Model loading at every inference

Hi,

I deployed an MXNet gluon model by creating a custom docker container.
While making inferences, model is getting loaded at almost every inference which is making cpu to utilize more memory for every inference.
With that I am getting “CPU out of memory” error while making 2 or more inferences at a time.

My question is: model should be loaded at the time of deploying and the loaded model will be used for inferences. But here why it’s loading at every inference?
I am following the “scikit bring your own container “ example. https://github.com/awslabs/amazon-sagemaker-examples/tree/master/advanced_functionality/scikit_bring_your_own

Can anyone please suggest why this happens and how to resolve this cpu memory issue.
I tried with both ml.m4.4xlarge and ml.c4.4xlarge instances. Model weights file size is 537MB. And input image size is 1920 x 1472.

Thanks in Advance,
Harathi