It is possible to use tf.Transform with other models frameworks Tensorflow?

In my company, we use XGBoost a lot to solve simples ML problems. After week two lectures, I’m wondering if the tf.Transformer can be used to avoid train-serving skew even if the final model is not built using Tensorflow.

Is it possible?

Hi @lucasrolim

Yes, it is possible, even though I do think it requires the right skills and some custom work.
A very nice example of a place where TFX and tf.Transform is used at large, in combination with “others” ML algorithms, like XGBoost, is Spotify.
You can read by their words here: MLOPs@Spotify