About the format of the data set?

hi @acbull
my case is that i should use a sample of the training data sets to build a graph but not all of the training data sets.
that is to say: Use the internal elements of the sample to construct a heterogeneous graph.
I was wonder if i can use GPT-GNN to fine-tuning it and then Do a classification task in the downstream ?

thanks!!!

Sure, pre-training is very suitable for your case.

About data format, you can refer to

GPT-GNN/GPT_GNN/data.py

Line 20 in 2409049

class Graph():

to check our data structure. Or you can use your own data structure, and adapt our code to your setting.

To construct such a graph, you can refer to prepreocess_*.py as a reference.

thank you!!!