AutoCompress: Automated Hardware-Guided Model Compression
cs.LG, cs.AI
Submitted: 2026-11-05
Updated: 2026-11-05
License: http://creativecommons.org/licenses/by/4.0/
The gist: LLM pruning must preserve quality while meeting hardware- and workload-specific constraints on latency, throughput, and memory.
Terminology
Abstract
LLM pruning must preserve quality while meeting hardware- and workload-specific constraints on latency, throughput, and memory. Existing pruning workflows often target parameter counts, requiring practitioners to manually select model sizes that meet latency, throughput, and GPU memory limits. However, parameter count alone does not predict these metrics, which depend on the model architecture, workload, hardware, and serving engine. We introduce AutoCompress, a hardware-guided compression framework that directly incorporates such constraints into differentiable architecture search. AutoCompress jointly optimizes global and per-layer structure to produce heterogeneous architectures, guided by serving-engine cost models validated against end-to-end latency, throughput, and memory measurements. We evaluate AutoCompress on long-context local coding, batched datacenter serving, and latency-sensitive in-game interaction. Compared to the state-of-the-art Minitron pruning framework, AutoCompress improves average accuracy across eight benchmarks by 6.57, 2.35, and 0.57 percentage points on RTX 5070, RTX 5080, and H100, respectively. AutoCompress also compresses Nemotron 3.5 Super from 120B to 74B parameters for single-B200 BF16 serving at 128K context, retaining 95% of its mean benchmark score.
Sources
- Multi-Objective Structured Pruning of LLMs for Latency and Model Size Optimization
- Puzzle: Distillation-Based NAS for Inference-Optimized LLMs
- Nemotron-Labs-3-Puzzle-75B-A9B: Compressing Hybrid MoE LLMs
- Training Verifiers to Solve Math Word Problems
- Distilling the Knowledge in a Neural Network
- Efficient Memory Management for Large Language Model Serving with PagedAttention
- Accurate Sublayer Pruning for Large Language Models by Exploiting Latency and Tunability Information
- Generalizing Verifiable Instruction Following
- SlimQwen: Exploring the Pruning and Distillation in Large MoE Model Pre-training
- Instruction-Following Evaluation for Large Language Models
- TraceLab: Characterizing Coding Agent Workloads for LLM Serving
Related papers
- A Physics-Regulated Neural Framework for Learning 3D Grain Growth Dynamics
- FlexMoE: Learning One-for-All Multi-Budget Expert Pruning for Flexible MoE Deployment
- OptiMind: Improving Small Language Models for Optimization Modeling through Formulation Guidance and Multi-Turn Learning
- Staying in the Sweet Spot: Boost Reasoning Evolution via Responsive Rollout Difficulty Manipulation
- Examining the Impact of Data and Training Methods on Emergent and Subliminal Misalignment
- Rethinking Black-Box On-Policy Distillation: A Rubric-based Analysis