AutoPrune:利用 LLM 自动设计视觉令牌剪枝策略的 AI4AI 框架
原标题:An AI4AI Framework for Visual Token Pruning
AI 摘要
AutoPrune 是一个由 Hugging Face 论文提出的 AI4AI 框架,利用大型语言模型自动设计视觉令牌剪枝策略,通过领域特定语言和残差搜索公式实现高效剪枝。实验表明,在移除 94.4% 视觉令牌时,AutoPrune 保留了超过 99% 的全令牌性能,同时将 FLOPs 减少 9.9 倍,预填充延迟降低 6.4 倍。该框架在 14 个多模态基准和三个 MLLM 骨干上验证了有效性、效率和可迁移性。
正文节选
An AI4AI Framework for Visual Token Pruning Abstract AutoPrune uses large language models to automatically design visual-token pruning policies for multimodal models via a domain-specific language and residual search formulation, achieving high efficiency with minimal performance loss. Visual-token pruning can substantially reduce the inference cost of multimodal large language models (MLLMs), yet existing methods largely rely on fixed, handcrafted heuristics and costly expert trial and error. A