SpecExec: Massively Parallel Speculative Decoding for Interactive LLM Inference on Consumer Devices

TL;DR

This paper introduces SpecExec, a parallel speculative decoding method enabling efficient inference of large language models on consumer GPUs with RAM offloading, achieving significant speedups over traditional methods.

Contribution

SpecExec is a novel parallel decoding approach that leverages token probability distributions to enable fast LLM inference on consumer hardware with offloaded parameters.

Findings

Achieves 4-6 tokens/sec with 4-bit quantization

Achieves 2-3 tokens/sec with 16-bit weights

Enables large LLM inference on consumer GPUs

Abstract

As large language models gain widespread adoption, running them efficiently becomes crucial. Recent works on LLM inference use speculative decoding to achieve extreme speedups. However, most of these works implicitly design their algorithms for high-end datacenter hardware. In this work, we ask the opposite question: how fast can we run LLMs on consumer machines? Consumer GPUs can no longer fit the largest available models (50B+ parameters) and must offload them to RAM or SSD. When running with offloaded parameters, the inference engine can process batches of hundreds or thousands of tokens at the same time as just one token, making it a natural fit for speculative decoding. We propose SpecExec (Speculative Execution), a simple parallel decoding method that can generate up to 20 tokens per target model iteration for popular LLM families. It utilizes the high spikiness of the token probabilities distribution in modern LLMs and a high degree of alignment between model output probabilities. SpecExec takes the most probable tokens continuation from the draft model to build a "cache" tree for the target model, which then gets validated in a single pass. Using SpecExec, we demonstrate inference of 50B+ parameter LLMs on consumer GPUs with RAM offloading at 4-6 tokens per second with 4-bit quantization or 2-3 tokens per second with 16-bit weights.