brainstorming
brainstorming
You MUST use this before any creative work - creating features, building components, adding functionality, modifying behavior, or planning anything new, in software or out of it (a talk, a business, a renovation).
Explorer les Skills
Développement · SKILL.md
graph-apis
graph-apisConstruct, inspect, transform, batch, and debug homogeneous and heterogeneous DGL graphs.
Independent GitHub snapshot · Automated classification · Original documentation governs
Documentation de la Skill
GitHubname: graph-apis description: "Construct, inspect, transform, batch, and debug homogeneous and heterogeneous DGL graphs." disable-model-invocation: true metadata: disco-role: operating license: Apache 2.0
DGL Graph APIs
Use this sub-skill when a task needs to create, inspect, transform, batch, or debug DGL homogeneous or heterogeneous graphs before model training or data loading.
Start Here
- For constructors, object relationships, feature views, idtype/device rules, conversions, blocks, batching, and subgraph APIs, read references/api-reference.md.
- For compact recipes, including homogeneous graphs, heterographs, canonical edge types, transforms, batching, and validation checks, read references/workflows.md.
- For failure diagnosis around canonical edge-type ambiguity, feature schema mismatches, device/idtype errors, zero-node assumptions, batching, and import/native-library errors, read references/troubleshooting.md.
- To verify a local DGL install can run the graph basics safely, run
python scripts/graph_smoke.pyfrom this sub-skill directory.
Route Elsewhere
- CSV ingestion,
DGLDataset,CSVDataset, graph caching, and persistent save/load depth belong in datasets-and-io. - Neighbor sampling, minibatch
DataLoader, and GraphBolt pipelines belong in dataloading-graphbolt. - Message passing,
update_all, built-in functions, GNN layers, and training loops belong in message-passing-training. - Distributed partitioning, launchers, and cluster/runtime tools belong in distributed-tools.