About Result #59
Description
INFO test result: OrderedDict([('recall@5', 0.000685), ('recall@10', 0.001005), ('recall@50', 0.00182), ('recall@200', 0.003395), ('ndcg@5', 0.0003777), ('ndcg@10', 0.0004817), ('ndcg@50', 0.0006559), ('ndcg@200', 0.0008877)])
实验结果为什么会这么差，几乎随机排序，计算完的item_feature：item_feature SHAPE=torch.Size([407083, 2048]) MEAN=-0.091796875 STD=3.03125
predict函数开始和结束部分进行调试，日志：
[start:eval-rank0] item_feature SHAPE=torch.Size([407083, 2048]) MEAN=-0.091796875 STD=3.03125
[end:eval-rank0] item_feature SHAPE=torch.Size([407083, 2048]) MEAN=-0.000667572021484375 STD=0.0220947265625
请问以上日志有什么问题吗？实在是不理解为什么实验结果是这个样子