cm

conf.yml

@@ -30,8 +30,8 @@ max_seq_len: 20
 num_workers: 1
 
 ##K-fold params
-use_kfold : True #n개의 fold를 이용하여 k-fold를 진행한다.
-use_stratify : True
+use_kfold : False #n개의 fold를 이용하여 k-fold를 진행한다.
+use_stratify : False
 n_fold : 5
 split_by_user : False #k-fold를 수행할 dataset을 user 기준으로 split
 

dkt/dataloader.py

@@ -267,7 +267,7 @@ def __getitem__(self,index):
         seq_len = len(row[0])
         print(f'row 값 : {len(row)}')
 
-        # test, question, tag, correct, test_level_diff, tag_mean, tag_sum, ans_rate
+        # test, question, tag, correct, solve_time
         cate_cols = [row[i] for i in range(len(row))]
 
 

dkt/model.py

@@ -1306,7 +1306,7 @@ def __init__(self, args):
 
         self.hidden_dim = self.args.hidden_dim
         self.n_layers = self.args.n_layers
-        self.cont_cols=self.args.cont_cols
+        self.cont_cols=1
 
         # Embedding 
         # interaction은 현재 correct로 구성되어있다. correct(1, 2) + padding(0)
@@ -1348,28 +1348,29 @@ def init_hidden(self, batch_size):
 
     def forward(self, input):
 
-        test, question, tag, _, mask, interaction, solve_time,_ = input
-
+        test, question, tag, _, mask,interaction,solve_time, _ = input
+        
         batch_size = interaction.size(0)
 
         # Embedding
-
+        solve_time=solve_time.unsqueeze(-1).type(torch.FloatTensor).to(self.device)
         embed_interaction = self.embedding_interaction(interaction)
         embed_test = self.embedding_test(test)
         embed_question = self.embedding_question(question)
         embed_tag = self.embedding_tag(tag)
 
         embed_cont=self.cont_proj(solve_time)
-
+        # print(embed_cont.shape)
+        # print("-"*80)
         embed = torch.cat([embed_interaction,
                            embed_test,
                            embed_question,
                            embed_tag,], 2)
 
         X = self.comb_proj(embed)
-        print(X.shape,embed_cont.shape)
+        print("범주형과 연속형의 shape: ",X.shape,embed_cont.shape)
         X=torch.cat([X, embed_cont], 2)
-
+        # print("둘은 concat한 shape: ",X.shape)
         hidden = self.init_hidden(batch_size)
         out, hidden = self.lstm(X, hidden)
         out = out.contiguous().view(batch_size, -1, self.hidden_dim)

dkt/trainer.py

@@ -448,10 +448,9 @@ def get_model(args):
 # 배치 전처리 일반화
 def process_batch_v2(batch, args):
     # batch : load_data_from 에서 return 시킬 feature(컬럼)
-    # test, question,tag,correct, test_level_diff, tag_mean,tag_sum,ans_rate,mask = batch
+    # test, question,tag,correct, mask = batch
     # 규칙 : mask는 항상 맨 뒤임
     # print(type(batch))
-    # print(batch)
     test = batch[0]
     question = batch[1]
     tag = batch[2]