MUDDFormer: Breaking Residual Bottlenecks in Transformers via Multiway Dynamic Dense Connections
Fuente:
arXiv
Saved in:
| Main Authors: | , , , |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866909625784205312 |
|---|---|
| author | Xiao, Da Meng, Qingye Li, Shengping Yuan, Xingyuan |
| author_facet | Xiao, Da Meng, Qingye Li, Shengping Yuan, Xingyuan |
| contents | We propose MUltiway Dynamic Dense (MUDD) connections, a simple yet effective method to address the limitations of residual connections and enhance cross-layer information flow in Transformers. Unlike existing dense connection approaches with static and shared connection weights, MUDD generates connection weights dynamically depending on hidden states at each sequence position and for each decoupled input stream (the query, key, value or residual) of a Transformer block. MUDD connections can be seamlessly integrated into any Transformer architecture to create MUDDFormer. Extensive experiments show that MUDDFormer significantly outperforms Transformers across various model architectures and scales in language modeling, achieving the performance of Transformers trained with 1.8X-2.4X compute. Notably, MUDDPythia-2.8B matches Pythia-6.9B in pretraining ppl and downstream tasks and even rivals Pythia-12B in five-shot settings, while adding only 0.23% parameters and 0.4% computation. Code in JAX and PyTorch and pre-trained models are available at https://github.com/Caiyun-AI/MUDDFormer . |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2502_12170 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | MUDDFormer: Breaking Residual Bottlenecks in Transformers via Multiway Dynamic Dense Connections Xiao, Da Meng, Qingye Li, Shengping Yuan, Xingyuan Machine Learning Artificial Intelligence Computation and Language We propose MUltiway Dynamic Dense (MUDD) connections, a simple yet effective method to address the limitations of residual connections and enhance cross-layer information flow in Transformers. Unlike existing dense connection approaches with static and shared connection weights, MUDD generates connection weights dynamically depending on hidden states at each sequence position and for each decoupled input stream (the query, key, value or residual) of a Transformer block. MUDD connections can be seamlessly integrated into any Transformer architecture to create MUDDFormer. Extensive experiments show that MUDDFormer significantly outperforms Transformers across various model architectures and scales in language modeling, achieving the performance of Transformers trained with 1.8X-2.4X compute. Notably, MUDDPythia-2.8B matches Pythia-6.9B in pretraining ppl and downstream tasks and even rivals Pythia-12B in five-shot settings, while adding only 0.23% parameters and 0.4% computation. Code in JAX and PyTorch and pre-trained models are available at https://github.com/Caiyun-AI/MUDDFormer . |
| title | MUDDFormer: Breaking Residual Bottlenecks in Transformers via Multiway Dynamic Dense Connections |
| topic | Machine Learning Artificial Intelligence Computation and Language |
| url | https://arxiv.org/abs/2502.12170 |