A research team trained a model in JAX and now wants to serve it on Vertex AI. None of the prebuilt prediction images include JAX. What is the best way to deploy it?
Show answer and explanation
Answer: B
Vertex AI lets you supply your own prediction image when no prebuilt container covers your framework. The image must install the required libraries, load the saved model, and serve predictions and health checks on the routes Vertex AI expects. Porting to TensorFlow adds risky rewrites, CSV weights cannot reproduce the forward pass, and retraining in scikit-learn changes the model family.
Why the other options are wrong
- A. Porting a JAX network to TensorFlow is a full rewrite that risks changing behaviour, and it is unnecessary when a custom image can run JAX directly.
- C. Weights exported as plain numbers lack the computation graph, so no serving framework can run inference from them, and this is not a deployment method.
- D. Retraining in a different framework produces a different model with different results, so it abandons the trained network instead of deploying it.
Exam tip. If no prebuilt prediction container matches your framework, a custom container is the standard answer.