🆕 dgl.graphbolt¶

dgl.graphbolt is a dataloading framework for GNN that provides well-defined APIs for each stage of the data pipeline and multiple standard implementations.

Dataset¶

A dataset is a collection of graph structure data, feature data and tasks.

`Dataset`	An abstract dataset which provides abstraction for accessing the data required for training.
`OnDiskDataset`	An on-disk dataset which reads graph topology, feature data and Train/Validation/Test set from disk.
`BuiltinDataset`	A utility class to download built-in dataset from AWS S3 and load it as `OnDiskDataset`.
`LegacyDataset`	A Graphbolt dataset for legacy DGLDataset.
`Task`	An abstract task which consists of meta information and Train/Validation/Test Set.

Graph¶

A graph is a collection of nodes and edges. It can be a homogeneous graph or a heterogeneous graph.

`SamplingGraph`	Class for sampling graph.
`FusedCSCSamplingGraph`	A sampling graph in CSC format.

Feature and FeatureStore¶

A feature is a collection of data(tensor, array). A feature store is a collection of features.

`Feature`	A wrapper of feature data for access.
`FeatureStore`	A store to manage multiple features for access.
`BasicFeatureStore`	A basic feature store to manage multiple features for access.
`TorchBasedFeature`	A wrapper of pytorch based feature.
`TorchBasedFeatureStore`	A store to manage multiple pytorch based feature for access.
`GPUCachedFeature`	GPU cached feature wrapping a fallback feature.

DataLoader¶

A dataloader is for iterating over a dataset and generate mini-batches.

DataLoader

Multiprocessing DataLoader.

ItemSet¶

An item set is an iterable collection of items.

`ItemSet`	A wrapper of iterable data or tuple of iterable data.
`ItemSetDict`	Dictionary wrapper of ItemSet.

ItemSampler¶

An item sampler is for sampling items from an item set.

`ItemSampler`	A sampler to iterate over input items and create subsets.
`DistributedItemSampler`	A sampler to iterate over input items and create subsets distributedly.

MiniBatch¶

A mini-batch is a collection of sampled subgraphs and their corresponding features. It is the basic unit for training a GNN model.

`MiniBatch`	A composite data class for data structure in the graphbolt.
`MiniBatchTransformer`	A mini-batch transformer used to manipulate mini-batch.

NegativeSampler¶

A negative sampler is for sampling negative items from mini-batches.

`NegativeSampler`	A negative sampler used to generate negative samples and return a mix of positive and negative samples.
`UniformNegativeSampler`	Sample negative destination nodes for each source node based on a uniform distribution.

SubgraphSampler¶

A subgraph sampler is for sampling subgraphs from a graph.

`SubgraphSampler`	A subgraph sampler used to sample a subgraph from a given set of nodes from a larger graph.
`SampledSubgraph`	An abstract class for sampled subgraph.
`NeighborSampler`	Sample neighbor edges from a graph and return a subgraph.
`LayerNeighborSampler`	Sample layer neighbor edges from a graph and return a subgraph.
`SampledSubgraphImpl`	Sampled subgraph of CSCSamplingGraph.
`InSubgraphSampler`	Sample the subgraph induced on the inbound edges of the given nodes.

FeatureFetcher¶

A feature fetcher is for fetching features from a feature store.

FeatureFetcher

A feature fetcher used to fetch features for node/edge in graphbolt.

CopyTo¶

This datapipe is for copying data to a device.

CopyTo

DataPipe that transfers each element yielded from the previous DataPipe to the given device.

Utilities¶

`fused_csc_sampling_graph`	Create a FusedCSCSamplingGraph object from a CSC representation.
`load_from_shared_memory`	Load a FusedCSCSamplingGraph object from shared memory.
`from_dglgraph`	Convert a DGLGraph to FusedCSCSamplingGraph.
`etype_str_to_tuple`	Convert canonical etype from string to tuple.
`etype_tuple_to_str`	Convert canonical etype from tuple to string.
`isin`	Tests if each element of elements is in test_elements.
`seed`	Set the random seed of Graphbolt.
`add_reverse_edges`	This function finds the reverse edges of the given edges and returns the composition of them.
`exclude_seed_edges`	Exclude seed edges with or without their reverse edges from the sampled subgraphs in the minibatch.
`compact_csc_format`	Relabel the row (source) IDs in the csc formats into a contiguous range from 0 and return the original row node IDs per type.
`unique_and_compact`	Compact a list of nodes tensor.
`unique_and_compact_csc_formats`	Compact csc formats and return unique nodes (per type).