Deploying models in Production with good performance


(QWERTY1) #1

Hey guys, after reading the forum posts on deploying model to production, I came around some different frameworks which are really helpful in deploying the machine learning models. You can deploy using spark for real time performance and low latency. But the following frameworks are also good.

  1. https://polyaxon.com/ - An open source platform for reproducible machine learning at scale.
  2. http://clipper.ai/ - Good for low latency and retraining the model.

All the above frameworks are open source.
Do comment if found any other frameworks.


Exposing DL models as api's/microservices
(Even Oldridge) #2

Polyaxon looks like it’s more for training than inference, but clipper looks very interesting. Here’s the paper on it https://arxiv.org/pdf/1612.03079.pdf

There’s a presentation on it as well from Spark Summit 2017.

Looks very promising. Thanks for sharing!