AI-300 Practice Question 75: Scoring Script Init
Microsoft Machine Learning Operations Engineer Associate (AI-300) · Test 8 · question 5 of 10
A custom scoring script for a managed online endpoint loads the model file from disk inside its run() function. Every request now takes several seconds. Where should the model be loaded to remove this delay?
Show answer and explanation
Answer:
A scoring script must define init() and run(). init() runs once when the container starts, so it is the place for expensive one-time work such as loading the model into a global variable. run() then only performs inference for each request (option D).
- A is wrong because the model would still be reloaded for every request. A longer timeout only prevents timeouts and does not remove the delay. - B is wrong because this guarantees that the load cost is paid on every single request. - C is wrong because an environment describes packages. It does not load a model into memory for the scoring code.
Reference: Deploy and score a machine learning model by using an online endpoint