Cross Compile YOLO from CPU to GPU (Python code)

We have a image object recognition service hosted in AWS written in python using Yolo darknet that uses CPU. This REST service takes image as input and returns objects detected with their labels, confidence level etc. Internally this service uses YOLO running on a CPU based docker container. When we try to run multiple REST API request we get "requestbusy on dataset" error signifying that YOLO CPU cannot run two instances in parallel. We want to run something described in this article to start using GPU for YOLO instead of CPU ([login to view URL]) what needs to be done: Cross compile current Python code to support GPU and deploy in our existing AWS cluster (Update existing Docker container). No change to logic or API format. The current code needs to run on GPU and execute in parallel. Performance: We would need to process 100 image API/REST requests in parallel and respond within 1 sec for each image request independently. Deployment: Deploy as docker container in AWS kubernetes GPU cluster, Need recomendation if we can use standard NVIDIA CUDA based containers.