Model soups: averaging weights of multiple fine-tuned models improves accuracy without increasing inference time

TL;DR

This paper introduces 'model soups', a method of averaging weights of multiple fine-tuned models to improve accuracy and robustness without increasing inference costs, demonstrating state-of-the-art results across various tasks.

Contribution

The paper proposes a novel weight-averaging technique called model soups that enhances model performance and robustness in fine-tuning large pre-trained models.

Findings

Model soups improve accuracy over individual fine-tuned models.

Model soups achieve state-of-the-art results on ImageNet with ViT-G.

The approach extends to multiple tasks and improves out-of-distribution and zero-shot performance.

Abstract

The conventional recipe for maximizing model accuracy is to (1) train multiple models with various hyperparameters and (2) pick the individual model which performs best on a held-out validation set, discarding the remainder. In this paper, we revisit the second step of this procedure in the context of fine-tuning large pre-trained models, where fine-tuned models often appear to lie in a single low error basin. We show that averaging the weights of multiple models fine-tuned with different hyperparameter configurations often improves accuracy and robustness. Unlike a conventional ensemble, we may average many models without incurring any additional inference or memory costs -- we call the results "model soups." When fine-tuning large pre-trained models such as CLIP, ALIGN, and a ViT-G pre-trained on JFT, our soup recipe provides significant improvements over the best model in a hyperparameter sweep on ImageNet. The resulting ViT-G model, which attains 90.94% top-1 accuracy on ImageNet, achieved a new state of the art. Furthermore, we show that the model soup approach extends to multiple image classification and natural language processing tasks, improves out-of-distribution performance, and improves zero-shot performance on new downstream tasks. Finally, we analytically relate the performance similarity of weight-averaging and logit-ensembling to flatness of the loss and confidence of the predictions, and validate this relation empirically. Code is available at https://github.com/mlfoundations/model-soups.