fixing model loading

app.py

@@ -131,7 +131,7 @@ with st.sidebar:
                 else:
                     synthesis_num_beams = st.slider("Num Beams", 1, 4, 1, key='synthesis_num_beams')
 
-                synthesis_max_new_tokens = st.slider("Max New Tokens", 100, 2000, 500, step=50, key='synthesis_max_new_tokens')
+                synthesis_max_new_tokens = st.slider("Max New Tokens", 100, 2000, 1500, step=50, key='synthesis_max_new_tokens')
         else:
             # Temperature
             synthesis_api_temperature = st.slider("Temperature", 0.0, .3, .5, help="Defines the randomness in the next token prediction. Lower: More predictable and focused. Higher: More adventurous and diverse.", key='a2t')
@@ -300,6 +300,15 @@ if submit_button_placeholder.button("AI Answer", type="primary"):
             # synthesis responses
             #-------------------------
             if st.session_state.synthesis_model in ["LLaMA-3.2-3B", "LLaMA-3.2-11B"]:
+
+                if st.session_state.expert_model == "LLaMA-3.2-11B":
+                    model_s = st.session_state.llama_model
+                    tokenizer_s = st.session_state.llama_tokenizer
+                    
+                elif st.session_state.expert_model == "LLaMA-3.2-3B":
+                    model_s = st.session_state.llama_model_3B
+                    tokenizer_s = st.session_state.llama_tokenizer_3B
+
                 synthesis_prompt = f"""
                 Question:
                 {st.session_state.question}
@@ -318,8 +327,8 @@ if submit_button_placeholder.button("AI Answer", type="primary"):
                 ]
 
                 synthesis_answer = generate_response(
-                    model=st.session_state.llama_model,
-                    tokenizer=st.session_state.llama_tokenizer,
+                    model=model_s,
+                    tokenizer=tokenizer_s,
                     messages=messages,
                     tokenizer_max_length=30000,
                     do_sample=synthesis_do_sample,