OpenViSTAI · Agents · Blockchain · Crypto
Submit a project
Menu
AI Models

Pykeen

PyKEEN is an open-source Python package for training and evaluating knowledge graph embedding models, including approaches that incorporate multimodal information. It provides model and dataset components, a high-level pipeline, training loops and evaluators, with support for user-provided triples.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Pykeen — OpenViST project profile in AI Models.
OpenViST project illustration. See the linked sources for product details.
Understand the project

Practical overview

When it fits

Use PyKEEN when you want a Python-based workflow to train and assess knowledge graph embedding models, from its documented TransE and Nations example to experiments using custom triples.

Who it is for

  • Researchers and practitioners building knowledge graph embedding experiments.
  • Python users who need to train and evaluate link-prediction or knowledge graph completion models.
  • Teams adapting embedding experiments to domain-specific graph data.

Documented uses

  • Train an embedding model on a built-in or custom knowledge graph dataset.
  • Evaluate a model using PyKEEN’s training and evaluation workflow.
  • Compare models through shared package APIs and repeatable experiments.
  • Explore embedding approaches that incorporate multimodal information.
Before you choose

Requirements & limits

The package is open source under the MIT license. The repository states that its latest stable version requires Python 3.11 or later and can be installed through PyPI or from GitHub.

  • PyKEEN is a library that runs in a Python environment, not a hosted modeling interface.
  • The documented installation requirement is Python 3.11 or later for the latest stable version.
  • The quickstart returns model, training-loop and evaluation information in a PipelineResult object.
  • Custom graph data can be supplied through user-generated triples factories; the repository links to a dedicated data tutorial.

Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.

Source-based practical guide

Train and evaluate TransE on the Nations dataset

Use PyKEEN’s pipeline to train a TransE knowledge graph embedding model and save its results.

Before you start

  • PyKEEN installed in a Python environment.

Steps

  1. Import pipeline from pykeen.pipeline.

  2. Call pipeline(dataset='Nations', model='TransE').

  3. Save the returned result with pipeline_result.save_to_directory('nations_transe').

Expected outcome

The pipeline returns a `PipelineResult` with the trained model, training loop, and evaluation; the example saves results to the `nations_transe` directory.

Guide limitations

  • The example uses the Nations dataset and TransE model.
  • The default training approach is sLCWA and the default evaluation is rank-based.
  • Do not pass negative_sampler when using the LCWA training approach.
Documentation 1 ↗
Evidence and transparency

Sources & review

Official repository ↗Guide documentation ↗
Source record details
Source availability
Active
Sources checked
11 October 2026
Original website title
pykeen/pykeen · Official homepage title · retained in its source language
How we review sourcesSuggest a correction