search

apache / incubator-hugegraph-ai

The integration of HugeGraph with AI/LLM & GraphRAG
https://hugegraph.apache.org/docs/quickstart/hugegraph-ai/
Apache License 2.0
56 stars 28 forks source link

feat(ml): node emdedding, node classify and graph classify with models based on DGL #83

Closed edw4rdyao closed 3 weeks ago

edw4rdyao commented 2 months ago

This PR is derived from the project of GLCC2024: https://www.gitlink.org.cn/hugegraph/hugegraph/wiki?wiki=graph-ai-1

Features:

  1. Read graphs from HugeGraph and convert them to DGL format

    • [x] Convert single graph dataset
    • [x] Convert mulltigraph dataset
    • [x] Convert heterogeneous graph dataset

  2. Implemented graph learning algorithms (for algorithim test, please visit hugegraph-integrate-model-test)

    • [x] DGI, GRACE, GRAND, JKNet
    • [x] DiffPool, GlobalPool
    • [x] HAN

  3. Encapsulated training and evaluation to simplify the end-to-end graph learning workflow.

    • [x] Node classify task
    • [x] Graph classify task
    • [x] Node embedding task

Additional Notes:

  1. Since the PGL hasn't been updated for a long time, priority is given to integrating the DGL.
edw4rdyao commented 2 months ago

Also note the py-lint ci failed

image

pylint failed for the hugegraph-llm and hugegraph-python-client, should we fix it? Also, should we add pylint CI for hugegraph-ml in workflows and code_format_and_analysis.sh?

imbajin commented 2 months ago

pylint failed for the hugegraph-llm and hugegraph-python-client, should we fix it? Also, should we add pylint CI for hugegraph-ml in workflows and code_format_and_analysis.sh?

It's a bit strange, let me take a look (Most of the issues should indeed be fixed, and if they are unnecessary, they should be eliminated)