Is graph-learn ready for all batch sampling results before training？ about graph-learn HOT 4 CLOSED

Hi,

1. Sampling is executed during training.
2. Traversal methods like EdgeSampler/NodeSampler only select vertices or edges from local machine(like data parallel)，
   but when you sample neighbors, GL will query the neighbors of each vertice, and these neighbors may be on other machines.

from graph-learn.

@baoleai Is this "Sampling is executed during training" a data pipeline? which means when the first batch is being trained, the next batch is being sampled at the same time?

from graph-learn.

@baoleai Is this "Sampling is executed during training" a data pipeline? which means when the first batch is being trained, the next batch is being sampled at the same time?

@baoleai @Seventeen17 @archwalker @alibaba-oss I also want to ask this question.

from graph-learn.

Yes, see GSL in user docs.

from graph-learn.