Mantéïa

Exemples/exemple_Classification1.py

@@ -5,5 +5,5 @@
 
 labels = ['funny','not funny']
 
-model = Model(model_name ='bert')
+model = Model(model_name ='bart')
 cl=Classification(model,documents,labels,process_classif=True)

Manteia/Model.py

@@ -104,6 +104,7 @@ def __init__(self,model_name ='bert',model_type=None,task='classification',num_l
 		self.epochs           = epochs
 		self.path            = path
 		self.verbose         = verbose
+		self.device          = None
 		self.history         = {}
 		self.history['loss'] = []
 		self.history['step'] = []
@@ -303,8 +304,8 @@ def fit(self,train_dataloader,validation_dataloader):
 		if self.early_stopping:
 			self.es=EarlyStopping(path=self.path)
 
-		if self.n_gpu > 1:
-			self.model = torch.nn.DataParallel(self.model)
+		#if self.n_gpu > 1:
+		#	self.model = torch.nn.DataParallel(self.model)
 
 		loss_values = []
 
@@ -334,10 +335,10 @@ def fit(self,train_dataloader,validation_dataloader):
 
 				self.model.zero_grad()        
 
-				if self.model_name != 'distilbert':
-						outputs = self.model(b_input_ids, token_type_ids=None, attention_mask=b_input_mask, labels=b_labels)
+				if self.model_name != 'distilbert' and self.model_name != 'bart':
+					outputs = self.model(b_input_ids, token_type_ids=None, attention_mask=b_input_mask, labels=b_labels)
 				else:
-						outputs = self.model(b_input_ids, attention_mask=b_input_mask, labels=b_labels)
+					outputs = self.model(b_input_ids, attention_mask=b_input_mask, labels=b_labels)
 
 
 				#loss = outputs[0]
@@ -384,7 +385,7 @@ def fit(self,train_dataloader,validation_dataloader):
 
 					with torch.no_grad():        
 
-						if self.model_name != 'distilbert':
+						if self.model_name != 'distilbert' and self.model_name != 'bart':
 							outputs = self.model(b_input_ids, token_type_ids=None,attention_mask=b_input_mask)
 						else:
 							outputs = self.model(b_input_ids,attention_mask=b_input_mask)
@@ -414,24 +415,34 @@ def fit(self,train_dataloader,validation_dataloader):
 					if self.verbose==True:
 						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')))
+
 		if self.verbose==True:
 			print("")
 			print("Training complete!")
 	"""
 	p_type='class' or 'probability' or 'logits'
 	"""
 	def predict(self,predict_dataloader,p_type='class'):
+		'''
 		if self.early_stopping:
 			#by torch
 			#pour charger uniquement la classe du modèle!
+			print('test')
+			self.load_type()
+			print('test')
 			self.load_class()
-			self.model.load_state_dict(torch.load(self.path+'state_dict_validation.pt'))
+			print('test')
+			self.model.load_state_dict(torch.load(os.path.join(self.path,'state_dict_validation.pt')))
+			print('test')
 
 			#by transformer
 			#self.model.from_pretrained(self.path)
 			if self.verbose==True:
 				print('loading model early...')
-				
+		'''
 		#self.model.cuda()
 		self.model.to(self.device)
 
@@ -566,12 +577,16 @@ def save(self,file_name):
 			else:
 				print ("Successfully created the directory %s " % self.path)
 		self.model.to(torch.device('cpu'))
-		torch.save(self.model.module.state_dict(),os.path.join(self.path,file_name))
+		#torch.save(self.model.module.state_dict(),os.path.join(self.path,file_name))
+		torch.save(self.model.state_dict(),os.path.join(self.path,file_name))
 		self.model.to(self.device)
 
 	def load(self,file_name):
+			self.load_type()
 			self.load_class()
 			self.model.load_state_dict(torch.load(os.path.join(self.path,file_name)))
+			if self.device is None:
+				self.devices()
 			self.model.to(self.device)
 
 def choose_from_top(probs, n=5):
@@ -734,8 +749,9 @@ def __init__(self, patience=2, delta=0,path=None, verbose=True):
 		self.acc_validation_min = 0
 		self.delta = delta
 		self.path=path
-		if os.path.isfile(self.path+'state_dict_validation.pt'):
-			os. remove(self.path+'state_dict_validation.pt') 
+
+		if os.path.isfile(os.path.join(self.path,'state_dict_validation.pt')):
+			os. remove(os.path.join(self.path,'state_dict_validation.pt')) 
 
 	def __call__(self, acc_validation , model,device_model):
 
