Brief summary of the paper on Model soups: Averaging weights of multiple fine-tuned models improves accuracy without increasing inference time