nip.graph_isomorphism.dataset_generation._generate_non_isomorphic_graphs

nip.graph_isomorphism.dataset_generation._generate_non_isomorphic_graphs#

nip.graph_isomorphism.dataset_generation._generate_non_isomorphic_graphs(config: GraphIsomorphicDatasetConfig, batch_size: int, device: device | str | int) → tuple[Int32[Tensor, 'pair batch node1 node2'], Int32[Tensor, 'batch'], Int32[Tensor, 'batch']][source]#

Generate non-isomorphic graphs.

Parameters:

  • config (GraphIsomorphicDatasetConfig) – The configuration for the dataset.

  • batch_size (int) – The number of pairs of graphs to generate.

  • device (TorchDevice) – The device to use for the computation. Note that all returned tensors will be on the CPU.

Returns:

  • adjacencies (Int32[Tensor, “pair batch node1 node2”], device=”cpu”) – The batch of adjacency matrices for the generated graphs.

  • sizes (Int32[Tensor, “batch”], device=”cpu”) – The number of nodes in each graph.

  • scores (Int32[Tensor, “batch”], device=”cpu”) – The Weisfeiler-Lehman scores for each pair of graphs.