PYRA: Parallel Yielding Re-Activation for Training-Inference Efficient Task Adaptation
Fuente:
arXiv
Guardado en:
| Autores principales: | , , , , , , , , |
|---|---|
| Formato: | Preprint |
| Publicado: |
2024
|
| Materias: | |
| Acceso en línea: | |
| Etiquetas: |
Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
|
| _version_ | 1866911960121999360 |
|---|---|
| author | Xiong, Yizhe Chen, Hui Hao, Tianxiang Lin, Zijia Han, Jungong Zhang, Yuesong Wang, Guoxin Bao, Yongjun Ding, Guiguang |
| author_facet | Xiong, Yizhe Chen, Hui Hao, Tianxiang Lin, Zijia Han, Jungong Zhang, Yuesong Wang, Guoxin Bao, Yongjun Ding, Guiguang |
| contents | Recently, the scale of transformers has grown rapidly, which introduces considerable challenges in terms of training overhead and inference efficiency in the scope of task adaptation. Existing works, namely Parameter-Efficient Fine-Tuning (PEFT) and model compression, have separately investigated the challenges. However, PEFT cannot guarantee the inference efficiency of the original backbone, especially for large-scale models. Model compression requires significant training costs for structure searching and re-training. Consequently, a simple combination of them cannot guarantee accomplishing both training efficiency and inference efficiency with minimal costs. In this paper, we propose a novel Parallel Yielding Re-Activation (PYRA) method for such a challenge of training-inference efficient task adaptation. PYRA first utilizes parallel yielding adaptive weights to comprehensively perceive the data distribution in downstream tasks. A re-activation strategy for token modulation is then applied for tokens to be merged, leading to calibrated token features. Extensive experiments demonstrate that PYRA outperforms all competing methods under both low compression rate and high compression rate, demonstrating its effectiveness and superiority in maintaining both training efficiency and inference efficiency for large-scale foundation models. Our code is available at https://github.com/THU-MIG/PYRA. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2403_09192 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | PYRA: Parallel Yielding Re-Activation for Training-Inference Efficient Task Adaptation Xiong, Yizhe Chen, Hui Hao, Tianxiang Lin, Zijia Han, Jungong Zhang, Yuesong Wang, Guoxin Bao, Yongjun Ding, Guiguang Computer Vision and Pattern Recognition Recently, the scale of transformers has grown rapidly, which introduces considerable challenges in terms of training overhead and inference efficiency in the scope of task adaptation. Existing works, namely Parameter-Efficient Fine-Tuning (PEFT) and model compression, have separately investigated the challenges. However, PEFT cannot guarantee the inference efficiency of the original backbone, especially for large-scale models. Model compression requires significant training costs for structure searching and re-training. Consequently, a simple combination of them cannot guarantee accomplishing both training efficiency and inference efficiency with minimal costs. In this paper, we propose a novel Parallel Yielding Re-Activation (PYRA) method for such a challenge of training-inference efficient task adaptation. PYRA first utilizes parallel yielding adaptive weights to comprehensively perceive the data distribution in downstream tasks. A re-activation strategy for token modulation is then applied for tokens to be merged, leading to calibrated token features. Extensive experiments demonstrate that PYRA outperforms all competing methods under both low compression rate and high compression rate, demonstrating its effectiveness and superiority in maintaining both training efficiency and inference efficiency for large-scale foundation models. Our code is available at https://github.com/THU-MIG/PYRA. |
| title | PYRA: Parallel Yielding Re-Activation for Training-Inference Efficient Task Adaptation |
| topic | Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2403.09192 |