Secretflow

Added

- Add Finetune and FedEval to SFXgboost
- Add SLModel support multi parties(>=2)

Changed

- FLModel supports most metrics of regression and classification

Fixed

- SLModel can be initialized without model.
- PSI doc typos.
- Fix the logic error bug of savemodel and loadmodel path in SLModel and FLModel

Added

- Add scorecard.
- Add replace/mode function to DataFrame.
- Add round function to VDataFrame.
- Add psi_join_csv and psi_join_df.
- Add preprocessing.LogroundTransformer.
- Add args to preprocessing.OneHotEncoder.

Changed

- Bump dependencies
- secretflow-ray to 2.0.0.dev2
- Update psi_df doc.
- Optimize sl_model by tf_funciton.
- Add curve parameter for ecdh psi.
- Protect biclassification, psi and pva with pyu object.
- Modify XgbModel predict api.

Fixed

- Raise exception if spu_fe.compile fails.
- Fix quantile security vulnerability.
- Fix woe bin bugs.
- Fix psi_join recv timeout.

Added

- omp_num_threads param for secretflow init().
- Regression and biclassification evaluation.
- Xgboost evaluation.
- Horizontal fl supports default naive aggreagte for metrics.
- PVA calculation.

Changed

- Remove graph util NodeDataLoader.

Fixed

- VDataFrame docstring.
- Remove dependencies
- dgl

Changed

- Get rid of import tensorflow/torch when import secretflow.

Added

- Add license file

Fixed

- Fix sl predict & remove reveal
- Fix typos in function docs.

Changed

- Bump dependencies
- TensorFlow to 2.10.0
- Jax to 0.3.17
- Jaxlib to 0.3.15

Changed

- Bump dependencies
- sf-heu to 0.2.0
- spu to 0.2.5

Fixed

- Missing requirements in dev-requirements.txt.