Update main.py

main.py

@@ -11,21 +11,15 @@ from sklearn.feature_extraction.text import TfidfVectorizer
 def load_models():
     st.session_state.loaded = True
 
-    # with open('models/tfidf_vectorizer_svm_model_2_classes_gpt_chatgpt_detection_tfidf_bg_0.886_F1_score.pkl', 'rb') as f:
-    #     st.session_state.tfidf_vectorizer_disinformation = pickle.load(f)
-
     with open('models/tfidf_vectorizer_untrue_inform_detection_tfidf_bg_0.96_F1_score_3Y_N_Q1_082023.pkl', 'rb') as f:
         st.session_state.tfidf_vectorizer_untrue_inf = pickle.load(f)
 
-    # with open('models/svm_model_2_classes_gpt_chatgpt_detection_tfidf_bg_0.886_F1_score.pkl', 'rb') as f:
-    #     st.session_state.gpt_detector = pickle.load(f)
-
     with open('models/SVM_model_untrue_inform_detection_tfidf_bg_0.96_F1_score_3Y_N_Q1_082023.pkl', 'rb') as f:
         st.session_state.untrue_detector = pickle.load(f)
 
     st.session_state.bert_disinfo = pipeline(task="text-classification",
-                                     model=BertForSequenceClassification.from_pretrained("TRACES/private-bert", use_auth_token=os.environ['ACCESS_TOKEN'],  num_labels=2),
-                                     tokenizer=AutoTokenizer.from_pretrained("TRACES/private-bert", use_auth_token=os.environ['ACCESS_TOKEN']))
+                                     model=BertForSequenceClassification.from_pretrained("usmiva/bert-desinform-bg", num_labels=2),
+                                     tokenizer=AutoTokenizer.from_pretrained("usmiva/bert-desinform-bg"))
     st.session_state.bert_gpt = pipeline(task="text-classification",
                                      model=BertForSequenceClassification.from_pretrained("usmiva/bert-deepfake-bg", num_labels=2),
                                      tokenizer=AutoTokenizer.from_pretrained("usmiva/bert-deepfake-bg"))
@@ -52,16 +46,13 @@ if all([
     'untrue_detector_result' not in st.session_state,
     'bert_disinfo_result' not in st.session_state
     ]):
-    # st.session_state.gpt_detector_result = ''
-    # st.session_state.gpt_detector_probability = [1, 0]
-
+    st.session_state.bert_gpt_result = [{'label': '', 'score': 1}]
 
     st.session_state.untrue_detector_result = ''
     st.session_state.untrue_detector_probability = 1
     
     st.session_state.bert_disinfo_result = [{'label': '', 'score': 1}]
         
-    st.session_state.bert_gpt_result = [{'label': '', 'score': 1}]
 
 content = load_content()
 if 'loaded' not in st.session_state:
@@ -98,10 +89,7 @@ if st.session_state.agree:
                                   content['text_placeholder'][st.session_state.lang]).strip('\n')
     
         if st.button(content['analyze_button'][st.session_state.lang]):
-            # user_tfidf_disinformation = st.session_state.tfidf_vectorizer_disinformation.transform([user_input])
-            # st.session_state.gpt_detector_result = st.session_state.gpt_detector.predict(user_tfidf_disinformation)[0]
-            # st.session_state.gpt_detector_probability = st.session_state.gpt_detector.predict_proba(user_tfidf_disinformation)[0]
-
+            st.session_state.bert_gpt_result = st.session_state.bert_gpt(user_input)
             
             user_tfidf_untrue_inf = st.session_state.tfidf_vectorizer_untrue_inf.transform([user_input])
             st.session_state.untrue_detector_result = st.session_state.untrue_detector.predict(user_tfidf_untrue_inf)[0]
@@ -110,19 +98,16 @@ if st.session_state.agree:
 
             st.session_state.bert_disinfo_result = st.session_state.bert_disinfo(user_input)
 
-            st.session_state.bert_gpt_result = st.session_state.bert_gpt(user_input)
-
             
-    
-        # if st.session_state.gpt_detector_result == 1:
-        #     st.warning(content['gpt_getect_yes'][st.session_state.lang] +
-        #                str(round(st.session_state.gpt_detector_probability[1] * 100, 2)) +
-        #                content['gpt_yes_proba'][st.session_state.lang], icon="⚠️")
-        # else:
-        #     st.success(content['gpt_getect_no'][st.session_state.lang] +
-        #                str(round(st.session_state.gpt_detector_probability[0] * 100, 2)) +
-        #                content['gpt_no_proba'][st.session_state.lang], icon="✅")
 
+        if st.session_state.bert_gpt_result[0]['label'] == 'LABEL_1':
+            st.warning(content['bert_gpt'][st.session_state.lang] +
+                       str(round(st.session_state.bert_gpt_result[0]['score'] * 100, 2)) +
+                       content['bert_gpt_prob'][st.session_state.lang], icon = "⚠️")
+        else:
+            st.success(content['bert_human'][st.session_state.lang] +
+                       str(round(st.session_state.bert_gpt_result[0]['score'] * 100, 2)) +
+                       content['bert_human_prob'][st.session_state.lang], icon="✅")
         
         if st.session_state.untrue_detector_result == 0:
             st.warning(content['untrue_getect_yes'][st.session_state.lang] +
@@ -142,15 +127,6 @@ if st.session_state.agree:
                        str(round(st.session_state.bert_disinfo_result[0]['score'] * 100, 2)) +
                        content['bert_no_2'][st.session_state.lang], icon="✅")
 
-        if st.session_state.bert_gpt_result[0]['label'] == 'LABEL_1':
-            st.warning(content['bert_gpt'][st.session_state.lang] +
-                       str(round(st.session_state.bert_gpt_result[0]['score'] * 100, 2)) +
-                       content['bert_gpt_prob'][st.session_state.lang], icon = "⚠️")
-        else:
-            st.success(content['bert_human'][st.session_state.lang] +
-                       str(round(st.session_state.bert_gpt_result[0]['score'] * 100, 2)) +
-                       content['bert_human_prob'][st.session_state.lang], icon="✅")
-
     
         st.info(content['disinformation_definition'][st.session_state.lang], icon="ℹ️")