Update app.py

app.py

@@ -8,7 +8,7 @@ hf_token = os.getenv("TUTOR_LLAMA")
 login(token=hf_token)
 
 # Load LLaMA model and tokenizer for Arabic and ESL tutoring
-model_name = "meta-llama/Llama-3.2-1B"  # Change to the model you're using
+model_name = "meta-llama/Llama-3.2-1B"  # Adjust to the LLaMA model you're using
 tokenizer = AutoTokenizer.from_pretrained(model_name)
 model = AutoModelForCausalLM.from_pretrained(model_name)
 
@@ -34,19 +34,31 @@ do_sample = st.sidebar.checkbox("Enable Random Sampling", value=True)  # Enable
 # Input field for the student
 student_question = st.text_input("Ask your question in English or Arabic!")
 
-# Generate and display response using the LLaMA model
-if student_question:
-    # Adjust prompt to encourage student-friendly responses
-    prompt = f"Please explain the answer step by step in simple terms to a young student: '{student_question}'"
-    
-    # Call the pipeline with adjusted parameters
+# Function to generate response with post-processing
+def generate_response(prompt, max_length=75):
+    # Generate the model's response
     response = model_pipeline(
         prompt,
-        max_length=75,  # Adjust this based on desired response length
-        temperature=temperature,  # Control randomness
-        top_p=top_p,  # Nucleus sampling
-        top_k=top_k,  # Top-k sampling
-        do_sample=do_sample  # Enable or disable sampling
+        max_length=max_length,
+        temperature=temperature,
+        top_p=top_p,
+        top_k=top_k,
+        do_sample=do_sample
     )
     
-    st.write("Tutor's Answer:", response[0]['generated_text'])
+    # Extract the generated text and remove the prompt (if necessary)
+    generated_text = response[0]['generated_text']
+    
+    # Find the first instance of the actual generated answer (post-prompt)
+    cleaned_text = generated_text.replace(prompt, "").strip()
+    return cleaned_text
+
+# Generate and display response using the LLaMA model
+if student_question:
+    # Format the prompt to guide the model to respond conversationally and concisely
+    prompt = f"Q: {student_question}\nA: Explain it simply to a young student in no more than 3 sentences."
+
+    # Call the function to generate and clean the response
+    answer = generate_response(prompt)
+    
+    st.write("Tutor's Answer:", answer)