Quick Start¶

Quick Start From API¶

Run experiments:¶

Running an existing baseline model on an existing benchmark

from openhgnn import Experiment
experiment = Experiment(model='RGCN', dataset='acm4GTN', task='node_classification', gpu=-1, lr=0.05, hidden_dim=64,
                        max_epoch=30, num_layers=3)
experiment.run()

Hyper-parameter optimization¶

Run an experiment with optuna

from openhgnn import Experiment
def search_space(trial):
    return {
        "lr": trial.suggest_categorical("lr", [1e-3, 5e-3, 1e-2]),
        "hidden_dim": trial.suggest_categorical("hidden_dim", [32, 64]),
        "dropout": trial.suggest_uniform("dropout", 0.0, 0.5),
        'num_layers': trial.suggest_int('num_layers', 2, 3)
    }
experiment = Experiment(model='RGCN', dataset='acm4GTN', task='node_classification', gpu=-1,
                        hpo_search_space=search_space, hpo_trials=20)
experiment.run()

OpenHGNN will determine hyperparameters in the following order:

If hpo_search_space is not None, search for the best hyperparameter by optuna. This is controlled by func_search() in ./openhgnn/auto/hpo.py. Please refer here for more details.
Otherwise, if --use_best_config is enabled, load the best hyperparameters built within OpenHGNN. The configurations are in ./openhgnn/utils/best_config.py.
Otherwise, load the hyperparameters in ./openhgnn/config.ini.

Quick Start From Source¶

Besides using API, you can also directly run the source code of OpenHGNN.

>>> python main.py -m model_name -d dataset_name -t task_name -g 0 --use_best_config
usage: main.py [-h] [--model MODEL] [--task TASK] [--dataset DATASET]
           [--gpu GPU] [--use_best_config]

Optional arguments are as follows:

--model MODEL, -m MODEL name of models. Please refer to the model list for supported models and their names.

--task TASK, -t TASK name of task. Please refer to tasks for the list of tasks.

--dataset DATASET, -d DATASET name of datasets. Please refer to datasets for the list of datasets.

--gpu GPU, -g GPU controls which GPU you will use. If you do not have GPU, set -g -1.

--use_best_config use the best config OpenHGNN has found in the dataset with the model. If you want to set the different hyper-parameter, modify the config.ini file manually.

--load_from_pretrained will load the model from a default checkpoint.

e.g.:

python main.py -m GTN -d imdb4GTN -t node_classification -g 0 --use_best_config