Hi there,
I’m about to finish the MLOps course and I wanted to get my hands a little bit dirty with a project using TFX pipeline.
But, I was wondering why ExampleGen generates only a train & eval split of the data instead of train, eval and test split?
I’m a little bit lost on how am I going to test my model and which data to use if the model has already seen the eval set during training?
looking forward to hearing from you!
Which lab are you referring to?
Dear @Aly_SEGNANE,
Welcome to the community.
ExampleGen generates a train and eval split by default because it aligns with standard machine learning practices and provides a solid foundation for model development and evaluation. If you need a test split in addition to the train and eval splits, you can customize ExampleGen or incorporate additional components into your TFX pipeline to accommodate your specific requirements.
No, it’s not a lab of the course. I just wanted to play with TFX on my own.
Thank you for confirming the source of your question. Moving forward, kindly tag such questions with general so that mentors don’t have to look for a notebook before answering.
Happy learning.