ML Model Crashes the Cloud Environment

0
Hi there, I have an ONNX model imported into Mendix which works smoothly when testing on local host but crashes when ran on a cloud environment. The crash fully takes the application offline and causes it to come back online about a minute afterwards Once deploying a package to the environment, the model will successfully run a few times but then constantly crash on subsequent calls. Any error handling on the call model action isn't possible as the application completely crashes, preventing the error handling flow to be run. Any help on what might be causing this or a way I could debug what is causing the Call ML Model microflow action to crash the application would be much appreciated.   Many thanks, Liam
asked
1 answers
1

So what I could tell from testing different models, this error occurs when the environment runs out of memory when attempting to load in the model.

 

For example, if you have a cloud environment with 1GiB of memory and your ONNX model size is 800MB, you're likely to run out of memory when the model mapping is loaded.

 

I can think of two solutions for this problem:

  1. Upgrade the cloud environment to have more memory
  2. Change your ONNX model to a model that takes up less storage.

 

Hope this helps anyone else who faces the same issue,

Liam

answered