def evaluate(args, model, tokenizer, prefix="", val_or_test="val"):
examples, predictions = predict(args,
model,
tokenizer,
prefix=prefix,
val_or_test=val_or_test)
# Compute the F1 and exact scores.
results, has_ans_eval, no_ans_eval = squad_evaluate(examples, predictions)
return results, has_ans_eval, no_ans_eval
def evaluate(args, model, tokenizer, prefix="", val_or_test="val"):
# prefix = epoch
examples, predictions, probabilites = predict(args,
model,
tokenizer,
prefix=prefix,
val_or_test=val_or_test)
# Compute the F1 and exact scores.
# predictions
results = squad_evaluate(examples,
predictions,
no_answer_probs=probabilites)
return results