new setting and test of our model

lambdaAPI/settings.py

@@ -110,12 +110,13 @@ AUTH_PASSWORD_VALIDATORS = [
     },
 ]
 
-ALLOWED_HOSTS = ['127.0.0.1','https://ferdinandpycode-lambda-hf.hf.space','http://localhost']
+ALLOWED_HOSTS = ['127.0.0.1','https://ferdinandpycode-lambda-hf.hf.space','http://localhost','192.168.100.28']
 
 CORS_ALLOWED_ORIGINS = [
     "http://localhost",
     "https://ferdinandpycode-lambda-hf.hf.space",
     "http://127.0.0.1",
+    "http://192.168.100.28"
 ]
 
 

utils/utils_function.py

@@ -1,8 +1,8 @@
 from fairseq.models.transformer import TransformerModel
-import os
 import torch
-import zipfile
-import shutil
+import re
+import string
+
 
 class Translator:
     def __init__(self, isFon:bool, device='cuda' if torch.cuda.is_available() else 'cpu'):
@@ -23,10 +23,33 @@ class Translator:
         
         # Mettre le modèle en mode évaluation (pas de mise à jour des poids)
         self.model.eval()
+    
+    def preprocess(self, data):
+        print('Preprocessing...')
+        # Convertir chaque lettre en minuscule
+        text = data.lower().strip()
+        
+        # Supprimer les apostrophes des phrases
+        text = re.sub("'", "", text)
+        
+        # Supprimer toute ponctuation
+        exclude = set(string.punctuation)
+        text = ''.join(ch for ch in text if ch not in exclude)
+        
+        # Supprimer les chiffres
+        digit = str.maketrans('', '', string.digits)
+        text = text.translate(digit)
+        
+        return text
 
     def translate(self, text):
+
+        print(text)
+        pre_traited = self.preprocess(text)
+        print(pre_traited)
+
         # Encodage du texte en tokens
-        tokens = self.model.encode(text)
+        tokens = self.model.encode(pre_traited)
         
         # Utilisation de la méthode generate avec le paramètre beam
         translations = self.model.generate(tokens, beam=5)
@@ -37,4 +60,4 @@ class Translator:
         # Décodage des tokens en traduction
         translations = [self.model.decode(best_translation_tokens[i]) for i in range(5)]
         
-        return "\n".join(translations)
+        return "\n".join(translations)