Update app.py

app.py

@@ -359,22 +359,22 @@ You are allowed to make up film and branding names, and do them like 80's, 90's
                 char_limit = compression_chars[compression_level]
                 base_prompt += f" Compress the output to be concise while retaining key visual details. MAX OUTPUT SIZE no more than {char_limit} characters."
 
-            messages = f"<|im_start|>system\nYou are a helpful assistant. Try your best to give best response possible to user.<|im_end|>"
-            messages += f"\n<|im_start|>user\n{base_prompt}\nDescription: {input_text}<|im_end|>\n<|im_start|>assistant\n"
+            messages = f"system\nYou are a helpful assistant. Try your best to give best response possible to user."
+            messages += f"\nuser\n{base_prompt}\nDescription: {input_text}\nassistant\n"
 
             stream = client.text_generation(messages, max_new_tokens=4000, do_sample=True, stream=True, details=True, return_full_text=False)
             output = ""
             for response in stream:
-                if not response.token.text == "<|im_end|>":
+                if not response.token.text == "":
                     output += response.token.text
 
             # Remove specific tokens based on the model
             if model == "Llama 3":
-                output = output.rstrip("<|eot_id|>")
+                output = output.rstrip("")
             elif model == "Mistral":
                 output = output.rstrip("</s>")
             elif model == "Mistral-Nemo":
-                output = output.rstrip("<|im_end|></s>")
+                output = output.rstrip("</s>")
             
             # Clean up the output
             if ": " in output:
@@ -396,7 +396,7 @@ def create_interface():
 
     with gr.Blocks(theme="Nymbo/Nymbo_Theme") as demo:
         
-        gr.HTML(title)
+        gr.HTML("<h1>My AI Art Prompt Generator</h1>")
 
         with gr.Row():
             with gr.Column(scale=2):
@@ -532,4 +532,4 @@ def create_interface():
 
 if __name__ == "__main__":
     demo = create_interface()
-    demo.launch()
+    demo.launch()