Alternatives to TFX for time series project pipelines

At the moment, TFX appears to not support order-sensitive data, with more details in this Github thread. In the meantime, is there a viable alternative for end2end time series project pipeline?

You can use Dataset and save the prepared data to disk. Batches of data can be loaded on the fly to train the model.

I’m not sure about the specifics of your situation, but maybe defining a custom ExampleGen with just class CustomTimeSeriesExampleGen(…) in tfx.proto might be useful since you can define how to split the data without shuffling.