Distributed Sampling

[alexbarghi-nv]

Work needed to implement distributed sampling, which will replace bulk sampling for most users.

Tasks

Beta Give feedback

1. [IMP] cuGraph-PyG Should Use torch_geometric.EdgeIndex #4051
   breaking improvement +2
   
2. [FEA] New Distributed Graph Store for cuGraph-DGL #4247
   feature request +1
   
3. [FEA] Split cuGraph-PyG Feature Store from Graph Store #4249
   feature request +1
   
4. New Distributed MG cuGraph GNN Examples #4252
   doc feature request +2
   
5. [FEA] Support Isolated Vertices in cuGraph-PyG/DGL #4326
   feature request +1
   
6. [BUG] CSR/CSC Output from Bulk Sampler is unusable in cuGraph-PyG #3942
   bug +1
   
7. [IMP] Don't Copy to Host When Creating CuGraphStore (cugraph-pyg) #3383
   improvement non-breaking +2
   
8. New TensorDictFeatureStore for cuGraph-DGL #4398
   feature request +1
   
9. WholeFeatureStore for cuGraph-PyG #4399
   feature request +1
   
10. WholeFeatureStorage for cuGraph-DGL #4400
    feature request +1
    
11. Support Heterogeneous Sample Output in cuGraph Distributed Sampler #4401
    feature request +1
    
12. Support Heterogeneous Sample Input in cuGraph-PyG #4402
    +0
    
13. Support Heterogeneous Input in cuGraph-DGL #4403
    feature request +1
    
14. API for Offline Sampling in cuGraph-PyG #4405
    feature request +1
    
15. API for Offline Sampling in cuGraph-DGL #4406
    feature request +1
    
16. Getter and setter functions of dgl.graph in cugraph storage #3303
    feature request +1
    
17. rapidsai/graph_dl#543