INNER CODE UNIT · Python
train
UKPLab/gpl · gpl/train.py:36
def train(
path_to_generated_data: str,
output_dir: str,
mnrl_output_dir: str = None,
mnrl_evaluation_output: str = None,
do_evaluation: str = False,
evaluation_data: str = None,
evaluation_output: str = "output",
qgen_prefix: str = "qgen",
base_ckpt: str = "distilbert-base-uncased",
generator: str = "BeIR/query-gen-msmarco-t5-base-v1",
cross_encoder: str = "cross-encoder/ms-marco-MiniLM-L-6-v2",
batch_size_gpl: int = 32,
batch_size_generation: int = 32,
pooling: str = None,
max_seq_length: int = 350,
new_size: int = None,
queries_per_passage: int = 3,