Update app.py

app.py

@@ -1,35 +1,29 @@
 import gradio as gr
-import pickle
 from sentence_transformers import SentenceTransformer
 import pandas as pd
 
 def find(query):
-    # transform query from user
-    model = SentenceTransformer('Bofandra/fine-tuning-use-cmlm-multilingual-quran-translation-qa')
-    encoded_query_text = model.encode(query)
-
-    # get encoded quran text
-    file = open('encoded_quran_fine-tuning-use-cmlm-multilingual-qa-quran-splitted.sav','rb')
-    encoded_quran_text = pickle.load(file)
-    file.close()
-
-    # compare query to each quran verse
-    i = 0
-    text_similarity = []
-    for encoded_quran_ayat in encoded_quran_text:
-      similarity = encoded_query_text @ encoded_quran_ayat.T
-      text_similarity.append(similarity)
-      i=i+1
-      print(i)
-
-    # insert the similarity value to dataframe & sort it
-    file = open('quran-splitted.sav','rb')
-    quran_splitted = pickle.load(file)
-    quran_splitted['similarity'] = text_similarity
-    sorted_quran = quran_splitted.sort_values(by='similarity', ascending=False)
+    def get_detailed_instruct(task_description: str, query: str) -> str:
+        return f'Instruct: {task_description}\nQuery: {query}'
+    
+    # Each query must come with a one-sentence instruction that describes the task
+    task = 'Given a web search query, retrieve relevant passages that answer the query'
+    queries = [
+        get_detailed_instruct(task, query)
+    ]
     
-    # insert the similarity value to dataframe & sort it
     quran = pd.read_csv('quran-simple-clean.txt', delimiter="|")
+    documents = quran['text'].tolist()
+    input_texts = queries + documents
+    
+    model = SentenceTransformer('intfloat/multilingual-e5-large-instruct')
+    
+    embeddings = model.encode(input_texts, convert_to_tensor=True, normalize_embeddings=True)
+    scores = (embeddings[:1] @ embeddings[1:].T) * 100
+
+    # insert the similarity value to dataframe & sort it
+    quran['similarity'] = scores.tolist()[0]
+    sorted_quran = quran.sort_values(by='similarity', ascending=False)
     
     results = ""
     i = 0