_path}")

4. Load it back — from any future session

This is the payoff: once it’s in Drive, any notebook can read it with a local path instead of a URL. The tutorials’ load_gold() already accepts either:

GOLD_PATH = "/content/drive/MyDrive/lda2/gold/cefr_sentences.json"
gold = load_gold(GOLD_PATH)      # same function, path instead of URL
print(len(gold), "items loaded from Drive")

That’s the whole round-trip: mount → folder → save → load. Nothing is lost when Colab disconnects, and you never touch GitHub.

Reusing this for your mini-project

The four steps don’t change — you only swap which dataset you build and what you name the file. Pick your track in mini-project-tracks.md, run that track’s download notebook, then:

  1. Mount your Drive (step 1).
  2. Save your track’s gold set — e.g. f"{DATA_DIR}/raamove_moves.json" (step 3).
  3. Load it by path in your evaluation notebook (step 4).

One dataset, one mechanic, learned once on familiar CEFR data — now applied to your own.

Troubleshooting

Symptom Fix
drive.mount does nothing / asks again You disconnected the runtime. Re-run the mount cell and re-approve.
FileNotFoundError on load Check the path spelling — it’s MyDrive (no space), and the folder must exist (re-run step 2).
Saved file isn’t in Drive Confirm the cell printed a /content/drive/MyDrive/... path, then look under My Drive → lda2 → gold in the Drive web UI.
“Mountpoint must not already contain files” Run drive.flush_and_unmount() once, then mount again.

See also