A Python library for neural architecture search
naszilla
A repository to compare many popular NAS algorithms seamlessly across three popular benchmarks (NASBench 101, 201, and 301). You can implement your own NAS algorithm, and then easily compare it with eleven algorithms across three benchmarks.
Installation
Clone this repository and install its requirements (which includes nasbench,
nas-bench-201, and nasbench301).
It may take a few minutes.
git clone https://github.com/naszilla/naszilla
cd naszilla
cat requirements.txt | xargs -n 1 -L 1 pip install
pip install -e .
You might need to replace line 32 of src/nasbench301/surrogate_models/surrogate_models.py
with a new path to the configspace file:
self.config_loader = utils.ConfigLoader(os.path.expanduser('~/naszilla/src/nasbench301/configspace.json'))
Next, download the nas benchmark datasets (either with the terminal commands below,
or from their respective websites
(nasbench,
nas-bench-201, and
nasbench301).
The versions recommended for use with naszilla are nasbench_only108.tfrecord,