Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

ChatNeuroSim: An LLM Agent Framework for Automated Compute-in-Memory Accelerator Deployment and Optimization

Ming-Yen Lee, Shimeng Yu

TL;DR

ChatNeuroSim, a large language model (LLM)-based agent framework for automated CIM accelerator deployment and optimization, which automates the entire CIM workflow, including task scheduling, request parsing and adjustment, parameter dependency checking, script generation, and simulation execution and integrates the proposed CIM optimizer using design space pruning.

Abstract

Compute-in-Memory (CIM) architectures have been widely studied for deep neural network (DNN) acceleration by reducing data transfer overhead between the memory and computing units. In conventional CIM design flows, system-level CIM simulators (such as NeuroSim) are leveraged for design space exploration (DSE) across different hardware configurations and DNN workloads. However, CIM designers need to invest substantial effort in interpreting simulator manuals and understanding complex parameter dependencies. Moreover, extensive design-simulation iterations are often required to identify optimal CIM configurations under hardware constraints. These challenges severely prolong the DSE cycle and hinder rapid CIM deployment. To address these challenges, this work proposes ChatNeuroSim, a large language model (LLM)-based agent framework for automated CIM accelerator deployment and optimization. ChatNeuroSim automates the entire CIM workflow, including task scheduling, request parsing and adjustment, parameter dependency checking, script generation, and simulation execution. It also integrates the proposed CIM optimizer using design space pruning, enabling rapid identification of optimal configurations for different DNN workloads. ChatNeuroSim is evaluated on extensive request-level testbenches and demonstrates correct simulation and optimization behavior, validating its effectiveness in automatic request parsing and task execution. Furthermore, the proposed design space pruning technique accelerates CIM optimization process compared to no-pruning baseline. In the case study optimizing Swin Transformer Tiny under 22 nm technology, the proposed CIM optimizer achieves a 0.42$\times$-0.79$\times$ average runtime reduction compared to the same optimization algorithm without design space pruning.

ChatNeuroSim: An LLM Agent Framework for Automated Compute-in-Memory Accelerator Deployment and Optimization

TL;DR

ChatNeuroSim, a large language model (LLM)-based agent framework for automated CIM accelerator deployment and optimization, which automates the entire CIM workflow, including task scheduling, request parsing and adjustment, parameter dependency checking, script generation, and simulation execution and integrates the proposed CIM optimizer using design space pruning.

Abstract

Compute-in-Memory (CIM) architectures have been widely studied for deep neural network (DNN) acceleration by reducing data transfer overhead between the memory and computing units. In conventional CIM design flows, system-level CIM simulators (such as NeuroSim) are leveraged for design space exploration (DSE) across different hardware configurations and DNN workloads. However, CIM designers need to invest substantial effort in interpreting simulator manuals and understanding complex parameter dependencies. Moreover, extensive design-simulation iterations are often required to identify optimal CIM configurations under hardware constraints. These challenges severely prolong the DSE cycle and hinder rapid CIM deployment. To address these challenges, this work proposes ChatNeuroSim, a large language model (LLM)-based agent framework for automated CIM accelerator deployment and optimization. ChatNeuroSim automates the entire CIM workflow, including task scheduling, request parsing and adjustment, parameter dependency checking, script generation, and simulation execution. It also integrates the proposed CIM optimizer using design space pruning, enabling rapid identification of optimal configurations for different DNN workloads. ChatNeuroSim is evaluated on extensive request-level testbenches and demonstrates correct simulation and optimization behavior, validating its effectiveness in automatic request parsing and task execution. Furthermore, the proposed design space pruning technique accelerates CIM optimization process compared to no-pruning baseline. In the case study optimizing Swin Transformer Tiny under 22 nm technology, the proposed CIM optimizer achieves a 0.42-0.79 average runtime reduction compared to the same optimization algorithm without design space pruning.
Paper Structure (25 sections, 6 equations, 16 figures, 4 tables, 3 algorithms)

This paper contains 25 sections, 6 equations, 16 figures, 4 tables, 3 algorithms.

Table of Contents

  1. Introduction
  2. Background
  3. Compute-in-Memory (CIM) and CIM Simulators
  4. CIM Design Space Exploration (DSE)
  5. Machine Learning (ML) Algorithms for AI Accelerator DSE
  6. Large Language Model (LLM)-based Agent Framework for EDA
  7. ChatNeuroSim: LLM-based agent framework of automated CIM deployment and optimization
  8. Overall Framework
  9. Prompt Design for Agents
  10. CIM Optimizer with Design Space Pruning
  11. Evaluations and Case Studies
  12. Experimental Setup
  13. Case Study 1: Accuracy of Request Dataset on ChatNeuroSim
  14. Case Study 2: FoM Optimization on ResNet-50 and Swin-T without Design Space Pruning
  15. Case Study 3: Optimizations on Swin-T with and without Design Space Pruning under Different Algorithms
  16. ...and 10 more sections

Figures (16)

  • Figure 1: (a) Compute-in-Memory (CIM) design workflow, (b) Conventional design space exploration (DSE) workflow and proposed LLM-based DSE workflow.
  • Figure 2: Compute-in-Memory (CIM) simulator: NeuroSim framework.
  • Figure 3: ChatNeuroSim: Overall framework.
  • Figure 4: ChatNeuroSim: Prompt design for agents.
  • Figure 5: Dialogue example between the user and ChatNeuroSim.
  • ...and 11 more figures