Currently, Span exists to represent groupings of data, but if a Span of data needs to be cleaned, there exists no way to track these changes beyond rerunning the entire pipeline on this changed data.
This CL adds support for a new Example artifact property, Version to the ExampleGen component. The ExampleGen driver will attempt to find the latest Span, then find the latest Version for that Span, through parsing a split's pattern, and will provide the correct split patterns to the ExampleGen executor for processing and conversion to Example artifacts.
该提问来源于开源项目:tensorflow/tfx