search

khalil-research / PyEPO

A PyTorch-based End-to-End Predict-then-Optimize Library for Linear and Integer Programming
MIT License
429 stars 56 forks source link

Using Graph-Based Models for Prediction #24

Closed ssakhavi closed 6 months ago

ssakhavi commented 7 months ago

Hi,

Thank you for the great repo! Is it possible to use GNNs as the model? For example, can we use a PyG model?

Such as the paper below? https://arxiv.org/pdf/2302.05636.pdf image

LucasBoTang commented 7 months ago

Hi @ssakhavi ,

Thank you for the kind words about the repository!

Given that PyEPO is built on PyTorch, it should be indeed compatible with PyTorch Geometric (PyG).

However, it's important to note that while theoretically, the compatibility between PyEPO and PyG should allow for the integration of PyG models within the PyEPO framework, this integration has not been explicitly tested in practice. I would like to see the use between PyG and PyEPO, and glad to contribute towards bridging any existing gaps.

Best regards,

Bo

ssakhavi commented 7 months ago

@LucasBoTang Sorry for the late reply.

Thank you. We'll consider this library when narrowing down our possible options.