create dataloaders for the other 2 tasks and called evaluation

multitask_classifier.py

@@ -13,7 +13,7 @@
 from datasets import SentenceClassificationDataset, SentencePairDataset, \
     load_multitask_data, load_multitask_test_data
 
-from evaluation import model_eval_sst, test_model_multitask
+from evaluation import model_eval_sst, test_model_multitask, model_eval_multitask
 
 
 TQDM_DISABLE=True
@@ -82,6 +82,7 @@ def predict_paraphrase(self,
         Note that your output should be unnormalized (a logit); it will be passed to the sigmoid function
         during evaluation, and handled as a logit by the appropriate loss function.
         '''
+
         ### TODO
         raise NotImplementedError
 
@@ -121,14 +122,33 @@ def train_multitask(args):
     # Create the data and its corresponding datasets and dataloader
     sst_train_data, num_labels,para_train_data, sts_train_data = load_multitask_data(args.sst_train,args.para_train,args.sts_train, split ='train')
     sst_dev_data, num_labels,para_dev_data, sts_dev_data = load_multitask_data(args.sst_dev,args.para_dev,args.sts_dev, split ='train')
-
+
+    #Sentiment analysis
     sst_train_data = SentenceClassificationDataset(sst_train_data, args)
     sst_dev_data = SentenceClassificationDataset(sst_dev_data, args)
 
     sst_train_dataloader = DataLoader(sst_train_data, shuffle=True, batch_size=args.batch_size,
                                       collate_fn=sst_train_data.collate_fn)
     sst_dev_dataloader = DataLoader(sst_dev_data, shuffle=False, batch_size=args.batch_size,
                                     collate_fn=sst_dev_data.collate_fn)
+
+    #Paraphrasing
+    paraphrase_train_data = SentencePairDataset(para_train_data, args, isRegression =False)
+    paraphrase_dev_data = SentencePairDataset(para_dev_data, args, isRegression =False)
+
+    paraphrase_train_dataloader = DataLoader(paraphrase_train_data, shuffle=True, batch_size=args.batch_size,
+                                    collate_fn=paraphrase_train_data.collate_fn)
+    paraphrase_dev_dataloader = DataLoader(paraphrase_dev_data, shuffle=True, batch_size=args.batch_size,
+                                    collate_fn=paraphrase_dev_data.collate_fn)
+
+    #sts
+    sts_train_data = SentencePairDataset(sts_train_data, args, isRegression =True)
+    sts_dev_data = SentencePairDataset(sst_dev_data, args, isRegression =True)
+
+    sts_train_dataloader = DataLoader(sst_train_data, shuffle=True, batch_size=256,
+                                    collate_fn=sst_train_data.collate_fn)
+    sts_dev_dataloader = DataLoader(sst_dev_data, shuffle=True, batch_size=256,
+                                collate_fn=sst_dev_data.collate_fn)
 
     # Init model
     config = {'hidden_dropout_prob': args.hidden_dropout_prob,
@@ -175,6 +195,22 @@ def train_multitask(args):
         train_acc, train_f1, *_ = model_eval_sst(sst_train_dataloader, model, device)
         dev_acc, dev_f1, *_ = model_eval_sst(sst_dev_dataloader, model, device)
 
+
+        ## Adding multitask evaluation
+        #train
+        (train_paraphrase_accuracy, train_para_y_pred, train_para_sent_ids,
+         train_sentiment_accuracy,train_sst_y_pred, train_sst_sent_ids,
+           train_sts_corr, train_sts_y_pred, train_sts_sent_ids) = model_eval_multitask(sst_train_dataloader,
+                                                                      paraphrase_train_dataloader,sts_train_dataloader,model, model.device  )
+
+        #dev
+        (dev_paraphrase_accuracy, dev_para_y_pred, dev_para_sent_ids,
+         dev_sentiment_accuracy,dev_sst_y_pred, dev_sst_sent_ids,
+           dev_sts_corr, dev_sts_y_pred, dev__sent_ids) = model_eval_multitask(sst_dev_dataloader,
+                                                                      paraphrase_dev_dataloader,sts_dev_dataloader,model, model.device  )
+
+        #We have to weight or average the three sores to save the best model.
+        # In the diven code only sst is used
         if dev_acc > best_dev_acc:
             best_dev_acc = dev_acc
             save_model(model, optimizer, args, config, args.filepath)