taobao-mnn/Qwen3-VL-4B-Thinking-Eagle3

English | 中文

EAGLE-3 Draft Model for Qwen3-VL-4B-Thinking

Model Overview

This repository contains an EAGLE-3 style draft model specifically trained to accelerate the inference of the Qwen3-VL-4B-Thinking large language model.

This is not a standalone model. It must be used in conjunction with its corresponding base model (Qwen3-VL-4B-Thinking) within a speculative decoding framework to achieve significant speedups in text generation.

• Base Model: Qwen3-VL-4B-Thinking
• Model Architecture: EAGLE-3 (Speculative Decoding Draft Model)
• Primary Benefit: Accelerates text generation throughput by 1.5x to 2.5x without compromising the generation quality of the base model.

What is EAGLE?

EAGLE (Extrapolative A* Generative Language Engine) is an advanced speculative decoding method. It uses a small draft model to generate a sequence of draft tokens in parallel. These tokens are then verified by the larger, more powerful base model in a single forward pass. If the draft is accepted, the generation process advances multiple steps at once, leading to a substantial increase in speed.

This model serves as the "draft model" in this process. Its average acceptance length (acc_length) on standard benchmarks is approximately 1.87 tokens (with 4 draft tokens), meaning on average, it helps the base model advance nearly 2 tokens per verification step.

Performance

This model was evaluated on a diverse set of benchmarks. The acc_length (average number of accepted draft tokens) indicates the efficiency of the acceleration. A higher value is better.

Benchmark	acc_length (num_draft_tokens=4)	acc_length (num_draft_tokens=8)
gsm8k	1.78	1.83
math500	1.71	1.75
humaneval	1.70	1.74
ceval	1.59	1.64
cmmlu	1.56	1.58
mtbench	1.53	1.55
Average	~1.65	~1.68

These results demonstrate consistent and effective acceleration across various tasks, including coding, math, and general conversation.

Training Details

• Training Framework: This model was trained using SpecForge, an open-source framework for speculative decoding research.
• Training Data: The model was trained on the EagleChat dataset. Available on Hugging Face and ModelScope.
• Training Duration: The model was trained for 3 epochs on 4x H20 GPUs, which took 75 hours and totaled 200 H20 GPU-hours.

---

适用于 Qwen3-VL-4B-Thinking 的 EAGLE-3 草稿模型

模型简介

本仓库包含一个 EAGLE-3 风格的草稿模型，专为加速 Qwen3-VL-4B-Thinking 大语言模型的推理而训练。

请注意：这是一个非独立模型。它必须与对应的基座模型 (Qwen3-VL-4B-Thinking) 在推测解码 (speculative decoding) 框架下配合使用，才能实现显著的文本生成加速效果。

• 基座模型: Qwen3-VL-4B-Thinking
• 模型架构: EAGLE-3 (推测解码草稿模型)
• 核心优势: 在不牺牲基座模型生成质量的前提下，将文本生成吞吐量提升 1.5 到 2.5 倍。

什么是 EAGLE？

EAGLE (Extrapolative A* Generative Language Engine) 是一种先进的推测解码方法。它利用一个轻量的草稿模型并行生成一系列草稿词元 (draft tokens)，然后由更大、更强的基座模型通过单次前向传播进行验证。如果草稿被接受，生成过程就能一次性前进多个步骤，从而实现显著的速度提升。

性能表现

本模型在一系列多样化的评测基准上进行了评估。acc_length (平均接受的草稿词元数) 反映了加速的效率，数值越高越好。

评测基准 (Benchmark)	acc_length (num_draft_tokens=4)	acc_length (num_draft_tokens=8)
gsm8k	1.78	1.83
math500	1.71	1.75
humaneval	1.70	1.74
ceval	1.59	1.64
cmmlu	1.56	1.58
mtbench	1.53	1.55
平均值	~1.65	~1.68

这些结果表明，该模型在编码、数学和通用对话等不同任务上都能提供稳定且高效的加速效果。

训练细节

• 训练框架: 本模型使用开源推测解码研究框架 SpecForge 进行训练。
• 训练数据: 训练数据使用了 EagleChat 数据集。您可以在 Hugging Face 或 ModelScope 上获取该数据集。
• 训练耗时: 训练使用 4x H20 训练 3 轮，耗时 75 小时，共 300 H20 卡时。