CoLA: Compute-Efficient Pre-Training of LLMs via Low-Rank Activation

Code

• github.com/alvin-zyl/cola
  OfficialIn paperpytorch★ 25

Abstract

The full-size MLPs and the projection layers in attention introduce tremendous model sizes of large language models (LLMs), imposing extremely demanding needs of computational resources in the pre-training stage. However, we empirically observe that the activations of pre-trained LLMs exhibit low-rank property. Motivated by such observations, we propose CoLA and its memory-efficient implementation, CoLA-M, to replace these full-size layers with compute-efficient auto-encoders that naturally enforce low-rank activations throughout training. This fundamental architectural change eliminates the activation redundancy and significantly boosts model capacity and training efficiency. Experiments on LLaMA models with 60 million to 7 billion parameters show that CoLA reduces the computing cost by 2 and improves training throughput by 1.86 while maintaining full-rank level performance. CoLA-M further squeezes memory cost without sacrificing throughput, offering a pre-training approach with collectively superior parameter, computing, and memory efficiency. The LLMs produced are also 2 smaller, enabling faster inference with lower memory cost on resource-constrained platforms.

Tasks

Reproductions