XFL 发行版 - Gitee.com

2024-02-27 18:04

最后提交信息为： support horizontal xgboost

v1.4.0

92121d5

2023-04-21 14:20

Release v1.4.0

最后提交信息为： update (#39)

v1.3.0

3989d3f

2022-11-09 15:30

v1.3.0

* Feat:
	* Operators:
		* Vertical：linear_regression, sampler;
		* Horizontal: bert, densenet, vgg, woe_iv;
		* Local: data_split, data_statistic, feature_preprocess;
	* New horizontal templates for TensorFlow, PaddlePaddle;
	* Vertical XGBoost supports category features and offline inference, the output model file changes to json format；
	* Vertical K-Means supports new initial method kmeans++；
* Improvement:
	* Support auto configuration of the assist trainer for vertical federation operator, e.g. vertical K-Means, no  need to provide configuration file for the assist trainer;
	* Simplify the vertical federation operator configuration, the same parameters only need to be configured on the label trainer side;
	* Optimize code of horizontal NbAFL、one-time pad;
* Fix:
	* Fix data alignment error in aggregation when transfering large amounts of data at once;
	* Fix label and feature data confusion in vertical K-Means under certain conditions ;  
	* Fix an index error in vertical pearson;
	* Fix incorrect bin number for 2^16 bins in vertical XGBoost;
	* Fix training error when goss is set to false and label trainer has no feature in vertical XGBoost.
* BREAKING CHANGE:
	* Configurations for local and vertical operators have been optimized or revised, old version may not be compatible.

最后提交信息为： [feat] (#32)