... 共batch_size*num_beams個 beam_scores = torch.zeros((batch_size, ... 可更新的情況:數量未飽和或超過最差得分 self.beams.append((score, ...
確定! 回上一頁