Requires Python 3.12 or above. Would recommend JAX 0.4.17 because 0.4.18 to 0.4.23 appears to sometimes yield CUDA internal errors randomly at runtime.
pip install git+https://github.com/cemlyn007/gpbo.git#egg=gpbo
Clone the repository at https://github.com/cemlyn007/gpbo and run:
pip install .
Clone the repository at https://github.com/cemlyn007/gpbo and run:
pip install -e .[dev]
If using Visual Studio Code with Pylance, you will want to run instead:
pip install -e .[dev] --config-settings editable_mode=strict
To get the type hinter to find the package.
You will need to install the example dependencies by running:
pip install -e .[example]
python examples/gp.py --objective_function=univariate --use_x64 --iterations 48
python examples/gp.py --objective_function=six_hump_camel --use_x64 --iterations 256
python examples/gp.py --objective_function=mnist_1d --use_x64 --iterations 256
python examples/gp.py --objective_function=mnist_2d --iterations 135
python examples/bo.py --objective_function=univariate --use_x64 --iterations 16
python examples/bo.py --objective_function=six_hump_camel --use_x64 --iterations 48
All commands are expected to be run from the root of the repository.
Run this command to generate the MNIST dataset:
./scripts/make_data.sh
Run this command to generate the results, passing gpu or cpu as the first argument to specify the device to use:
./scripts/run_experiments.sh gpu
To export the 3D results to be visualised using VTK via something like Paraview:
./scripts/export_3d.sh
Run this command to generate the MNIST dataset:
sbatch --chdir=$(pwd) ./scripts/make_data.sh
Run this command to generate the results, passing gpu or cpu as the first argument to specify the device to use:
sbatch --chdir=$(pwd) ./scripts/run_experiments.sh gpu