Loss calculation problem

[onlyzer0]

Hello, the MtetaQA-kb code only outputs one entity with the highest probability. For questions with multiple entity answers, is this way of calculating the answer incomplete?

scores, idx = torch.max(e_score, dim = 1) # [bsz], [bsz]
match_score = torch.gather(answers, 1, idx.unsqueeze(-1)).squeeze().tolist()