Saved in:
| Main Authors: | , |
|---|---|
| Format: | Preprint |
| Published: |
2024
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2401.17790 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866917579580243968 |
|---|---|
| author | Menes, Thibaut Risser-Maroix, Olivier |
| author_facet | Menes, Thibaut Risser-Maroix, Olivier |
| contents | Model Soups, extending Stochastic Weights Averaging (SWA), combine models fine-tuned with different hyperparameters. Yet, their adoption is hindered by computational challenges due to subset selection issues. In this paper, we propose to speed up model soups by approximating soups performance using averaged ensemble logits performances. Theoretical insights validate the congruence between ensemble logits and weight averaging soups across any mixing ratios. Our Resource ADjusted soups craftINg (RADIN) procedure stands out by allowing flexible evaluation budgets, enabling users to adjust his budget of exploration adapted to his resources while increasing performance at lower budget compared to previous greedy approach (up to 4% on ImageNet). |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2401_17790 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | RADIN: Souping on a Budget Menes, Thibaut Risser-Maroix, Olivier Machine Learning Computer Vision and Pattern Recognition Model Soups, extending Stochastic Weights Averaging (SWA), combine models fine-tuned with different hyperparameters. Yet, their adoption is hindered by computational challenges due to subset selection issues. In this paper, we propose to speed up model soups by approximating soups performance using averaged ensemble logits performances. Theoretical insights validate the congruence between ensemble logits and weight averaging soups across any mixing ratios. Our Resource ADjusted soups craftINg (RADIN) procedure stands out by allowing flexible evaluation budgets, enabling users to adjust his budget of exploration adapted to his resources while increasing performance at lower budget compared to previous greedy approach (up to 4% on ImageNet). |
| title | RADIN: Souping on a Budget |
| topic | Machine Learning Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2401.17790 |