search

Hanjun-Dai / graph_comb_opt

Implementation of "Learning Combinatorial Optimization Algorithms over Graphs"
https://arxiv.org/abs/1704.01665
MIT License
491 stars 135 forks source link

how to build graphnn #6

Open SHUht opened 6 years ago

SHUht commented 6 years ago

Could you tell me how to build graphnn? And I want to know whether use pytorch version graphnn?

Hanjun-Dai commented 6 years ago

Hi there,

you can find the build instruction here: https://github.com/Hanjun-Dai/graphnn

The pytorch version of graphnn is not compatible with this repository yet. But you can write your own Q-learning with the pytorch_structure2vec: https://github.com/Hanjun-Dai/pytorch_structure2vec

neurohn commented 6 years ago

@Hanjun-Dai : What all needs to change here in this repo for it to be compatible with pytorch_structure2vec?

Hanjun-Dai commented 5 years ago

Sorry it is not that easy to be compatible with pytorch directly. I've implemented everything (including gnn, q-learning) in C++. But you can definitely implement this with pytorch_structure2vec with additional effort on the implementation of q-learning part.