Clover-2: Accurate Inference for Regressive Lightweight Speculative Decoding

TL;DR

Clover-2 introduces an improved RNN-based draft model for lightweight speculative decoding, achieving high accuracy and efficiency in large language model inference, outperforming existing methods.

Contribution

Clover-2 enhances lightweight speculative decoding by combining architectural improvements and knowledge distillation, matching attention-based models' accuracy with minimal computational cost.

Findings

Clover-2 outperforms existing methods across multiple architectures.

Achieves comparable accuracy to attention decoder models.

Demonstrates robustness and efficiency in experiments.

Abstract

Large Language Models (LLMs) frequently suffer from inefficiencies, largely attributable to the discord between the requirements of auto-regressive decoding and the architecture of contemporary GPUs. Recently, regressive lightweight speculative decoding has garnered attention for its notable efficiency improvements in text generation tasks. This approach utilizes a lightweight regressive draft model, like a Recurrent Neural Network (RNN) or a single transformer decoder layer, leveraging sequential information to iteratively predict potential tokens. Specifically, RNN draft models are computationally economical but tend to deliver lower accuracy, while attention decoder layer models exhibit the opposite traits. This paper presents Clover-2, an advanced iteration of Clover, an RNN-based draft model designed to achieve comparable accuracy to that of attention decoder layer models while maintaining minimal computational overhead. Clover-2 enhances the model architecture and incorporates knowledge distillation to increase Clover's accuracy and improve overall efficiency. We conducted experiments using the open-source Vicuna 7B and LLaMA3-Instruct 8B models. The results demonstrate that Clover-2 surpasses existing methods across various model architectures, showcasing its efficacy and robustness.