Mantéïa

Exemples/exemple_Summarize1.py

@@ -0,0 +1,6 @@
+from Manteia.Summarize import Summarize
+
+documents = ['What should you do before criticizing Pac-Man? WAKA WAKA WAKA mile in his shoe.','What did Arnold Schwarzenegger say at the abortion clinic? Hasta last vista, baby.']
+
+#model = Model(model_name ='bart')
+su=Summarize(documents=documents)

Exemples/test.py

@@ -0,0 +1,9 @@
+from transformers import BartTokenizer, BartForConditionalGeneration, BartConfig
+# see ``examples/summarization/bart/evaluate_cnn.py`` for a longer example
+model = BartForConditionalGeneration.from_pretrained('bart-large-cnn')
+tokenizer = BartTokenizer.from_pretrained('bart-large-cnn')
+ARTICLE_TO_SUMMARIZE = "My friends are cool but they eat too many carbs."
+inputs = tokenizer.batch_encode_plus([ARTICLE_TO_SUMMARIZE], max_length=1024, return_tensors='pt')
+# Generate Summary
+summary_ids = model.generate(inputs['input_ids'], num_beams=4, max_length=5, early_stopping=True)
+print([tokenizer.decode(g, skip_special_tokens=True, clean_up_tokenization_spaces=False) for g in summary_ids])

Manteia/ActiveLearning.py

@@ -51,6 +51,26 @@ def query(self,predictions,unlabeled_idx,nb_question):
 			print(entropie_trie[:nb_question])
 		return idx_entropie
 
+class DAL():
+	"""
+	The basic discriminative strategy.
+	"""
+
+	def __init__(self,verbose=False):
+		self.verbose=verbose
+		if self.verbose:
+			print('DAL')
+
+	def query(self,predictions,unlabeled_idx,nb_question):
+		#dal est une liste de tuple idx non labélisé et probabilité de ne pas etre labellisé
+		dal=[(idx,p[1])for idx,p in zip(unlabeled_idx,predictions)]
+		dal=sorted(dal, key=itemgetter(1),reverse=True)
+		print(dal[:3])
+		idx_dal=[tup[0] for tup in dal[:nb_question]]
+		if self.verbose:
+			print(dal[:nb_question])
+		return idx_dal
+
 
 
 

Manteia/Dataset.py

@@ -79,7 +79,7 @@ def load(self):
 		if self.name=="trec":
 			self.load_trec()
 
-		if self.name=="agnews":
+		if self.name=="Agnews":
 			self.load_agnews()
 
 		if self.name=="DBPedia":

Manteia/Model.py

@@ -93,7 +93,7 @@ class Model:
 			
 		Attributes:
 	"""
-	def __init__(self,model_name ='bert',model_type=None,task='classification',num_labels=0,epochs=None,MAX_SEQ_LEN = 128,early_stopping=False,path='./model',verbose=True): 
+	def __init__(self,model_name ='bert',model_type=None,task='classification',num_labels=0,epochs=None,MAX_SEQ_LEN = 128,early_stopping=False,path='./model/',verbose=True): 
 
 		self.model_name      = model_name
 		self.model_type      = model_type
@@ -297,7 +297,7 @@ def configuration(self,train_dataloader,batch_size = 16,epochs = 20,n_gpu=1):
 		self.devices()
 
 
-	def fit(self,train_dataloader,validation_dataloader):
+	def fit(self,train_dataloader=None,validation_dataloader=None):
 
 		self.model.to(self.device)
 		#self.model.cuda()
@@ -315,7 +315,8 @@ def fit(self,train_dataloader,validation_dataloader):
 
 				print("")
 				print('======== Epoch {:} / {:} ========'.format(epoch_i + 1, self.epochs))
-				print('Training :')
+				#print('')
+				#print('Training :')
 
 			t0 = time.time()
 			total_loss = 0
@@ -365,21 +366,22 @@ def fit(self,train_dataloader,validation_dataloader):
 				if self.verbose:
 					progress(count=step+1, total=len(train_dataloader))
 
-			if self.verbose==True:
-				print("")
+			if self.verbose:
 				print("  Average training loss: {0:.2f}".format(avg_train_loss))
-				print("  Training epoch took: {:}".format(format_time(time.time() - t0)))
-				print("Validation :")
-
-			t0 = time.time()
-
-			self.model.eval()
+				print("")
 
