C3W3-On Premise Distributed Training with MultiworkerMirroredStrategy

I want to distribute train my model using an on-premise cluster of machines with MultiWorkerMirroredStrategy within TFX environment to take advantage of Apache Beam for parallelizing compute. The examples in Week 3 of Course 3 are based on on cloud computing (on Google Colab, GCP etc.). I also have not encountered examples of distribute training within TFX environment which I was expecting from this course.

Any help on this? Am I missing anything here?


I’m guessing you are referring to C3 W3 Lab1 Distributed Training. There’s nothing that ties the training to a specific cloud provider. Please clarify and provide more context into which part of the notebook is specific to a cloud provider.