This Pull Request is part of the Distributed Graph Partitioning feature pipeline to support massively large graphs for distributed training. This is to add support to the ParMETIS to read input graph files which are in multiple file format. Here it is assumed that each process is mapped to a input graph file for each of the graph files (nodes, edges, node features and edge features). Each process reads nodes/edges file which is assigned to the current process, using the process-id as suffix string to the appropriate input graph file name. After reading the input graph file, we as
This Pull Request is part of the Distributed Graph Partitioning feature pipeline to support massively large graphs for distributed training. This is to add support to the ParMETIS to read input graph files which are in multiple file format. Here it is assumed that each process is mapped to a input graph file for each of the graph files (nodes, edges, node features and edge features). Each process reads nodes/edges file which is assigned to the current process, using the process-id as suffix string to the appropriate input graph file name. After reading the input graph file, we as