-			tab_logits = None
-			tab_labels = None
+				#print("  Training epoch took: {:}".format(format_time(time.time() - t0)))
+			if validation_dataloader is not None:
+				if self.verbose:
+					print("")
+					#print("Validation :")
 
+				t0 = time.time()
+				self.model.eval()
+				tab_logits = None
+				tab_labels = None
 
-			for step,batch in enumerate(validation_dataloader):
+				for step,batch in enumerate(validation_dataloader):
 
 					batch = tuple(t.to(self.device) for t in batch)
 
@@ -405,30 +407,34 @@ def fit(self,train_dataloader,validation_dataloader):
 						progress(count=step+1, total=len(validation_dataloader))
 
 
-			acc_validation=accuracy(tab_logits, tab_labels)
-			if self.verbose==True:
-				print("")
-				print("  Accuracy: {0:.2f}".format(acc_validation))
-				print("  Validation took: {:}".format(format_time(time.time() - t0)))
-
-			if self.early_stopping:
-				self.es(acc_validation, self.model,self.device)
+				acc_validation=accuracy(tab_logits, tab_labels)
+				if self.verbose==True:
+					print("")
+					print("  Validation : Accuracy : {0:.2f}".format(acc_validation))
+					#print("  Validation took: {:}".format(format_time(time.time() - t0)))
+
+				if self.early_stopping:
+					self.es(acc_validation, self.model,self.device)
 
-				if self.es.early_stop:
-					if self.verbose==True:
-						print("Early stopping")
-					break
+					if self.es.early_stop:
+						if self.verbose==True:
+							print("")
+							print("Early stopping")
+						break
 		#a la fin de l'entrainement on charge le meilleur model.
 		if self.early_stopping:
 			self.model.load_state_dict(torch.load(os.path.join(self.path,'state_dict_validation.pt')))
+			self.model.to(self.device)
+
 
 		if self.verbose==True:
 			print("")
 			print("Training complete!")
+
 	"""
 	p_type='class' or 'probability' or 'logits'
 	"""
-	def predict(self,predict_dataloader,p_type='class'):
+	def predict(self,predict_dataloader,p_type='class',mode='eval'):
 		'''
 		if self.early_stopping:
 			#by torch
@@ -446,10 +452,11 @@ def predict(self,predict_dataloader,p_type='class'):
 			if self.verbose==True:
 				print('loading model early...')
 		'''
-		#self.model.cuda()
 		self.model.to(self.device)
-
-		self.model.eval()
+		if mode=='eval':
+			self.model.eval()
+		if mode=='train':
+			self.model.train()
 		predictions = None
 		if self.verbose:
 			print('Predicting :')
@@ -793,11 +800,8 @@ def save_checkpoint(self, acc_validation, model,device_model):
 		#save by torch
 		device = torch.device('cpu')
 		model.to(device)
-		if self.verbose:
-			print(type(model))
 		#torch.save(model.module.state_dict(),os.path.join(self.path,'state_dict_validation.pt'))
 		torch.save(model.state_dict(),os.path.join(self.path,'state_dict_validation.pt'))
-
 		model.to(device_model)
 		#save by transformer
 		#model.save_pretrained(self.path)

Manteia/Utils.py

@@ -1,5 +1,16 @@
 import sys,time
+import numpy as np
 
+def coss_validation_idx(nb_pass,nb_docs):
+  docs_idx = [idx for idx in range(nb_docs)]
+  train_idx, test_idx = [], []
+  for pli in range(nb_pass):
+    test_pli_idx = list(np.random.choice(docs_idx,int(len(docs_idx)/nb_pass) , replace=False))
+    train_pli_idx  = [idx for idx in docs_idx if idx not in test_pli_idx]
+    train_idx.append(train_pli_idx)
+    test_idx.append(test_pli_idx)
+  return train_idx, test_idx
+
 def progress(count, total):
 	bar_len = 60
 	filled_len = int(round(bar_len * count / float(total)))

Manteia/__init__.py

@@ -28,7 +28,7 @@
 	reminiscent.
 """
 
-__version__ = "0.0.31"
+__version__ = "0.0.32"
 
 
 from Manteia import Classification