remove top_k openai

app.py

@@ -178,20 +178,19 @@ with st.sidebar:
             model = st.selectbox("Choose the LLM model", ["gpt-4o-mini", "gpt-3.5-turbo", "llama-tommi-0.35"], key='a1model')
 
             if model == "llama-tommi-0.35":
-                tommi_do_sample = st.toggle("Enable Sampling", value=True, key='tommi_sample')
+                tommi_do_sample = st.toggle("Enable Sampling", value=False, key='tommi_sample')
 
                 if tommi_do_sample:
                     tommi_temperature = st.slider("Temperature", 0.0, 1.5, 0.7, key='tommi_temp')
                     tommi_top_k = st.slider("Top K", 0, 100, 50, key='tommi_top_k')
                     tommi_top_p = st.slider("Top P", 0.0, 1.0, 0.95, key='tommi_top_p')
                 else:
-                    tommi_num_beams = st.slider("Num Beams", 1, 10, 4, key='tommi_num_beams')
+                    tommi_num_beams = st.slider("Num Beams", 1, 4, 1, key='tommi_num_beams')
 
                 tommi_max_new_tokens = st.slider("Max New Tokens", 100, 2000, 500, step=50, key='tommi_max_new_tokens')
             else:
                 expert_temperature = st.slider("Temperature", 0.0, 1.5, 0.7, key='a1t')
                 expert_top_p = st.slider("Top P", 0.0, 1.0, 0.9, key='a1p')
-                expert_top_k = st.slider("Top K", 0, 100, 50, key='a1k')
 
     with st.expander('Synthesis model',expanded=False):
 
@@ -324,8 +323,7 @@ if submit_button_placeholder.button("AI Answer", type="primary"):
                         st.session_state.question,
                         model=model,
                         temperature=expert_temperature,
-                        top_p=expert_top_p,
-                        top_k=expert_top_k
+                        top_p=expert_top_p
                     )
             else:
                 st.session_state.expert_answer = 'No Expert Answer. Only use the context.'

prompts.py

@@ -14,7 +14,7 @@ def fix_latex(text):
 
 
 # Step 1: Domain-Specific Answer Generation
-def openai_domain_specific_answer_generation(subject, question, model="gpt4o-mini", temperature=0.3, top_p=0.1, top_k=0.1):
+def openai_domain_specific_answer_generation(subject, question, model="gpt4o-mini", temperature=0.3, top_p=0.1):
     system_prompt = f"""
     You are a highly specialized assistant for the subject {subject}. Provide a direct and focused answer to the following question based on your specialized training.
     """
@@ -40,7 +40,6 @@ def openai_domain_specific_answer_generation(subject, question, model="gpt4o-min
         ],
         temperature=temperature,  # Set low for deterministic and precise responses.
         top_p=top_p,  # Focus on high-probability outputs to ensure accuracy.
-        top_k=top_k,
         frequency_penalty=0.1,  # Reduce repetition of technical terms.
         presence_penalty=0.0  # Prevent introduction of unrelated ideas.
     )