Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Learning Library Cell Representations in Vector Space

Rongjian Liang, Yi-Chen Lu, Wen-Hao Liu, Haoxing Ren

TL;DR

Lib2Vec addresses the challenge of learning meaningful vector representations for library cells to enable ML-driven circuit analysis, by proposing a self-supervised framework that derives training data from Liberty files and evaluates semantics with automated regularity tests. The approach combines an attention-based model with separate functional and electrical branches to produce property- and arc-aware embeddings, validated through tests that reveal functional and electrical relationships such as $vector(BUF) - vector(INV) + vector(AND) \,\approx \, vector(NAND)$. Empirical results show Lib2Vec captures functional/electrical properties and provides significant gains in downstream circuit tasks, especially under limited labeled data, highlighting potential for circuit foundation models and cross-library transfer. This work thus enables data-efficient, transferable circuit representations and sets the stage for broader adoption in ML-assisted circuit design and optimization.

Abstract

We propose Lib2Vec, a novel self-supervised framework to efficiently learn meaningful vector representations of library cells, enabling ML models to capture essential cell semantics. The framework comprises three key components: (1) an automated method for generating regularity tests to quantitatively evaluate how well cell representations reflect inter-cell relationships; (2) a self-supervised learning scheme that systematically extracts training data from Liberty files, removing the need for costly labeling; and (3) an attention-based model architecture that accommodates various pin counts and enables the creation of property-specific cell and arc embeddings. Experimental results demonstrate that Lib2Vec effectively captures functional and electrical similarities. Moreover, linear algebraic operations on cell vectors reveal meaningful relationships, such as vector(BUF) - vector(INV) + vector(NAND) ~ vector(AND), showcasing the framework's nuanced representation capabilities. Lib2Vec also enhances downstream circuit learning applications, especially when labeled data is scarce.

Learning Library Cell Representations in Vector Space

TL;DR

Lib2Vec addresses the challenge of learning meaningful vector representations for library cells to enable ML-driven circuit analysis, by proposing a self-supervised framework that derives training data from Liberty files and evaluates semantics with automated regularity tests. The approach combines an attention-based model with separate functional and electrical branches to produce property- and arc-aware embeddings, validated through tests that reveal functional and electrical relationships such as . Empirical results show Lib2Vec captures functional/electrical properties and provides significant gains in downstream circuit tasks, especially under limited labeled data, highlighting potential for circuit foundation models and cross-library transfer. This work thus enables data-efficient, transferable circuit representations and sets the stage for broader adoption in ML-assisted circuit design and optimization.

Abstract

We propose Lib2Vec, a novel self-supervised framework to efficiently learn meaningful vector representations of library cells, enabling ML models to capture essential cell semantics. The framework comprises three key components: (1) an automated method for generating regularity tests to quantitatively evaluate how well cell representations reflect inter-cell relationships; (2) a self-supervised learning scheme that systematically extracts training data from Liberty files, removing the need for costly labeling; and (3) an attention-based model architecture that accommodates various pin counts and enables the creation of property-specific cell and arc embeddings. Experimental results demonstrate that Lib2Vec effectively captures functional and electrical similarities. Moreover, linear algebraic operations on cell vectors reveal meaningful relationships, such as vector(BUF) - vector(INV) + vector(NAND) ~ vector(AND), showcasing the framework's nuanced representation capabilities. Lib2Vec also enhances downstream circuit learning applications, especially when labeled data is scarce.

Paper Structure

This paper contains 18 sections, 5 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Challenges
  3. Contributions
  4. Library Cell Properties
  5. Lib2Vec Overview
  6. Definition and Test Sets for Cell Semantics
  7. Inverting Functionality Tests
  8. Functional Similarity Tests
  9. Electrical Similarity Tests
  10. Library Cell Self-Supervised Learning Scheme
  11. Model Architecture
  12. Utilizing Lib2Vec for Downstream Applications
  13. Experimental Results
  14. Lib2Vec Implementation Details
  15. Regularity Test Results
  16. ...and 3 more sections

Figures (5)

  • Figure 1: Overview of the lib2vec framework.
  • Figure 2: Functional similarity calculation for cells with identical input pins.
  • Figure 3: Accuracy comparison in (a) inverting functionality, (b) functional similarity and (c) electrical similarity test sets between random guess and Lib2Vec with various embedding sizes.
  • Figure 4: Visualization of cell representations.
  • Figure 5: Impacts of integrating Lib2Vec into ML models for different prediction tasks.