• Get Started
  • Dataset
  • Features
  • Tutorial
Download
GitHub
  • Get Started
  • Dataset
  • Features
  • Tutorial
Download
GitHub
  • Get Started
  • Dataset

    • Introduction
    • Download
    • Overview
  • Features

    • Basic Properties
    • Routability
    • IR drop
    • Graph
    • Timing
  • Tutorial
  • Change Log
  • FAQ
  • License

Introduction

CircuitNet is an open-source dataset dedicated to machine learning (ML) applications in electronic design automation (EDA). We have collected more than 20K samples from versatile runs of commercial design tools based on open-source designs with various features for multiple ML for EDA applications.

Available Datasets

We currently maintain three variations of datasets: CircuitNet-N28, CircuitNet-N14, and CircuitNet-ISPD15. All datasets are collected from commercial design tools, but differ in underlying designs and technology nodes.

CircuitNet-N28

  • Design Basis: RISC-V designs
  • Technology: 28nm planar technology
  • Status: Most comprehensive support for all tasks
  • Applications: Full support for all prediction tasks

CircuitNet-N14

  • Design Basis: Multiple designs including RISC-V, GPU, and ML accelerator
  • Technology: 14nm FinFET technology
  • Current Support: Congestion prediction, IR drop prediction, Net delay prediction
  • Future Plans: DRC prediction support coming soon

CircuitNet-ISPD15

  • Design Basis: ISPD2015 contest benchmark
  • Current Support: Congestion prediction
  • Future Plans: DRC prediction support coming soon

Feature Support Matrix

The following table shows the feature support status for each dataset:

DatasetCongestionDRCIR DropNet Delay
CircuitNet-N28✅✅✅✅
CircuitNet-N14✅❌✅✅
CircuitNet-ISPD15✅❌N/AN/A

Legend: ✅ = Supported, ❌ = Not Supported, N/A = Not Applicable

Directory Structure

All datasets follow a similar directory structure to enable flexible feature combinations and custom applications.

Feature Organization

📁 .
├─ 📁 routability_features
│  ├─ 📄 cell_density
│  ├─ 📁 congestion
│  │  ├─ 📁 congestion_early_global_routing
│  │  │  ├─ 📁 overflow_based
│  │  │  │  ├─ 📄 congestion_eGR_horizontal_overflow
│  │  │  │  └─ 📄 congestion_eGR_vertical_overflow
│  │  │  └─ 📁 utilization_based
│  │  │     ├─ 📄 congestion_eGR_horizontal_util
│  │  │     └─ 📄 congestion_eGR_vertical_util
│  │  └─ 📁 congestion_global_routing
│  │     ├─ 📁 overflow_based
│  │     │  ├─ 📄 congestion_GR_horizontal_overflow
│  │     │  └─ 📄 congestion_GR_vertical_overflow
│  │     └─ 📁 utilization_based
│  │        ├─ 📄 congestion_GR_horizontal_util
│  │        └─ 📄 congestion_GR_vertical_util
│  ├─ 📁 DRC
│  │  ├─ 📄 DRC_all
│  │  └─ 📄 DRC_seperated
│  ├─ 📄 macro_region
│  └─ 📁 RUDY
│     ├─ 📄 RUDY
│     ├─ 📄 RUDY_long
│     ├─ 📄 RUDY_short
│     ├─ 📄 RUDY_pin
│     └─ 📄 RUDY_pin_long
└─ 📁 IR_drop_features
   ├─ 📄 power_i
   ├─ 📄 power_s
   ├─ 📄 power_sca
   ├─ 📄 power_all
   ├─ 📄 power_t
   └─ 📄 IR_drop

Feature Usage

  • Features are organized in separate directories to enable:

    • Custom feature combinations
    • Flexible preprocessing methods
    • Task-specific feature selection
  • Provided scripts can help you:

    • Preprocess the raw features
    • Combine features for specific tasks
    • Prepare training and testing datasets
  • You are encouraged to:

    • Implement your own preprocessing methods
    • Experiment with different feature combinations
    • Develop novel feature engineering approaches

Support

If you encounter any issues:

  • Open an issue in our GitHub repository
  • Contact us via email
  • Check our FAQ page for common questions
Next
Download