Update app.py

app.py

@@ -103,8 +103,6 @@ def bot(history, text_input, image, temperature, length_penalty, repetition_pena
         yield history + [["Please input image or wait for image to be uploaded before clicking submit.", None]]
         return
     
-    chat_history = " ".join([item for sublist in history for item in sublist if item is not None])  # Flatten history
-    
     if arnold_mode:
         system_prompt = """You are Arnold Schwarzenegger, the famous bodybuilder, actor, and former Mr. Olympia. 
         Respond in his iconic style, using his catchphrases and focusing on fitness, bodybuilding, and motivation. 
@@ -112,7 +110,8 @@ def bot(history, text_input, image, temperature, length_penalty, repetition_pena
     else:
         system_prompt = "You are a helpful AI assistant. Provide clear and concise responses to the user's questions about the image and text input."
     
-    prompt = f"{system_prompt}\n{chat_history}\nUSER: <image>\n{text_input}\nASSISTANT:"
+    # Use only the current input for generating the response
+    prompt = f"{system_prompt}\nUSER: <image>\n{text_input}\nASSISTANT:"
     
     response = infer(image, prompt, temperature, length_penalty, repetition_penalty, max_length, min_length, top_p)
     
@@ -124,7 +123,7 @@ def bot(history, text_input, image, temperature, length_penalty, repetition_pena
         history[-1][1] = response[:i+1]
         time.sleep(0.05)
         yield history
-
+        
 with gr.Blocks() as demo:
     gr.Markdown(DESCRIPTION)
     gr.Markdown("""## LLaVA, one of the greatest multimodal chat models is now available in Transformers with 4-bit quantization! ⚡️