Main Hyper-Parameter Reference

Name	Description
activation	Activation function used in the model layers.
batch_size	Number of samples per batch during training.
bias	Whether to include bias terms in the model layers.
conv_dim	Dimension of convolution layers when the input and output dimensions must be the same.
concat	Whether to concatenate the outputs from multiple heads in multi-head attention.
dataset	Dataset to be used for training or evaluation.
dropout	Dropout rate applied to the model to prevent overfitting.
emb_dim	Dimension of the embedding layer.
epochs	Number of training epochs.
head_dim	Dimension of each attention head in multi-head attention mechanisms.
hidden_dim	Dimension of the hidden layers within the model.
in_dim	Dimension of the input data.
lr	Learning rate for training.
metadata	Metadata of graph or tabular data, including node and edge types, and other related information.
num_classes	Number of classes in the classification task.
num_feats	Number of features in the dataset.
num_heads	Number of attention heads in multi-head attention mechanisms.
num_layers	Number of layers in the model.
out_dim	Dimension of the model’s final output.
patience	Early stopping criterion, specifying the number of epochs to wait for improvement before halting training.
seed	Random seed for reproducibility of results.
wd	Weight decay parameter to regularize the model.