All notable changes to this repository will be documented in this file.
The format is based on Keep a Changelog.
Latest Release: 0.0.6
- Introduce BaseIngestion and GraphDBConfig (#115)(#116)
- Add retry logic when creating dataproc cluster (#124)
- Introduce
should_include_isolated_nodes_in_trainingfield in shared config (#134)
- Config Validation now also checks GiglResourceConfig and PreprocessedMetadata (#123)(#125)
- Simplified Dockerfiles to be 40-50% smaller (#198)
- Corrected invalid transductive node classification test (#128)
- Corrected some inconsistencies in component cleanup logic (#196)
- Introduce TaskMetadata and GraphMetadata in GbmlConfig (#830)
- Sphinx Support (#19)
stop_afterparameter in KFP pipeline (#49)- UDLAnchorBasedSupervisionEdge Split Strategy (#17)
- Added build and push docker image script (#109)
- SortedDict Class for unsorted proto fields used in HGS (#150)
- Add SimpleHGN model for HGS (#122)
- Logger Refactor (#29)
- Unsupervised Node Anchor (UNA) -> Node Anchor (NA) Refactor (#40)
- GiGL Modeling Task Spec Infer/Forward Refactor (#43)
- UNALP Batch + Sample Definitions (#32)
- RNN Samples paths and introduce GbmlConfigPbWrapper (#10)
- Complex Modeling Task Spec - Testing (#61)
- Complex Modeling Task Spec - Task Spec(#60)
- Complex Modeling Task Spec - Task Initialization Changes (#71)
- Complex Modeling Task Spec - Utility Functions for Training (#57, #62)
- HGS Support for Trainer (#89)
- HGS Support for Inferencer - Proto Changes (#121)
- HGS Support for Inferencer - Codebase Changes (#113)
- HGS Support for Trainer/Inferencer - Mocking/Testing Changes (#110)
- Deprecation of
UserDefLabelType; consolidation intoEdgeUsageType(#18) - Deprecate cora_assets.py and toy_graph.py in favour of
PassthroughPreprocessorConfigForMockedAssets(#25)
- Make feature order determinisitc in FeatureEmbeddingLayer (#23)
- Introduce KFPOrchestrator
- Enable CPU training
- Cost Tooling Script
- Torch, TFT, TF, PyG upgrades + mac arm64 support (#807)
- None