python - How would one go about creating a custom TFX component to perform feature engineering/building

Question

TFX Transform component operates with Tensors that cannot be executed eagerly within its preprocessing_fn. Therefore the values for each feature cannot be passed to a python function that generates new features based on those values.

I tried creating a custom component but failed to modify the input data as tf.Example objects can't be modified.

What is the recommended way to construct complex/custom features from within the pipeline?

question from:https://stackoverflow.com/questions/65893255/how-would-one-go-about-creating-a-custom-tfx-component-to-perform-feature-engine

Answer 1

Waitting for answers