Update app.py

app.py

@@ -1,10 +1,10 @@
-import numpy as np
 import streamlit as st
 from transformers import DPRQuestionEncoder, DPRQuestionEncoderTokenizer
 from transformers import BartForConditionalGeneration, BartTokenizer
 from sentence_transformers import SentenceTransformer
 import pdfplumber
 from sklearn.metrics.pairwise import cosine_similarity
+import numpy as np  # Import NumPy
 import torch
 
 # Load the Question Encoder, Context Encoder, and Tokenizers
@@ -60,9 +60,8 @@ query = st.text_input("🔍 Enter your query")
 
 if st.button("💬 Get Answer"):
     if query:
-        # Step 1: Encode the query
-        question_inputs = question_tokenizer(query, return_tensors="pt")
-        question_embedding = question_encoder(**question_inputs).pooler_output.detach().cpu().numpy()
+        # Step 1: Encode the query with the same SentenceTransformer model
+        question_embedding = sentence_model.encode([query])  # Use the same model
 
         # Step 2: Calculate Cosine Similarity
         similarity_scores = cosine_similarity(question_embedding, doc_embeddings)