fix: prompt

app.py

@@ -513,6 +513,7 @@ def display_llm_output():
             st.session_state["model_filename"],
             st.session_state["model_type"],
             st.session_state["gpu_layers"],
+            "You are a medical assistant",
             llm_message)
         st.write(llm_response)
         st.write('Done displaying LLM response')

utils/epfl_meditron_utils.py

@@ -1,7 +1,7 @@
 
 
 
-def get_llm_response(repo, filename, model_type, gpu_layers, prompt):
+def get_llm_response(repo, filename, model_type, gpu_layers, system_message, prompt):
     from transformers import AutoModelForCausalLM, AutoTokenizer, pipeline
 
     model_name_or_path = "TheBloke/meditron-7B-GPTQ"
@@ -40,7 +40,7 @@ def get_llm_response(repo, filename, model_type, gpu_layers, prompt):
     <|im_start|>user
     {prompt}<|im_end|>
     <|im_start|>assistant
-    '''.format(system_message="You are an assistant", prompt=prompt)
+    '''
 
     response = pipe(prompt_template)[0]['generated_text']
     print(response)