in leaderboard/tasks.py [0:0]
def requires(self):
folds = ["dev"]
if Path(conf["squad"]["submission_predictions"]["test"]).exists():
folds.append("test")
for fold in folds:
for irt_type in IRT_TYPES:
for evaluation in IRT_EVALUATION_TYPES:
if "no_irt" in irt_type or "no_irt" in evaluation:
continue
# Only need a trained model for the CAT sampling stability
# experiment
if evaluation == "subject_heldout":
if irt_type == "2PL" and fold == "dev":
yield PyroTrainIRT(
evaluation=evaluation,
irt_type=irt_type,
device=self.device,
fold=fold,
)
else:
yield EvaluateIRT(
evaluation=evaluation, irt_type=irt_type, device=self.device, fold=fold,
)