@@ -773,7 +789,8 @@ def save_checkpoint(self, acc_validation, model,device_model):
 		device = torch.device('cpu')
 		model.to(device)
 		print(type(model))
-		torch.save(model.module.state_dict(),self.path+'state_dict_validation.pt')
+		#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

Manteia/Summarize.py

@@ -43,19 +43,19 @@ class Summarize:
 			
 			
 	"""
-	def __init__(self,model=None,documents = [],verbose=True):
+	def __init__(self,model=None,documents = [],verbose=True,process_summarize=False):
 
-		self.process_classif = process_classif
-		self.verbose         = verbose
-		self.model           = model
-		self.documents_train = documents_train
-		self.labels_train    = labels_train
-		self.documents_test  = documents_test
-		self.labels_test     = labels_test
+		self.process_summarize = process_summarize
+		self.verbose           = verbose
+		self.model             = model
+		self.documents         = documents
 
 		self.load_model()
 		inputs=self.process_text()
-		print(self.predict(inputs))
+		summary_ids = self.model.model.generate(inputs['input_ids'], num_beams=6, max_length=8, early_stopping=True)
+		print([self.model.tokenizer.decode(g, skip_special_tokens=True, clean_up_tokenization_spaces=False) for g in summary_ids])
+
+		#print(self.predict(inputs))
 
 	def load_model(self):
 		"""
@@ -66,6 +66,7 @@ def load_model(self):
 		"""
 		if self.model is None:
 			self.model = Model(model_name ='bart',model_type='bart-large-cnn',task='summarize')
+
 		self.model.load_type()
 		self.model.load_tokenizer()
 		self.model.load_class()
@@ -80,7 +81,7 @@ def process_text(self):
 		
 			from Manteia.Summarize import Summarize
 		"""
-		inputs = self.model.tokenizer.batch_encode_plus(self.documents, max_length=1024, return_tensors='pt')
+		inputs = self.model.tokenizer.batch_encode_plus(self.documents, max_length=1024,pad_to_max_length=True, return_tensors='pt')
 		return inputs
 
 	def predict(self,inputs):

README.md

@@ -14,8 +14,11 @@ You can install it with pip :
 For use with GPU and cuda we recommend the use of [Anaconda](https://www.anaconda.com/open-source) :
 
      __conda create -n manteia_env python=3.7__
+
      __conda activate manteia_env__
+
      __conda install pytorch__
+
      __pip install manteia__
 
 Example of use Classification :

