def __init__(
self,
name,
extractive_class,
abstractive_class,
extractive_args,
abstractive_args,
):
super().__init__(name)
self.extractive = baselines.use(extractive_class, **extractive_args)
self.abstractive = baselines.use(abstractive_class, **abstractive_args)
args = parse_json_file(json_args.run_args_file)
# Load dataset
dataset = load_dataset(
args.dataset.name, split=args.dataset.split, cache_dir=args.dataset.cache_dir
)
# Compute baselines
scores = {}
for baseline in args.baselines:
print(f"Compute {baseline.baseline_class}...")
dataset, score = use(baseline.baseline_class, **baseline.init_kwargs).compute_rouge(
dataset,
args.dataset.document_column_name,
args.dataset.summary_colunm_name,
list(args.run.rouge_types.keys()),
**baseline.run_kwargs,
)
scores[baseline.init_kwargs["name"]] = score
# Save results
Path(args.run.hypotheses_folder).mkdir(parents=True, exist_ok=True)
utils.write_references(
dataset, args.run.hypotheses_folder, args.dataset.summary_colunm_name,
)
utils.write_documents(
dataset, args.run.hypotheses_folder, args.dataset.document_column_name,
)
utils.write_hypotheses(dataset, args.run.hypotheses_folder)
if args.run.csv_file != None: