Create a prefix for each SVO file as necessary, according to how they were embedded.
Create a prefix for each SVO file as necessary, according to how they were embedded.
If the edges are embedded, we need to differentiate the latent representations if they were not made together. That is, if we have two or more embedded SVO files, and they have embeddings files that are _different_, that means that a +L1 edge from one and a +L1 edge from another are not the same edge type. So we add a prefix to the edge type that is specific to each embedding. This isn't a problem with KB edges vs. SVO edges, because the "alias" relation assures that the two kinds of edges will never share the same space.
Runs GraphChi's preprocessing (sharding) on the graph.
Runs GraphChi's preprocessing (sharding) on the graph. This produces a number of shard files, and if the files are already present, this is a no-op. So it's only run once for each graph, no matter how many times you run GraphChi code.