Ruyi2 Technical Report
This work provides a more efficient deployment strategy for LLMs, particularly for organizations facing high computational costs, by enabling a "Train Once, Deploy Many" paradigm.
This paper addresses the high deployment costs and latency of Large Language Models (LLMs) by introducing Ruyi2, an adaptive model designed for efficient variable-depth computation. Ruyi2, built on a stable "Familial Model" using 3D parallel training, achieves a 2-3 times speedup over its predecessor, Ruyi, while maintaining performance comparable to same-sized Qwen3 models.
Large Language Models (LLMs) face significant challenges regarding deployment costs and latency, necessitating adaptive computing strategies. Building upon the AI Flow framework, we introduce Ruyi2 as an evolution of our adaptive model series designed for efficient variable-depth computation. While early-exit architectures offer a viable efficiency-performance balance, the Ruyi model and existing methods often struggle with optimization complexity and compatibility with large-scale distributed training. To bridge this gap, Ruyi2 introduces a stable "Familial Model" based on Megatron-LM. By using 3D parallel training, it achieves a 2-3 times speedup over Ruyi, while performing comparably to same-sized Qwen3 models. These results confirm that family-based parameter sharing is a highly effective strategy, establishing a new "Train Once, Deploy Many" paradigm and providing a key reference for balancing architectural efficiency with high-performance capabilities.