# dmlc/dgl

Python package built to ease deep learning on graph, on top of existing DL frameworks.

Repository: https://github.com/dmlc/dgl
Canonical: https://ross.abutalabs.com/products/dgl
Homepage: http://dgl.ai
Language: Python
License: Apache-2.0
License Family: permissive
Topics: deep-learning, graph-neural-networks
Last push: 2025-07-31T23:53:10+00:00

## Health v2 (maintenance only)
Score: 38/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 34, release rhythm 8, longevity 100
- inputs: {"age_days": 3057, "days_push": 398, "days_rel": 729, "gap_med": null, "n_releases_24m": 1}
- flags: none
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 14282, forks 3054 (observed 2026-08-28T04:11:06.669813+00:00)

## What it is
DGL (Deep Graph Library) is a Python package for deep learning on graph-structured data, built on top of existing frameworks like PyTorch, TensorFlow, and Apache MXNet. It provides efficient message-passing primitives, GNN layers, GPU-accelerated data pipelines, and distributed training for graphs with hundreds of millions of nodes and edges.

## Use cases
- train graph neural networks on social network data
- predict molecular properties with GNNs
- build knowledge graph embeddings at scale
- run node classification on large graphs with mini-batch sampling
- implement graph transformer models
- benchmark GNN models on OGB datasets
- scale GNN training across multiple GPUs

## When to choose
- you need a mature, GPU-ready library for graph neural networks
- you want framework-agnostic GNN support across PyTorch, TensorFlow, or MXNet
- you need to train on very large graphs with distributed or stochastic training
- you want ready-made GNN layers, model examples, and domain projects like DGL-KE or DGL-LifeSci

## When to avoid
- your data is tabular or unstructured text with no graph structure
- you only need classical graph algorithms like shortest paths without deep learning
- you need a lightweight pure-PyTorch-native solution and prefer PyG's ecosystem
- you require a framework with no dependency on external DL backends

## Facets
- artifact type: library
- maturity: stable
- function: deep-learning, machine-learning, gpu-computing, data-science
- domain: deep-learning, machine-learning, graph-processing, artificial-intelligence
- platform: python, windows
- tags: graph-neural-networks, gnn, message-passing, framework-agnostic, pytorch, tensorflow, mxnet, graph-sampling, distributed-training, linux, macos, gpu

## Member repositories
- dmlc/dgl (main) score 38

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:11:06.669813+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-29T17:12:40.499951+00:00, confidence not recorded.
  - readme: https://github.com/dmlc/dgl (fetched 2026-08-28T04:11:06.669813+00:00, sha 1d68fcb337f3)
  - homepage: http://dgl.ai (fetched 2026-08-29T08:07:05.946695+00:00, sha 710dd8466aec)
  - site_page: https://www.dgl.ai/pages/about.html (fetched 2026-08-29T08:07:05.956061+00:00, sha f989c31b95ee)
  - site_page: https://www.dgl.ai/dgl_docs/graphtransformer/index.html (fetched 2026-08-29T08:07:05.958101+00:00, sha 3cd307665f67)
  - site_page: https://www.dgl.ai/dgl_docs (fetched 2026-08-29T08:07:05.959775+00:00, sha 249792cfeb36)
  - registry_pypi: https://pypi.org/pypi/dgl/json (fetched 2026-08-29T08:07:05.964420+00:00, sha 3ba0ebf6604c)
  - site_page: https://www.dgl.ai/pages/start.html (fetched 2026-08-29T08:07:05.962937+00:00, sha 299414466b6b)
- Data as of 2026-08-30T08:39:29.467469+00:00.