Batch Processing with SFACE Question

I am using the SFace model for a project and I was wondering if anyone has made any batch processing framework so that if their are multiple faces in a frame at one time the faces don’t have to be run through with the .infer method one by one. I was looking to see if anyone has managed to do this in parallel where all the faces are ran through a different instance of the model at one time ? I am thinking about just creating three instances of the model and then running 1 face through each instance at the same time and if their are more than three faces in the video feed then the code just loops to get the rest of the faces. Would the be efficient or would that be wasting memory ?