• 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

Download

Choose the appropriate dataset based on your target tasks:

TaskRecommended Dataset
Congestion PredictionRoutability Features
DRC Violation PredictionRoutability Features
IR Drop AnalysisIR Drop Features
Net Delay PredictionGraph Features / Timing Features

Available Datasets

CircuitNet-N28

  • Google Drive
  • Baidu Netdisk

CircuitNet-N14

  • 🤗 Hugging Face

Decompress and Preprocess

Note

The decompression process may take at least 15 minutes and require 200GB of storage space. Please ensure you have sufficient disk space before proceeding. For the CircuitNet-N14 on Hugging Face, the detailed decompress instructions are coming soon.

Routability/IR Drop Features

Step 1: Decompress the Dataset

Choose one of the following commands based on your needs:

# For Routability features
python decompress_routability.py

# For IR Drop features
python decompress_IR_drop.py

Important

Make sure your directory structure matches the one in Google Drive or Baidu Netdisk, and you are using the latest version of the script from the drive.

Step 2: Generate Training Set

Run the preprocessing script to generate the training set for your specific task:

python generate_training_set.py \
    --task [congestion/DRC/IR_drop] \
    --data_path [path_to_decompressed_dataset] \
    --save_path [path_to_save_output]

Step 3: Start Training

You can now:

  • Set up your own model for training
  • Use our tutorial code from the tutorial page
  • Check our GitHub repository for implementation details

Graph Features

Step 1: Decompress the Dataset

tar -xf PATH_TO_THE_FILE

Step 2: Construct Graph

Sample code for graph construction will be available soon.

Raw Data

We provide the following raw data formats (CircuitNet-N28) for custom feature extraction:

  • Netlist files
  • LEF/DEF files

Raw data for CircuitNet-N14 is coming soon. Currently available upon request.

Feature Extraction

You can use our feature extraction toolkit to:

  • Extract custom features from raw data
  • Process and transform the data for your specific needs
  • Integrate with your own machine learning pipeline

For detailed instructions on feature extraction, please refer to our feature documentation.

Note

Make sure to follow the data format specifications when working with raw data files. The feature extraction toolkit includes example scripts and documentation to help you get started.

Prev
Introduction
Next
Overview