Graphsage data prep for large datasets

[parhamouni]

Hello there,
I would like to apply the code for a dataset with a millon nodes, however using the dense matrix would cause a memory error in graphsage data prep. do you have any idea how to fix it?

[matenure]

I did not fully understand your question. Not sure whether you want to convert the format of GraphSAGE into the format of FastGCN or inverse.
Anyway, my suggestion is to first check the memory load at each important step and find the bottleneck.
And possible solutions may be:

1. Change the dense matrix into sparse.
2. Split your graph into different blocks or batches when doing the data prep.