docs/_build/html/Model.html

@@ -182,7 +182,7 @@
 
 <dl class="py class">
 <dt id="Manteia.Model.Model">
-<em class="property">class </em><code class="sig-prename descclassname">Manteia.Model.</code><code class="sig-name descname">Model</code><span class="sig-paren">(</span><em class="sig-param"><span class="n">model_name</span><span class="o">=</span><span class="default_value">'bert'</span></em>, <em class="sig-param"><span class="n">model_type</span><span class="o">=</span><span class="default_value">None</span></em>, <em class="sig-param"><span class="n">num_labels</span><span class="o">=</span><span class="default_value">0</span></em>, <em class="sig-param"><span class="n">epochs</span><span class="o">=</span><span class="default_value">None</span></em>, <em class="sig-param"><span class="n">MAX_SEQ_LEN</span><span class="o">=</span><span class="default_value">128</span></em>, <em class="sig-param"><span class="n">early_stopping</span><span class="o">=</span><span class="default_value">False</span></em>, <em class="sig-param"><span class="n">path</span><span class="o">=</span><span class="default_value">'./model'</span></em>, <em class="sig-param"><span class="n">verbose</span><span class="o">=</span><span class="default_value">True</span></em><span class="sig-paren">)</span><a class="headerlink" href="#Manteia.Model.Model" title="Permalink to this definition">¶</a></dt>
+<em class="property">class </em><code class="sig-prename descclassname">Manteia.Model.</code><code class="sig-name descname">Model</code><span class="sig-paren">(</span><em class="sig-param"><span class="n">model_name</span><span class="o">=</span><span class="default_value">'bert'</span></em>, <em class="sig-param"><span class="n">model_type</span><span class="o">=</span><span class="default_value">None</span></em>, <em class="sig-param"><span class="n">task</span><span class="o">=</span><span class="default_value">'classification'</span></em>, <em class="sig-param"><span class="n">num_labels</span><span class="o">=</span><span class="default_value">0</span></em>, <em class="sig-param"><span class="n">epochs</span><span class="o">=</span><span class="default_value">None</span></em>, <em class="sig-param"><span class="n">MAX_SEQ_LEN</span><span class="o">=</span><span class="default_value">128</span></em>, <em class="sig-param"><span class="n">early_stopping</span><span class="o">=</span><span class="default_value">False</span></em>, <em class="sig-param"><span class="n">path</span><span class="o">=</span><span class="default_value">'./model'</span></em>, <em class="sig-param"><span class="n">verbose</span><span class="o">=</span><span class="default_value">True</span></em><span class="sig-paren">)</span><a class="headerlink" href="#Manteia.Model.Model" title="Permalink to this definition">¶</a></dt>
 <dd><p>This is the class to construct model.</p>
 <p>Args:</p>
 <blockquote>
@@ -219,6 +219,29 @@
 </pre></div>
 </div>
 <p>Attributes:</p>
+<dl class="py method">
+<dt id="Manteia.Model.Model.predict">
+<code class="sig-name descname">predict</code><span class="sig-paren">(</span><em class="sig-param"><span class="n">predict_dataloader</span></em>, <em class="sig-param"><span class="n">p_type</span><span class="o">=</span><span class="default_value">'class'</span></em><span class="sig-paren">)</span><a class="headerlink" href="#Manteia.Model.Model.predict" title="Permalink to this definition">¶</a></dt>
+<dd><dl>
+<dt>if self.early_stopping:</dt><dd><p>#by torch
+#pour charger uniquement la classe du modèle!
+print(‘test’)
+self.load_type()
+print(‘test’)
+self.load_class()
+print(‘test’)
+self.model.load_state_dict(torch.load(os.path.join(self.path,’state_dict_validation.pt’)))
+print(‘test’)</p>
+<p>#by transformer
+#self.model.from_pretrained(self.path)
+if self.verbose==True:</p>
+<blockquote>
+<div><p>print(‘loading model early…’)</p>
+</div></blockquote>
+</dd>
+</dl>
+</dd></dl>
+
 </dd></dl>
 
 <dl class="py function">

docs/_build/html/_sources/index.rst.txt

@@ -15,6 +15,27 @@ of your results on several dataset ( 20newsgroups, Agnews Amazon Review Full, Am
 | .. image:: images/train.png     | .. image:: images/boxplot.png| 
 +---------------------------------+------------------------------+
 
+Installation
+------------
+
+You can install it with pip :
+
+     pip install Manteia
+
+.. _Anaconda: https://www.anaconda.com/open-source>
+
+For use with GPU and cuda we recommend the use of `Anaconda`_. :
+
+     conda create -n manteia_env python=3.7
+
+     conda activate manteia_env
+
+     conda install pytorch
+
+     pip install manteia
+
+
+
 Classes
 =======
 

docs/_build/html/genindex.html

@@ -470,7 +470,11 @@ <h2 id="P">P</h2>
       <li><a href="Visualisation.html#Manteia.Visualisation.Visualisation.plot_boxplot">plot_boxplot() (Manteia.Visualisation.Visualisation method)</a>
 </li>
       <li><a href="Classification.html#Manteia.Classification.Classification.predict">predict() (Manteia.Classification.Classification method)</a>
+
+      <ul>
+        <li><a href="Model.html#Manteia.Model.Model.predict">(Manteia.Model.Model method)</a>
 </li>
+      </ul></li>
       <li>
     Preprocess
 

docs/_build/html/index.html

@@ -188,6 +188,16 @@ <h1>Simple Documentation Tutorial : Mantéïa<a class="headerlink" href="#simple
 </tr>
 </tbody>
 </table>
+<div class="section" id="installation">
+<h2>Installation<a class="headerlink" href="#installation" title="Permalink to this headline">¶</a></h2>
+<p>You can install it with pip :</p>
+<p>     pip install Manteia</p>
+<p>For use with GPU and cuda we recommend the use of <a class="reference external" href="https://www.anaconda.com/open-source&gt;">Anaconda</a>. :</p>
+<p>     conda create -n manteia_env python=3.7</p>
+<p>     conda activate manteia_env</p>
+<p>     conda install pytorch</p>
+<p>     pip install manteia</p>
+</div>
 </div>
 <div class="section" id="classes">
 <h1>Classes<a class="headerlink" href="#classes" title="Permalink to this headline">¶</a></h1>