Code for "Optimal Transport Kernels for Sequential and Parallel Neural Architecture Search" at ICML2021
- tensorflow == 1.14.0
- pytorch == 1.2.0, torchvision == 0.4.0
- matplotlib, jupyter
- nasbench101 (follow the installation instructions here)
- nasbench201 (follow the installation instructions here)
To run on NASBench101, download nasbench_only108.tfrecord and place it in the top level folder of this repo.
To run on NASBench201, download NAS-Bench-201-v1_0-e61699.pth and place it in the top level folder of this repo.
python run_experiments/run_experiments_sequential.py
This will run the sequential NAS setting including the BO algorithm against several other sequential NAS algorithms on the NASBench101 search space.
python run_experiments/run_experiments_batch.py
This will run the batch NAS setting including the k-DPP quality algorithm against several other batch baseline algorithms on the NASBench201 search space.
To customize your experiment, open params.py. Here, you can change the hyperparameters and the algorithms to run.
We adapt the source code from BANANAS to enable the fair comparison with BANANAS and other baselines https://github.com/naszilla/bananas