indices[0,0] = 3046 is not in [0, 2681)

在用keras训练deepFM时，评估的时候出现这个错误。说明评估的数据维度和训练的维度不一致，建议对sparse_features、dense_features重新划分。

部分代码如下：

cclicol = ['position', 'hour', 'advert_place', 'province_id', 'W', 'H', 'click_nums_all',
               'click_nums_yesterday', 'user_modelMake', 'holiday_clicks', 'weekend_clicks', 'workday_clicks',
               'ad_click_all', 'ad_click_yesterday']

 # sparse_features = cclicol
 # dense_features = cclicol
 # 稀疏
 sparse_features = ['hour', 'advert_place', 'province_id', 'holiday_clicks', 'weekend_clicks', 'workday_clicks']
 # 稠密
 dense_features = ['position', 'W', 'H', 'click_nums_all', 'click_nums_yesterday', 'user_modelMake', 'ad_click_all', 'ad_click_yesterday']

 sparse_feature_columns = [SparseFeat(feat, vocabulary_size=pdtrain[feat].max() + 1)
                           for i, feat in enumerate(sparse_features)]
 dense_feature_columns = [DenseFeat(feat, 1)
                          for feat in dense_features]

https://blog.csdn.net/qq_41904729/article/details/115763492

https://blog.csdn.net/u011585024/article/details/88938194