v0.2.1

Summary

• advance the package version to 0.2.1 after the failed v0.2.0 publish mismatch
• restore a clean release path where repo version, latest tag, and PyPI can agree
• keep script bootstrap and artifact smoke contracts green for wheel and sdist releases

Verification

• metadata_consistency.py against refs/tags/v0.2.1
• pytest tests/test_metadata_consistency.py -q
• python -m build
• release_contract_scan.py --artifact-dir dist
• twine check dist/.whl dist/.tar.gz
• release_smoke.py --artifact-dir dist --kind all
• install_release_extras.py --artifact-dir dist --extras pyg dgl
• release_smoke.py --artifact-dir dist --kind all --interop-backend all

v0.2.0

• 重写了README
• 增加测试cli

v0.1.7: Normalize trainer profile payloads to one stable schema

Move the canonical simple-profiler shape into TrainingHistory so
profile payloads are normalized consistently whether they come from a
fresh fit, a restored history state, or future checkpoint-driven
reconstruction. Missing counters now default cleanly, derived values
are recomputed, and malformed profile sections fail with explicit
ValueError messages.

Constraint: The simple-profiler key order and required fields are already part of the repo's documented contract
Rejected: Keep profile schema logic only inside Trainer | leaves restored history payloads unsanitized and allows shape drift across entry points
Confidence: high
Scope-risk: narrow
Reversibility: clean
Directive: Keep TrainingHistory and Trainer on one shared profile schema source; do not duplicate profiler key lists in multiple modules again
Tested: python -m ruff check vgl tests/train; python -m mypy vgl; PYTHONPATH=. pytest -q tests/train/test_history.py tests/train/test_trainer_plus.py -k "profile or profiler or history"
Not-tested: Full repository pytest on this branch; resume flows that persist a completed profile and then continue training

vX.Y.Z: finalize data sampling compatibility surface

Highlights

This release expands the data loading, transform, dataset, and compatibility surface across vgl, with a focus on smoother migration from legacy imports and broader coverage for practical graph training
workflows.

• Added broader advanced dataloading support, including random-walk, Node2Vec, GraphSAINT, Cluster-GCN, and ShaDow-style sampling surfaces.
• Added public dataset coverage and example modules for Planetoid, TU datasets, GraphSAINT, and Cluster-GCN style workflows.
• Split and expanded transform exports into clearer public modules, including base, compose, feature, split, and structure-oriented transforms.
• Extended legacy compatibility shims for vgl.data, vgl.data.sampler, vgl.data.transform, vgl.data.plan, vgl.data.executor, vgl.data.requests, and vgl.data.materialize.
• Extended representative compatibility re-exports for legacy vgl.train and vgl.core import paths.
• Improved link prediction compatibility around query ids, negative sampling behavior, and related loader/split coverage.
• Improved docs and README guidance for preferred imports, compatibility imports, testing, and local release verification.

Compatibility

For new code, prefer importing from the domain packages directly:

• vgl.graph
• vgl.dataloading
• vgl.transforms
• vgl.engine

Legacy import paths remain available as compatibility shims to help existing projects migrate incrementally:

• vgl.core
• vgl.data
• vgl.data.sampler
• vgl.data.transform
• vgl.data.plan
• vgl.data.executor
• vgl.train

Included Examples

This release adds or strengthens example coverage for:

• Planetoid node classification
• TU graph classification
• GraphSAINT node classification
• Cluster-GCN node classification

Verification

Validated locally with:

• python -m pytest -q -> 1118 passed, 2 warnings
• python scripts/public_surface_scan.py -> 84/84 passed
• python scripts/full_scan.py -> 100/100 passed
• python scripts/docs_link_scan.py -> 11/11 passed