Evaluation

slicemap

Find the slices where a new model regressed.

0 downloads/month on PyPI

Install

pip install slicemap-cli

If you use uv, also as an isolated CLI:

uv tool install slicemap-cli

What it does

A new model can lift the overall metric while getting worse on a segment that matters: a country, an age band, a category. slicemap scores every slice and lists where the new model lost ground, ranked by impact.

In action

A real recorded session of the CLI running.

Read it as text

slicemap: accuracy overall 0.877 -> 0.857
feature  slice       size    old    new  regression
plan     enterprise    53  0.830  0.660      -0.170

$ slicemap compare preds.csv --true y_true --old old_pred --new new_pred --features plan
slicemap: accuracy overall 0.877 -> 0.857
feature  slice       size    old    new  regression
plan     enterprise    53  0.830  0.660      -0.170

Features

—Categorical and quantile slicing.
—Accuracy, error and MAE metrics.
—Impact ranking: size times degradation.
—CI gate and JSON output.

View the code on GitHub →

Other tools

dsdiff

A git-style diff between two datasets, with distribution drift.

splitcheck

Detect rows leaking between train, validation and test splits.

pii-sweep

Scan datasets for PII before sharing them.