Aktualisieren scripts/ActorOnly+CEL/main_ActorOnly.py, scripts/ActorOnly+CEL/main_CrossEntropy.py

scripts/ActorOnly+CEL/main_ActorOnly.py

 import torch
 from torch import nn
 import numpy as np
-import stanza
-from gensim.models import KeyedVectors
-from datasets import load_dataset
 from structures import *
+from models import *
+import utils
 
-
-# loads skipgram gensim
-file_name = "data/1-billion-word-language-modeling-benchmark-r13output.word2vec.vec"
-model_gensim = KeyedVectors.load_word2vec_format(file_name)
-
-# initialize tokenizer, => sentences splitting and tokenizing, !pip install stanza 
-nlp = stanza.Pipeline(lang='en', processors='tokenize')
-
-# loads dataset cnn_dailymail, !pip install datasets
-dataset = load_dataset('cnn_dailymail', '3.0.0', split='train[:100]+validaton[:20]+test[:20]') # extract subset for testing
-dataset = PreprocessedDataSet(dataset, model_gensim, nlp)  
+# loads dataset cnn_dailymail
+train_dataset = PreprocessedDataSet('/workspace/students/kreuzer/train')
+validation_dataset = PreprocessedDataSet('/workspace/students/kreuzer/val')
+test_dataset = PreprocessedDataSet('/workspace/students/kreuzer/test')
 
 # hyperparameters
 epochs=20
 batch_size=20
 learning_rate=0.001
 
-training_dataloader = torch.utils.data.DataLoader(dataset.train, batch_size=batch_size, shuffle=True)
-test_dataloader = torch.utils.data.DataLoader(dataset.test, batch_size=batch_size, shuffle=True)
+training_dataloader = torch.utils.data.DataLoader(train_dataset, batch_size=batch_size, shuffle=True, collate_fn=lambda x:x)
 
 m = ActorOnlySummarisation()
 
-
 since = time.time()
 val_rouge_history = []
 
@@ -37,16 +27,17 @@ best_model_wts = copy.deepcopy(m.state_dict())
 
 for epoch in range(epochs):
 
-    print('Epoch {}/{}'.format(epoch, epochs - 1))
+    print()
+    print('Epoch {}/{}'.format(epoch+1, epochs))
     print('-' * 10)
 
     # train phase
-    epoch_loss, epoch_rouge = m.training_epoch(D.train)
+    epoch_loss, epoch_rouge = m.training_epoch(training_dataloader)
     print('Train Loss: {:.4f} Rouge Score: {:.4f}'.format(epoch_loss, epoch_rouge))
     
     # validation phase 
-    epoch_rouge = m.validation(dataset.validation)
-    val_rouge_history.append(epoch_rouge)    
+    epoch_rouge = m.validation(validation_dataset)
+    val_rouge_history.append(epoch_rouge)
     print('Validation Rouge Score: {:.4f}'.format(epoch_rouge))
 
     # epoch completed, deep copy the best model sofar
@@ -56,9 +47,15 @@ for epoch in range(epochs):
 
 # after training completed
 time_elapsed = time.time() - since
-print('Training complete in {:.0f}m {:.0f}s'.format(time_elapsed // 60, time_elapsed % 60))
+print('Training + validation complete in {:.0f}m {:.0f}s'.format(time_elapsed // 60, time_elapsed % 60))
 print('Best val rouge: {:4f}'.format(best_rouge))
+
 # write val_rouge_history in file
+with open("ActorOnly_val_history.txt", "a") as f:
+    f.write(">>>Start\n")
+    for event in val_rouge_history:
+        f.write(str(event)+"\n")
+    f.write("\n")
 
 # load best model weights
 m.load_state_dict(best_model_wts)
@@ -66,14 +63,16 @@ m.load_state_dict(best_model_wts)
 # for Critic: save model_actor_only   
 torch.save(m.state_dict(), 'model_actor_only_wts.pth')
 # set sent_vecs 
-dataset.compute_sent_vecs(model)
+train_dataset.compute_sent_vecs(m)
+validation_dataset.compute_sent_vecs(m)
+test_dataset.compute_sent_vecs(m)
 
