Enabling Performant and Flexible Model-Internal Observability for LLM Inference

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Yu, Nengneng, Xiong, Sixian, Zhao, Yibo, Wang, Wei, Liu, Zaoxing
Format: Preprint
Published: 2026
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866909034113662976
author Yu, Nengneng
Xiong, Sixian
Zhao, Yibo
Wang, Wei
Liu, Zaoxing
author_facet Yu, Nengneng
Xiong, Sixian
Zhao, Yibo
Wang, Wei
Liu, Zaoxing
contents Today's inference-time workloads increasingly depend on timely access to a model's internal states. We present DMI-Lib, a high-speed deep model inspector that treats internal observability as a first-class systems primitive, decoupling it from the inference hot path via an asynchronous observability substrate built from Ring^2, a GPU-CPU memory abstraction for capturing and staging tensors, and a policy-controlled host backend that exports them. DMI-Lib enables the placement of observation points across a rich space of internal signals and diverse inference backends while preserving serving optimizations and adhering to tight GPU memory budgets. Our experiments demonstrate that DMI-Lib incurs only 0.4%--6.8% overhead in offline batch inference and an average of 6% in moderate online serving, reducing latency overhead by 2x-15x compared to existing baselines with similar observability features. DMI-Lib is open-sourced at https://github.com/ProjectDMX/DMI.
format Preprint
id arxiv_https___arxiv_org_abs_2605_11093
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Enabling Performant and Flexible Model-Internal Observability for LLM Inference
Yu, Nengneng
Xiong, Sixian
Zhao, Yibo
Wang, Wei
Liu, Zaoxing
Machine Learning
Artificial Intelligence
Performance
Software Engineering
Systems and Control
Today's inference-time workloads increasingly depend on timely access to a model's internal states. We present DMI-Lib, a high-speed deep model inspector that treats internal observability as a first-class systems primitive, decoupling it from the inference hot path via an asynchronous observability substrate built from Ring^2, a GPU-CPU memory abstraction for capturing and staging tensors, and a policy-controlled host backend that exports them. DMI-Lib enables the placement of observation points across a rich space of internal signals and diverse inference backends while preserving serving optimizations and adhering to tight GPU memory budgets. Our experiments demonstrate that DMI-Lib incurs only 0.4%--6.8% overhead in offline batch inference and an average of 6% in moderate online serving, reducing latency overhead by 2x-15x compared to existing baselines with similar observability features. DMI-Lib is open-sourced at https://github.com/ProjectDMX/DMI.
title Enabling Performant and Flexible Model-Internal Observability for LLM Inference
topic Machine Learning
Artificial Intelligence
Performance
Software Engineering
Systems and Control
url https://arxiv.org/abs/2605.11093