Fusing finetuned models for better pretraining
This provides a low-cost alternative to expensive pretraining for machine learning practitioners, though it is incremental as it builds on existing fine-tuning methods.
The paper tackles the high cost of pretraining by fusing multiple fine-tuned models through weight averaging, resulting in a base model that surpasses pretrained model performance and is less dependent on the target task.
Pretrained models are the standard starting point for training. This approach consistently outperforms the use of a random initialization. However, pretraining is a costly endeavour that few can undertake. In this paper, we create better base models at hardly any cost, by fusing multiple existing fine tuned models into one. Specifically, we fuse by averaging the weights of these models. We show that the fused model results surpass the pretrained model ones. We also show that fusing is often better than intertraining. We find that fusing is less dependent on the target task. Furthermore, weight decay nullifies intertraining effects but not those of fusing.