ProfilingAgent: Profiling-Guided Agentic Reasoning for Adaptive Model Optimization

Fuente: arXiv
Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Jafari, Sadegh, Sarkar, Aishwarya, Bilwal, Mohiuddin, Jannesari, Ali
Format: Preprint
Veröffentlicht: 2025
Schlagworte:
Online-Zugang:
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866912574279254016
author Jafari, Sadegh
Sarkar, Aishwarya
Bilwal, Mohiuddin
Jannesari, Ali
author_facet Jafari, Sadegh
Sarkar, Aishwarya
Bilwal, Mohiuddin
Jannesari, Ali
contents Foundation models face growing compute and memory bottlenecks, hindering deployment on resource-limited platforms. While compression techniques such as pruning and quantization are widely used, most rely on uniform heuristics that ignore architectural and runtime heterogeneity. Profiling tools expose per-layer latency, memory, and compute cost, yet are rarely integrated into automated pipelines. We propose ProfilingAgent, a profiling-guided, agentic approach that uses large language models (LLMs) to automate compression via structured pruning and post-training dynamic quantization. Our modular multi-agent system reasons over static metrics (MACs, parameter counts) and dynamic signals (latency, memory) to design architecture-specific strategies. Unlike heuristic baselines, ProfilingAgent tailors layer-wise decisions to bottlenecks. Experiments on ImageNet-1K, CIFAR-10, and CIFAR-100 with ResNet-101, ViT-B/16, Swin-B, and DeiT-B/16 show pruning maintains competitive or improved accuracy (about 1% drop on ImageNet-1K, +2% gains for ViT-B/16 on smaller datasets), while quantization achieves up to 74% memory savings with <0.5% accuracy loss. Our quantization also yields consistent inference speedups of up to 1.74 times faster. Comparative studies with GPT-4o and GPT-4-Turbo highlight the importance of LLM reasoning quality for iterative pruning. These results establish agentic systems as scalable solutions for profiling-guided model optimization.
format Preprint
id arxiv_https___arxiv_org_abs_2509_05584
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle ProfilingAgent: Profiling-Guided Agentic Reasoning for Adaptive Model Optimization
Jafari, Sadegh
Sarkar, Aishwarya
Bilwal, Mohiuddin
Jannesari, Ali
Machine Learning
Computer Vision and Pattern Recognition
Performance
Foundation models face growing compute and memory bottlenecks, hindering deployment on resource-limited platforms. While compression techniques such as pruning and quantization are widely used, most rely on uniform heuristics that ignore architectural and runtime heterogeneity. Profiling tools expose per-layer latency, memory, and compute cost, yet are rarely integrated into automated pipelines. We propose ProfilingAgent, a profiling-guided, agentic approach that uses large language models (LLMs) to automate compression via structured pruning and post-training dynamic quantization. Our modular multi-agent system reasons over static metrics (MACs, parameter counts) and dynamic signals (latency, memory) to design architecture-specific strategies. Unlike heuristic baselines, ProfilingAgent tailors layer-wise decisions to bottlenecks. Experiments on ImageNet-1K, CIFAR-10, and CIFAR-100 with ResNet-101, ViT-B/16, Swin-B, and DeiT-B/16 show pruning maintains competitive or improved accuracy (about 1% drop on ImageNet-1K, +2% gains for ViT-B/16 on smaller datasets), while quantization achieves up to 74% memory savings with <0.5% accuracy loss. Our quantization also yields consistent inference speedups of up to 1.74 times faster. Comparative studies with GPT-4o and GPT-4-Turbo highlight the importance of LLM reasoning quality for iterative pruning. These results establish agentic systems as scalable solutions for profiling-guided model optimization.
title ProfilingAgent: Profiling-Guided Agentic Reasoning for Adaptive Model Optimization
topic Machine Learning
Computer Vision and Pattern Recognition
Performance
url https://arxiv.org/abs/2509.05584