Update app.py

app.py

@@ -11,6 +11,23 @@ import chardet
 from pathlib import Path
 from detect_delimiter import detect
 
+data = pd.read_csv("training_data.csv")
+#Change Information - Sammenligning to information
+data['Category'] = data['Category'].replace('Information - Sammenligning', 'Informational')
+data['Category'] = data['Category'].replace('Information', 'Informational')
+data = data.groupby('Category').apply(lambda x: x.sample(1500, replace=True)).reset_index(drop=True)
+train_df = pd.DataFrame()
+train_df['text'] = data['keywords']
+train_df['labels'] = data['Category']
+train_df['labels'] = train_df['labels'].astype('category').cat.codes
+n_labels = len(train_df['labels'].unique())
+from simpletransformers.ner import NERModel
+from simpletransformers.classification import ClassificationModel
+
+model = ClassificationModel('bert', 'Maltehb/danish-bert-botxo', num_labels=n_labels, use_cuda=True, args={'reprocess_input_data': True, 'overwrite_output_dir': True})
+
+model.train_model(train_df)
+
 label_dict = {
     0: "Brandsøgning",
     1: "Informational",
@@ -19,7 +36,7 @@ label_dict = {
     4: "Transactional"
 }
 upload_file = st.file_uploader("Choose a file",type="csv" )
-model = pickle.load(open("finalized_model.sav","rb"))
+#model = pickle.load(open("finalized_model.sav","rb"))
 
 if upload_file is not None:
     result = chardet.detect(upload_file.getvalue())