hyper-parameters-configuration.txt

# ChemProt - the pre-trained model: BioBERT-large
{
    "seed": 1,
    "epoch": 10.0,
    "learning_rate": 1e-05,
    "n_gpu": 2,
    "num_of_datasets": 1,
    "per_device_train_batch_size": 8,
    "warmup_ratio": 0.0,
    "weight_decay": 0.0
}

# DDI - the pre-trained model: PubMedBERT-fulltext
{
    "seed": 1,
    "epoch": 10.0,
    "learning_rate": 3e-05,
    "n_gpu": 2,
    "num_of_datasets": 1,
    "per_device_train_batch_size": 8,
    "warmup_ratio": 0.0,
    "weight_decay": 0.0
}

# GAD - the pre-trained model: PubMedBERT-fulltext
{
    "seed": 1,
    "epoch": 3.0,
    "learning_rate": "1e-05",
    "n_gpu": 2,
    "num_of_datasets": 1,
    "per_device_train_batch_size": 4,
    "warmup_ratio": 0.0,
    "weight_decay": 0.0
}

# EU-ADR - the pre-trained model: BioBERT-base
{
    "seed": 1,
    "epoch": 3.0,
    "learning_rate": "1e-05",
    "n_gpu": 2,
    "num_of_datasets": 10,
    "per_device_train_batch_size": 4,
    "warmup_ratio": 0.0,
    "weight_decay": 0.0
}

# PPI (AIMed, BioInfer) - the pre-trained model: BioBERT-base
{
    "seed": 1,
    "epoch": 10.0,
    "learning_rate": "5e-05",
    "n_gpu": 2,
    "num_of_datasets": 10,
    "per_device_train_batch_size": 16,
    "warmup_ratio": 0.0,
    "weight_decay": 0.0
}

# PPI (HPRD50, IEPA, LLL)- the pre-trained model: BioBERT-base
{
    "seed": 1,
    "epoch": 10.0,
    "learning_rate": "5e-05",
    "n_gpu": 2,
    "num_of_datasets": 10,
    "per_device_train_batch_size": 8,
    "warmup_ratio": 0.0,
    "weight_decay": 0.0
}