This is a containerized model of YOLO built to run and train with CUDA. It has functionality to convert coco datasets to yolo datasets and output an evaluation as well.
To use each of the commands, an example is provided below for ease of use. The arguments are (mode, model size (n,s,m,l,x), epochs/model directory)
docker build -t yolo_cuda .
docker run --rm --gpus all -v /your/data/folder:/app/data yolo_cuda convert n 0
docker run --gpus all --rm -v /your/data/folder:/app/data yolo_cuda train m 20
docker run --rm --gpus all -v /your/data/folder:/app/data yolo_cuda eval n /app/data/yolo_annotations/models_and_results/model.pt