 # testing
 
 since = time.time()
 
-epoch_rouge_1, epoch_rouge_2, epoch_rouge_l = m.test(dataset.test)
-print('Test rouge_1: {:.4f} rouge_2: {:.4f} rouge_l: {:.4f}'.format(epoch_rouge_1, epoch_rouge_2, epoch_rouge_l))
+epoch_rouge_1, epoch_rouge_2, epoch_rouge_l = m.test(test_dataset)
+print('Test rouge_1: {:.4f} rouge_2: {:.4f} rouge_l: {:.4f} mean: {:.4f}'.format(epoch_rouge_1, epoch_rouge_2, epoch_rouge_l, (epoch_rouge_1+epoch_rouge_2+epoch_rouge_l)/3.0))
 
 # after testing completed
 time_elapsed = time.time() - since

scripts/ActorOnly+CEL/main_CrossEntropy.py

@@ -3,15 +3,19 @@ from torch import nn
 import numpy as np
 from structures import *
 from models import *
+import utils
 
-dataset = torch.load("dataset.data")
+# loads dataset cnn_dailymail
+train_dataset = PreprocessedDataSet('/workspace/students/kreuzer/train')
+validation_dataset = PreprocessedDataSet('/workspace/students/kreuzer/val')
+test_dataset = PreprocessedDataSet('/workspace/students/kreuzer/test')
 
 # hyperparameters
-epochs=2
+epochs=20
 batch_size=20
 learning_rate=0.001
 
-train_dataloader = torch.utils.data.DataLoader(dataset.train, batch_size=batch_size, shuffle=True, collate_fn=lambda x:x)
+training_dataloader = torch.utils.data.DataLoader(train_dataset, batch_size=batch_size, shuffle=True, collate_fn=lambda x:x)
 
 m = SummarisationModelWithCrossEntropyLoss()
 
@@ -28,11 +32,11 @@ for epoch in range(epochs):
     print('-' * 10)
 
     # train phase
-    epoch_loss = m.training_epoch(train_dataloader)
+    epoch_loss = m.training_epoch(training_dataloader)
     print('Train Loss: {:.4f}'.format(epoch_loss))
     
     # validation phase 
-    epoch_rouge = m.validation(dataset.validation)
+    epoch_rouge = m.validation(validation_dataset)
     val_rouge_history.append(epoch_rouge)    
     print('Validation Rouge Score: {:.4f}'.format(epoch_rouge))
 
@@ -43,9 +47,15 @@ for epoch in range(epochs):
 
 # after training completed
 time_elapsed = time.time() - since
-print('Training complete in {:.0f}m {:.0f}s'.format(time_elapsed // 60, time_elapsed % 60))
+print('Training + validation complete in {:.0f}m {:.0f}s'.format(time_elapsed // 60, time_elapsed % 60))
 print('Best val rouge: {:4f}'.format(best_rouge))
+
 # write val_rouge_history in file
+with open("CEL_val_history.txt", "a") as f:
+    f.write(">>>Start\n")
+    for event in val_rouge_history:
+        f.write(str(event)+"\n")
+    f.write("\n")
 
 # load best model weights
 m.load_state_dict(best_model_wts)
@@ -53,7 +63,7 @@ m.load_state_dict(best_model_wts)
 # testing
 since = time.time()
 
-epoch_rouge_1, epoch_rouge_2, epoch_rouge_l = m.test(dataset.test)
+epoch_rouge_1, epoch_rouge_2, epoch_rouge_l = m.test(test_dataset)
 print('Test rouge_1: {:.4f} rouge_2: {:.4f} rouge_l: {:.4f} mean: {:.4f}'.format(epoch_rouge_1, epoch_rouge_2, epoch_rouge_l, (epoch_rouge_1+epoch_rouge_2+epoch_rouge_l)/3.0))
 
 # after testing completed