Update app.py

app.py

@@ -393,7 +393,7 @@ def generate_answer(prompt, max_length=860, temperature=0.2):
     model_f = models['llm_model']
     inputs = tokenizer_f(prompt, return_tensors="pt", truncation=True)
     # Start timing
-    start_time = time.time()
+    #start_time = time.time()
     # Generate the output
     output_ids = model_f.generate(
         inputs.input_ids,
@@ -403,9 +403,9 @@ def generate_answer(prompt, max_length=860, temperature=0.2):
         pad_token_id=tokenizer_f.eos_token_id
     )
     # End timing
-    end_time = time.time()
+    #end_time = time.time()
     # Calculate the duration
-    duration = end_time - start_time
+    #duration = end_time - start_time
     # Decode the answer
     answer = tokenizer_f.decode(output_ids[0], skip_special_tokens=True)
     # Extract keywords from the passage and answer
@@ -413,9 +413,9 @@ def generate_answer(prompt, max_length=860, temperature=0.2):
     answer_keywords = set(answer.lower().split())
     # Verify if the answer aligns with the passage
     if passage_keywords.intersection(answer_keywords):
-        return answer, duration
+        return answer  #, duration
     else:
-        return "Sorry, I can't help with that.", duration
+        return "Sorry, I can't help with that." #, duration
   
 def remove_answer_prefix(text):
     prefix = "Answer:"
@@ -477,8 +477,8 @@ entities = extract_entities(query_text)
 passage = enhance_passage_with_entities(combined_parts, entities)
 # Generate answer with the enhanced passage
 prompt = create_prompt(query_text, passage)
-answer, generation_time = generate_answer(prompt)
-print(f"\nTime taken to generate the answer: {generation_time:.2f} seconds")
+answer = generate_answer(prompt)
+#print(f"\nTime taken to generate the answer: {generation_time:.2f} seconds")
 answer_part = answer.split("Answer:")[-1].strip()
 cleaned_answer = remove_answer_prefix(answer_part)
 final_answer = remove_incomplete_sentence(cleaned_answer)