Releases: FederatedAI/FATE-Flow
Releases · FederatedAI/FATE-Flow
Release v1.9.1
Bug Fixes
- Fix parameter inheritance when loading non-model modules from ModelLoader
- Fix job inheritance after adding or removing roles from training configuration
- Fix delimiter error in uploaded/downloaded data
- Fix anonymous feature name renewal
Release v1.9.0
Major Features and Improvements
- Support high availability and load balancing to improve system availability and stability
- Added support for site authentication and data set authority authentication, and supports hook mode for users to customize authentication schemes
- Component registration optimization, support participants to use different versions of algorithm components
- Upload, reader support feature anonymity, support specifying id column
- Scheduling optimization, asynchronous time-consuming operations, component scheduling performance improved by more than 5 times This optimization obvious benefits for multi-component tasks
- Added component ApiReader to get feature data by id
- Model storage optimization, support model data synchronization between local and other storage
- The scheduler now can obtain the error information from other participant's algorithm components
Release v1.8.0
Major Features and Improvements
- Optimize the model migration function to reduce user operation steps;
- Add version compatibility check in component center to support multiple parties to use different versions;
- Add data table disable/enable function, and support batch delete disable table
Release v1.7.2
Major Features and Improvements
- Separate the base connection address of the data storage table from the data table information, and compatible with historical versions;
- Optimize the component output data download interface.
Release v1.7.1
Major Features and Improvements
- Added the writer component, which supports exporting data to mysql and saving data as a new table;
- Added job reuse function, which supports the reuse of successful status components of historical tasks in new jobs;
- Optimize the time-consuming problem of submitting tasks and the time-consuming problem of stopping tasks;
- Component registration supports automatic setting of PYTHONPYTH.
Bug Fixes
- Fix the problem of OOM when uploading hdfs table;
- Fix the problem of incompatibility with the old version of serving;
- The parameter partitions of the toy test is set to 4, and a timeout prompt is added.
Release v1.7.0
Release 1.7.0
Major Features and Improvements
- Independent repository instead of all code in the main FATE repository
- Component registry, which can hot load many different versions of component packages at the same time
- Hot update of component parameters, component-specific reruns, automatic reruns
- Model Checkpoint to support task hot start, model deployment and other
- Data, Model and Cache can be reused between jobs
- Reader component supports more data sources, such as MySQL, Hive
- Realtime recording of dataset usage derivation routes
- Multi-party permission control for datasets
- Automatic push to reliable storage when model deployment, support Tencent Cloud COS, MySQL, Redis
- REST API authentication