Alink Versions Save

Release version 1.6.2

Optimize performance and fixed bugs.

Optimize performance and fixed bugs.

1. Add more rules about exception.
2. Add more graph algorithms.

1. Add pipeline model support in model stream.
2. Refine online learning.
3. Fixed some bugs.

1. Add partition in ak, csv and parquet source/sink.
2. Add model stream as initial model in online learning
3. Add lazyVizDive and lazyVizStatistics
4. Add hbase connector
5. Add custom path in resource plugin

1. Add datahub catalog.
2. Add Kernel Dense Estimate
3. Fixed some bugs

1. Add parquet source
2. Add Export2FileSinkStreamOp https://www.yuque.com/pinshu/alink_tutorial/book_java_3_2_5
3. Add Model File Path on predictop and pipeline.
4. Add TFRecordDataset source/sink
5. Add ONNX predict operators

1. Add xgboost wrapper using plugin
2. Add PyTorch model predictor, close #207.
3. Support pyalink on vvp. https://www.yuque.com/pinshu/alink_tutorial/flink_vvp
4. Support pyalink on dsw. https://www.yuque.com/pinshu/alink_tutorial/pai_dsw
5. Support pyalink on pai-designer. https://www.yuque.com/pinshu/alink_tutorial/pai_designer
6. Add serializer of MTable and Tensor types
7. Add code of pyalink, close #208.
8. Fix #199, #193.

1. Improve the performance of model stream.
2. Add ToTensor, ToVector, ToMTable and support tensor, vector, mtable types in csv source.
3. Keras sequential operators now support string/int types; Improve plugin mechanism in TF predictor.
4. Add redis to plugin and add lookup redis operator.
5. PyAlink: StreamOperator print function now supports specifying port.