Ship and explain the score labeler

A shippable lab artifact includes: how to run it, the metric result (boundary labels at 0.49/0.50/0.51 and empty-list positive count 0), the failure you can still reproduce (string scores that compare lexicographically, or IndentationError that hides a wrong cutoff), the security gate for eval() on pasted score text or logging raw learner identifiers beside scores, and a rollback note. The user decision it supports remains: accept or reject a model score using one shared cutoff.

def validate(scores, threshold):
    if not isinstance(scores, list):
        raise TypeError("scores must be a list")
    values = []
    for index, value in enumerate(scores):
        if isinstance(value, bool) or not isinstance(value, (int, float)):
            raise TypeError(f"score[{index}] must be a number")
        value = float(value)
        if not 0.0 <= value <= 1.0:
            raise ValueError(f"score[{index}] must be between 0 and 1")
        values.append(value)
    if isinstance(threshold, bool) or not isinstance(threshold, (int, float)):
        raise TypeError("threshold must be a number")
    threshold = float(threshold)
    if not 0.0 <= threshold <= 1.0:
        raise ValueError("threshold must be between 0 and 1")
    return values, threshold

def label_scores(scores, threshold):
    scores, threshold = validate(scores, threshold)
    labels = ["yes" if score >= threshold else "no" for score in scores]
    return labels, labels.count("yes")

assert label_scores([0.49, 0.50, 0.51], 0.50) == (["no", "yes", "yes"], 2)
assert label_scores([], 0.50) == ([], 0)
try:
    label_scores(["0.9"], 0.50)
except TypeError as error:
    assert str(error) == "score[0] must be a number"
else:
    raise AssertionError("numeric strings must be rejected")

print(label_scores([0.2, 0.9, 0.4], 0.5))
print("3/3 release checks passed")

Expected output:

(['no', 'yes', 'no'], 1)
3/3 release checks passed

Save this as scores.py and run python3 scores.py. It uses only the Python standard library.

State operating limits for the score labeler in plain language: fixture size, offline vs live dependencies, and what would require a new eval set. Shipping python-only-what-you-need is honest scoping, not maximal confidence language.

Fill this table in your notes for the score labeler:

• Proved on scores=[0.2,0.9,0.4], threshold=0.5: …
• Unproved beyond the fixture: …
• Metric that blocks release: boundary labels at 0.49/0.50/0.51 and empty-list positive count 0
• Failure still reproducible: string scores that compare lexicographically, or IndentationError that hides a wrong cutoff
• Security gate: eval() on pasted score text or logging raw learner identifiers beside scores
• Rollback: …

Ship the narrative only when the unproved list is honest. Reviewers trust narrow claims that support accept or reject a model score using one shared cutoff more than maximal language that collapses under the first production oddity.

Hand your ship note to a peer and ask them to recreate a proved/unproved ship note with rollback without watching you type. If they cannot, your evidence is still tribal knowledge. Tighten the run command and the metric line until a stranger can validate the score labeler against scores=[0.2,0.9,0.4], threshold=0.5.

Ship a temperatures.py artifact with the same structure: explicit contract, one shared boundary rule, empty-input behavior, a rejected string case, aggregate-only logs, and a six-line ship note. A peer should be able to run and review it without this lesson.

ML Python starter

Previous