jittor_geometric.dataloader

class jittor_geometric.dataloader.ClusterLoader(dataset, num_parts, mini_splits, fixed=False)[source]

The graph dataset loader, using metis to partition graph into ‘mini_splits’ groups and randomly merge them into ‘num_parts’ mini-batches for training. The dataset loader yields the induced graph of the selected nodes iteratively.

Parameters:

  • dataset (InMemoryDataset) – The original graph dataset.

  • num_parts (int) – Number of expected mini-batches.

  • mini_splits (int) – Number of parts that metis partitions.

  • fixed (bool, optional) – If set to ‘True’, the dataset loader will yield identical mini-batches every round.

get_node_indices()[source]
partition(edge_index, num_nodes, mini_splits)[source]
class jittor_geometric.dataloader.DataLoader(dataset, batch_size=1, shuffle=False, follow_batch=None, exclude_keys=None, collate_fn=None, **kwargs)[source]

A data loader which merges data objects from a jittor_geometric.data.Dataset to a mini-batch. Data objects can be either of type Data or HeteroData.

Parameters:

  • dataset (Dataset) – The dataset from which to load the data.

  • batch_size (int, optional) – How many samples per batch to load. (default: 1)

  • shuffle (bool, optional) – If set to True, the data will be reshuffled at every epoch. (default: False)

  • follow_batch (List[str], optional) – Creates assignment batch vectors for each key in the list. (default: None)

  • exclude_keys (List[str], optional) – Will exclude each key in the list. (default: None)

  • **kwargs (optional) – Additional arguments.

class jittor_geometric.dataloader.GeneralLoader(dataset, shuffle=None)[source]
dataset: InMemoryDataset
itercnt: int
itermax: int
class jittor_geometric.dataloader.NeighborLoader(dataset, source_node=None, num_neighbors=None, batch_size=None, fixed=False)[source]

The graph dataset loader, samples the neighbors of source nodes for graph convolution iteratively and randomly. The dataset loader yields all the tuples of (source_node, sampled neighbor) as the graph edge. The yielded data block will contain the node map and node index.

Parameters:

  • dataset (InMemoryDataset) – The original graph dataset.

  • source_node (Var) – The list of source_node.

  • num_neighbors (List[int]) – Number of sampled neighbors per layer. For example, [2, 3] represents sample 2 neighbors, and 3 neighbors of each sampled ones.

  • batch_size (int) – Size of each mini-batch

  • fixed (bool, optional) – If set to ‘True’, the dataset loader will yield identical mini-batches every round.

get_node_indices()[source]
neighbor_sampling(source_nodes)[source]
unique(x)[source]
class jittor_geometric.dataloader.RandomNodeLoader(dataset, num_parts, fixed=True)[source]

The graph dataset loader, randomly split all of the nodes into ‘num_parts’ mini-batches. The dataset loader yields the induced graph of the selected nodes iteratively.

Parameters:

  • dataset (InMemoryDataset) – The original graph dataset.

  • num_parts (int) – Number of expected mini-batches.

  • fixed (bool, optional) – If set to ‘True’, the dataset loader will yield identical mini-batches every round.

get_node_indices()[source]

```