Serve Model Azure Functions How-To

[luisquintanilla]

Serve Model Azure Functions How-To

Internal Review URL

#AB1484868

[JRAlexander]

LGTM!

[natke]

Make sure it is clear that these are alternatives

Pre-trained model.
- Use the ML.NET Sentiment Analysis tutorial to build your own model or (<-- or similar)
- Download this [pre-trained sentiment analysis machine learning model]

[natke]

This is a real nit but can we indent the code samples to align with the numbered list. I think this would increase readability

[natke]

LGTM. A couple of minor comments

[CESARDELATORRE]

This is wrong doing that on a static object because the PredictionEngine is not threadsafe.
_predictionEngine = _mlContext.Model.CreatePredictionEngine<SentimentData, SentimentPrediction>(_model);

[CESARDELATORRE]

You could probably use a static PredictionEnginePool because the pool is thread-safe.
Similar to how we use it in ASP.NET Core but without using dependency injection.

But I would really prefer to use dependency injection in the Azure Function in a similar way it is explained in this blog post:

https://blog.wille-zone.de/post/azure-functions-dependency-injection/

Then, the implementation using PredictionEnginePool would be very similar to the one recommended in ASP.NET Core WebAPI or apps.