TY - GEN
T1 - SmartTrim
T2 - Joint 30th International Conference on Computational Linguistics and 14th International Conference on Language Resources and Evaluation, LREC-COLING 2024
AU - Wang, Zekun
AU - Chen, Jingchang
AU - Zhou, Wangchunshu
AU - Liang, Haichao Zhu Jiafeng
AU - Shan, Liping
AU - Liu, Ming
AU - Xu, Dongliang
AU - Yang, Qing
AU - Qin, Bing
N1 - Publisher Copyright:
© 2024 ELRA Language Resource Association: CC BY-NC 4.0.
PY - 2024
Y1 - 2024
N2 - Despite achieving remarkable performance on various vision-language tasks, Transformer-based Vision-Language Models (VLMs) suffer from redundancy in inputs and parameters, significantly hampering their efficiency in real-world applications. Moreover, the degree of redundancy in token representations and model parameters, such as attention heads, varies significantly for different inputs. In light of the challenges, we propose SmartTrim, an adaptive acceleration framework for VLMs, which adjusts the computational overhead per instance. Specifically, we integrate lightweight modules into the original backbone to identify and prune redundant token representations and attention heads within each layer. Furthermore, we devise a self-distillation strategy to enhance the consistency between the predictions of the pruned model and its fully-capacity counterpart. Experimental results across various vision-language tasks consistently demonstrate that SmartTrim accelerates the original model by 2-3 times with minimal performance degradation, highlighting the effectiveness and efficiency compared to previous approaches. Code will be available at https://github.com/kugwzk/SmartTrim.
AB - Despite achieving remarkable performance on various vision-language tasks, Transformer-based Vision-Language Models (VLMs) suffer from redundancy in inputs and parameters, significantly hampering their efficiency in real-world applications. Moreover, the degree of redundancy in token representations and model parameters, such as attention heads, varies significantly for different inputs. In light of the challenges, we propose SmartTrim, an adaptive acceleration framework for VLMs, which adjusts the computational overhead per instance. Specifically, we integrate lightweight modules into the original backbone to identify and prune redundant token representations and attention heads within each layer. Furthermore, we devise a self-distillation strategy to enhance the consistency between the predictions of the pruned model and its fully-capacity counterpart. Experimental results across various vision-language tasks consistently demonstrate that SmartTrim accelerates the original model by 2-3 times with minimal performance degradation, highlighting the effectiveness and efficiency compared to previous approaches. Code will be available at https://github.com/kugwzk/SmartTrim.
KW - Adaptive Inference
KW - Dynamic Network
KW - Pruning
KW - Vision-Language Model
UR - https://www.scopus.com/pages/publications/85195962037
M3 - 会议稿件
AN - SCOPUS:85195962037
T3 - 2024 Joint International Conference on Computational Linguistics, Language Resources and Evaluation, LREC-COLING 2024 - Main Conference Proceedings
SP - 14937
EP - 14953
BT - 2024 Joint International Conference on Computational Linguistics, Language Resources and Evaluation, LREC-COLING 2024 - Main Conference Proceedings
A2 - Calzolari, Nicoletta
A2 - Kan, Min-Yen
A2 - Hoste, Veronique
A2 - Lenci, Alessandro
A2 - Sakti, Sakriani
A2 - Xue, Nianwen
PB - European Language Resources Association (ELRA)
Y2 - 20 May 2024 through 25 May 2024
ER -