sjain-stanford / deep-route

Training a deep FCN network in PyTorch to route circuit layouts

Deep-Route

This repository contains PyTorch implementation (pretrained weights provided) and dataset generation code for the paper

Training a Fully Convolutional Neural Network to Route Integrated Circuits
Sambhav R. Jain, Kye Okabe
arXiv-cs.CV (Computer Vision and Pattern Recognition) 2017

We design and train a deep, fully convolutional neural network that learns to route a circuit layout net with appropriate choice of metal tracks and wire class combinations. Inputs to the network are the encoded layouts containing spatial location of pins to be routed. After 15 fully convolutional stages followed by a score comparator, the network outputs 8 layout layers (corresponding to 4 route layers, 3 via layers and an identity-mapped pin layer) which are then decoded to obtain the routed layouts.

Proposed FCN Model

Training samples (left: data, right: labels) from the generated dataset

Install (Linux)

1. Fork this GitHub repository
2. Setup virtualenv and install dependencies
   • ./setup_virtualenv.sh
3. Install PyTorch
   • ./setup_pytorch.sh
4. Activate virtualenv, start Jupyter notebook
   • ./start_jupyter_env.sh

Generate Dataset

Run the script ./datagen/gen_data.py to generate training data of shape (N, 1, H, W) and labels of shape (N, 8, H, W) stored using HDF5 (h5py). Default parameters used for the paper are H = W = 32, and pin_range = (2, 6), but feel free to modify as desired. Generating 50,000 image dataset should take < 1 minute.

python ./datagen/gen_data.py
>> Enter the number of images to be generated: 50000
mv ./data/layout_data.hdf5 ./model/data/train_50k_32pix.hdf5

python ./datagen/gen_data.py
>> Enter the number of images to be generated: 10000
mv ./data/layout_data.hdf5 ./model/data/val_10k_32pix.hdf5

Train FCN Network (in PyTorch)

Switch to ./model/ dir and run the script ./train_fcn_pytorch.py to train the FCN model with default options, or use the switch --help to display a list of options and their defaults.

cd ./model/
python ./train_fcn_pytorch.py --help

usage: train_fcn_pytorch.py [-h] [--data PATH] [--batch_size N]
                            [--num_workers N] [--num_epochs N] [--use_gpu]
                            [--pretrained] [--lr LR] [--adapt_lr] [--reg REG]
                            [--print-freq N]

Deep-Route: Training a deep FCN network to route circuit layouts.

optional arguments:
  -h, --help       show this help message and exit
  --data PATH      path to dataset (default: ./data/)
  --batch_size N   mini-batch size (default: 100)
  --num_workers N  number of data loading workers (default: 4)
  --num_epochs N   number of total epochs to run (default: 200)
  --use_gpu        use GPU if available
  --pretrained     use pre-trained model
  --lr LR          initial learning rate (default: 5e-4)
  --adapt_lr       use learning rate schedule
  --reg REG        regularization strength (default: 1e-5)
  --print-freq N   print frequency (default: 10)

To run on GPU, provide switch --use_gpu. Best model parameters (based on F-1 score on validation set) are saved to ./model/training/ dir every epoch, along with loss and training curves. If the switch --pretrained is provided, model is pre-loaded with saved parameters before training. Pretrained weights (for batch size 10 and 100) are made available here. With --adapt_lr, a learning rate decay factor of 10 is applied every 30 epochs.

Cite

If you find this work useful in your research, please cite:

@article{jain2017route,
  title={Training a Fully Convolutional Neural Network to Route Integrated Circuits},
  author={Jain, Sambhav R and Okabe, Kye},
  journal={arXiv preprint arXiv:1706.08948},
  year={2017}
}