pmlb

library1.0.1.post3pypypi✓ verified 84d ago

A Python wrapper for the Penn Machine Learning Benchmark (PMLB) data repository, which provides curated datasets for evaluating classification and regression algorithms. Current version is 1.0.1.post3, with low release cadence.

ai-ml

pip install pmlb

github ↗homepage ↗

INSTALL

IMPORT

SIG · PMLB

pmlb

ai-mlpythonv1.0.1.post3

harness data pending

Install & Compatibility

Where this runs

No compatibility data collected yet for this library.

Code

Verified usage

Verified import paths — ran on the pinned version, not inferred.

fetch_data

✓ from pmlb import fetch_data

✗ from pmlb.pmlb import fetch_data

Old import path no longer works.

dataset_names

✓ from pmlb import dataset_names

✗ from pmlb.datasets import dataset_names

Use top-level import.

Load the iris dataset as a DataFrame or as (X, y) arrays.

from pmlb import fetch_data, dataset_names # List all available datasets print(len(dataset_names)) # Fetch a dataset as a pandas DataFrame df = fetch_data('iris') print(df.head()) # Fetch with 'return_X_y=True' for sklearn-style split X, y = fetch_data('iris', return_X_y=True) print(X.shape, y.shape)

Debug

Known issues

deprecatedThe function `pmlb.classification_dataset_names` and `pmlb.regression_dataset_names` are deprecated as of version 1.0.0. Use `dataset_names` instead.

fix

Use `from pmlb import dataset_names` and filter manually if needed.

affects: <1.0.0

gotchaDatasets are loaded into memory entirely as pandas DataFrames. Some datasets can be large (e.g., 'adult' with ~48k rows), which may cause memory issues on low-RAM systems.

fix

Use `return_X_y=True` to get NumPy arrays (still in memory). Consider subsampling if necessary.

affects: all

gotchaSome datasets contain missing values represented as '?' or other sentinel values. They are not automatically cleaned.

fix

Check dataset metadata or manually handle missing values after loading.

affects: all

Errors

Common errors & fixes

ModuleNotFoundError: No module named 'pmlb'

Library not installed or installed in a different environment.

fix

Run `pip install pmlb` in the correct Python environment.

AttributeError: module 'pmlb' has no attribute 'fetch_ucirepo'

Confusion with another library (e.g., ucimlrepo) or outdated code expecting a different function name.

fix

Use `fetch_data` from pmlb, not `fetch_ucirepo`.

ValueError: Unknown dataset: 'adult_original'

The requested dataset name does not exist in PMLB (e.g., it may be named 'adult' instead).

fix

Check available datasets with `pmlb.dataset_names`.

Upgrade

Version history

1.0.1.post3latest on PyPI · released Sep 10, 2020

Audit

Dependencies

pandasrequiredData handling; returned datasets are DataFrames.

scikit-learnoptionalOptional for downstream ML tasks; not required for fetching data.

Agent activity

9 hits · last 30 days

node

Resources

githubhttps://github.com/EpistasisLab/penn-ml-benchmarks ↗

homepagehttps://github.com/EpistasisLab/penn-ml-benchmarks ↗