inorder to reduce model size and compute, can a self-supervised model be distilled to predict only a single object of interest, say cars. By feeding it a few examples of cars, and then prune anything that doesn't largely contribute to the prediction of "cars".
Hi there,
inorder to reduce model size and compute, can a self-supervised model be distilled to predict only a single object of interest, say cars. By feeding it a few examples of cars, and then prune anything that doesn't largely contribute to the prediction of "cars".