scorers.from_function()

Make a scorer from a function, with a name, a version, and a batch size.

Usage

scorers.from_function(
    fn,
    name=None,
    version=None,
    batch_size=None,
    per_row=False,
)

Any function of a batch of rows can be used as a scorer in Validate.col_vals_scored(). This wraps one to give it the attributes of the Scorer protocol: a stable name for the score cache and the report, a version that’s part of the key of the cache, and the number of rows to score at a time. With per_row=True, it also wraps a function that scores one row at a time.

Parameters

fn: Callable[…, Any]: The function. It takes a list of rows (dictionaries of the values of the columns of the step) and returns a score for each, or with per_row=True, it takes one row and returns its score. A score is a number, or None for a row that can’t be scored.
name: str | None = None: A stable name for the scorer. A function defined at the top level of a module is named after its module and its name by default; a lambda or a function defined inside another function has no name unless it’s given here, and its scores can only be cached in memory.
version: str | None = None: A version of the scorer (e.g., of its model). Scores in the cache are kept by name and version, so changing the version means scoring the rows again.
batch_size: int | None = None: The number of rows that the function gets at a time (256 by default).
per_row: bool = False: Whether the function scores one row at a time (True) or a list of rows (False, the default).

Returns

Scorer: A scorer for Validate.col_vals_scored().

Examples

Here, a function that scores one row (a ratio of two columns) becomes a scorer with a name and a version.

import pointblank as pb
import polars as pl

tbl = pl.DataFrame({"clicks": [5, 40, 2, 18], "views": [100, 200, 120, 90]})

ctr = pb.scorers.from_function(
    lambda row: row["clicks"] / row["views"], name="ctr", version="1", per_row=True
)

pb.Validate(data=tbl).col_vals_scored(
    columns=["clicks", "views"], scorer=ctr, min_val=0.03
).interrogate()

		STEP	COLUMNS	VALUES	TBL	EVAL	UNITS	PASS	FAIL	W	E	C	EXT
#4CA64C66	1	col_vals_scored()	clicks, views	≥0.03 ctr		✓	4	3 0.75	1 0.25	—	—	—
Notes Step 1 (scores) Scored by `ctr` (version 1): 4 distinct rows scored. Scores from 0.01667 to 0.2 (median 0.125).

The scorer’s name, ctr, is shown in the report. One row has a click-through rate below 3%.