Frac-Connections: Fractional Extension of Hyper-Connections
This addresses memory consumption issues for deep learning practitioners, but it is incremental as it builds on Hyper-Connections.
The paper tackles the memory inefficiency of Hyper-Connections in deep networks by proposing Frac-Connections, which divides hidden states instead of expanding width, and shows that it significantly outperforms residual connections in large-scale language tasks, such as a 7B MoE model trained on 3T tokens.
Residual connections are central to modern deep learning architectures, enabling the training of very deep networks by mitigating gradient vanishing. Hyper-Connections recently generalized residual connections by introducing multiple connection strengths at different depths, thereby addressing the seesaw effect between gradient vanishing and representation collapse. However, Hyper-Connections increase memory access costs by expanding the width of hidden states. In this paper, we propose Frac-Connections, a novel approach that divides hidden states into multiple parts rather than expanding their width. Frac-Connections retain partial benefits of Hyper-Connections while reducing memory consumption. To validate their effectiveness, we conduct large-scale experiments on language tasks, with the largest being a 7B MoE model trained on up to 3T tokens, demonstrating that Frac-Connections significantly outperform residual connections.