API reference

Auto-generated reference for the public surface of GraphNetz. The reference is organised by module so deep links stay stable as the package grows; the top-level convenience imports below cover everything you’ll typically need.

• graphnetz.datasets
• graphnetz.models
• graphnetz.benchmark
• graphnetz.training
• graphnetz.plotting

Top-level convenience imports

Importable directly from graphnetz:

Models

GCN	Two-layer Graph Convolutional Network.
GAT	Two-layer Graph Attention Network.
GIN	Graph Isomorphism Network for graph-level prediction.
GraphSAGE	Two-layer GraphSAGE for node-level prediction.
GraphTransformer	Two-layer graph transformer based on TransformerConv.
DGI	Deep Graph Infomax for unsupervised node representation learning.

Datasets

CATEGORIES	dict() -> new empty dictionary dict(mapping) -> new dictionary initialized from a mapping object's (key, value) pairs dict(iterable) -> new dictionary initialized as if via: d = {} for k, v in iterable: d[k] = v dict(**kwargs) -> new dictionary initialized with the name=value pairs in the keyword argument list. For example: dict(one=1, two=2).
Netz	Netzschleuder network dataset.
download_all_networks_netz	Download and process every network in a Netzschleuder dataset.
list_datasets	Return loader names organized by category and task.

Benchmark

BENCHMARK_TASKS	dict() -> new empty dictionary dict(mapping) -> new dictionary initialized from a mapping object's (key, value) pairs dict(iterable) -> new dictionary initialized as if via: d = {} for k, v in iterable: d[k] = v dict(**kwargs) -> new dictionary initialized with the name=value pairs in the keyword argument list. For example: dict(one=1, two=2).
BenchmarkReport	Structured outcome of a multi-seed benchmark run.
ModelSpec	How to instantiate a model and which task tasks it supports.
Task	A single benchmark task_type: a dataset loader plus its training task.
iter_benchmark_tasks	Flatten BENCHMARK_TASKS to a list, optionally filtered by category/task.
plot_benchmark	Grouped bar chart with mean ± CI error bars.
register_model	Register a model with the benchmark dispatcher.
register_task	Register task under category in BENCHMARK_TASKS.
run_benchmark	Run a benchmark across one or more (model, task, seed) combinations.
task_from_dataset	Wrap an already-loaded dataset as a Task.
unregister_task	Remove a previously registered task; returns it, or None if absent.

Training utilities

train_node_classification	Train a node classifier with Planetoid-style train/val/test masks.
train_graph_classification	Train a graph-level classifier.
train_graph_regression	Train a graph-level regressor (MSE loss, MAE on val).
train_link_prediction	Train a link predictor with binary cross-entropy on RandomLinkSplit.
train_dgi	Train a Deep Graph Infomax model (unsupervised).
train_node_degree_regression	Self-supervised node-level regression: predict log node degree.
train_relational_link_prediction	Train a relational link predictor (DistMult) on knowledge graph triples.

Plotting

figure	Create a sized figure.
panel_label	Add a bold panel label (a, b, ...) to an axis.
plot_grouped_bars	Grouped bar chart from a {group: {series: value}} mapping.
plot_history	Plot a training history dict.
save_figure	Save fig to one path stem in multiple formats; returns the saved paths.
set_plot_style	Apply the rcParams and color cycle.