Update app.py

app.py

@@ -6,17 +6,15 @@ from transformers import AutoModelForCausalLM, AutoTokenizer, TextIteratorStream
 import gradio as gr
 from threading import Thread
 
-# Define constants and configuration
 MODEL_LIST = ["mistralai/mathstral-7B-v0.1"]
 HF_TOKEN = os.environ.get("HF_TOKEN", None)
 MODEL = os.environ.get("MODEL_ID")
 
+TITLE = "<h1><center>MathΣtral</center></h1>"
+
 PLACEHOLDER = """
 <center>
-<p>MathΣtral - Your Math advisor</p>
-<p>Hi! I'm MisMath. A Math advisor. My model is based on mathstral-7B-v0.1. Feel free to ask your questions</p>
-<p>Mathstral 7B is a model specializing in mathematical and scientific tasks, based on Mistral 7B.</p>
-<p>mathstral-7B-v0.1 is the first Mathstral model</p>
+<p>MathΣtral - I'm MisMath,Your Math advisor</p>
 </center>
 """
 
@@ -27,86 +25,78 @@ CSS = """
     background: black !important;
     border-radius: 100vh !important;
 }
-h1 {
+h3 {
     text-align: center;
-    font-size: 2em;
-    color: #333;
 }
 """
 
-TITLE = "<h1><center>MathΣtral - Your Math advisor</center></h1>"
-
-device = "cuda"  # for GPU usage or "cpu" for CPU usage
+device = "cuda" # for GPU usage or "cpu" for CPU usage
 
-# Configuration for model quantization
 quantization_config = BitsAndBytesConfig(
     load_in_4bit=True,
     bnb_4bit_compute_dtype=torch.bfloat16,
     bnb_4bit_use_double_quant=True,
-    bnb_4bit_quant_type="nf4"
-)
+    bnb_4bit_quant_type="nf4")
 
-# Initialize tokenizer and model
 tokenizer = AutoTokenizer.from_pretrained(MODEL)
 model = AutoModelForCausalLM.from_pretrained(
     MODEL,
     torch_dtype=torch.bfloat16,
     device_map="auto",
-    quantization_config=quantization_config
-)
+    quantization_config=quantization_config)
 
-# Define the chat streaming function
 @spaces.GPU()
 def stream_chat(
-    message: str,
+    message: str, 
     history: list,
     system_prompt: str,
-    temperature: float = 0.8,
-    max_new_tokens: int = 1024,
-    top_p: float = 1.0,
-    top_k: int = 20,
+    temperature: float = 0.8, 
+    max_new_tokens: int = 1024, 
+    top_p: float = 1.0, 
+    top_k: int = 20, 
     penalty: float = 1.2,
 ):
-    # Prepare the conversation context
-    conversation_text = system_prompt + "\n"
-    for _, answer in history:
-        conversation_text += f"MisMath: {answer}\n"
-
-    conversation_text += f"User: {message}\nMisMath:"
-
-    # Tokenize the conversation text
-    input_ids = tokenizer(conversation_text, return_tensors="pt").input_ids.to(model.device)
-
+    print(f'message: {message}')
+    print(f'history: {history}')
+
+    conversation = [
+        {"role": "system", "content": system_prompt}
+    ]
+    for prompt, answer in history:
+        conversation.extend([
+            {"role": "user", "content": prompt}, 
+            {"role": "assistant", "content": answer},
+        ])
+
+    conversation.append({"role": "user", "content": message})
+
+    input_ids = tokenizer.apply_chat_template(conversation, add_generation_prompt=True, return_tensors="pt").to(model.device)
+    
     streamer = TextIteratorStreamer(tokenizer, timeout=60.0, skip_prompt=True, skip_special_tokens=True)
-
+    
     generate_kwargs = dict(
-        input_ids=input_ids,
-        max_new_tokens=max_new_tokens,
-        do_sample=False if temperature == 0 else True,
-        top_p=top_p,
-        top_k=top_k,
-        temperature=temperature,
-        eos_token_id=[128001, 128008, 128009],
+        input_ids=input_ids, 
+        max_new_tokens = max_new_tokens,
+        do_sample = False if temperature == 0 else True,
+        top_p = top_p,
+        top_k = top_k,
+        temperature = temperature,
+        eos_token_id=[128001,128008,128009],
         streamer=streamer,
     )
 
     with torch.no_grad():
         thread = Thread(target=model.generate, kwargs=generate_kwargs)
         thread.start()
-
+        
     buffer = ""
-    final_output = ""
     for new_text in streamer:
         buffer += new_text
-        # Extract only the final response after the last "MisMath:"
-        if "MisMath:" in buffer:
-            final_output = buffer.split("MisMath:")[-1].strip()
-        yield final_output
+        yield buffer
 
-# Define the Gradio chatbot component
+            
 chatbot = gr.Chatbot(height=500, placeholder=PLACEHOLDER)
 
-# Define the footer with links
 footer = """
 <div style="text-align: center; margin-top: 20px;">
     <a href="https://www.linkedin.com/in/pejman-ebrahimi-4a60151a7/" target="_blank">LinkedIn</a> |
@@ -117,7 +107,6 @@ footer = """
 </div>
 """
 
-# Create and launch the Gradio interface
 with gr.Blocks(css=CSS, theme="Ajaxon6255/Emerald_Isle") as demo:
     gr.HTML(TITLE)
     gr.DuplicateButton(value="Duplicate Space for private use", elem_classes="duplicate-button")
@@ -128,7 +117,7 @@ with gr.Blocks(css=CSS, theme="Ajaxon6255/Emerald_Isle") as demo:
         additional_inputs_accordion=gr.Accordion(label="⚙️ Parameters", open=False, render=False),
         additional_inputs=[
             gr.Textbox(
-                value="You are a helpful assistant for Math questions and complex calculations and programming and your name is MisMath",
+                value="You are a helpful assistant in mathematical and scientific tasks",
                 label="System Prompt",
                 render=False,
             ),
@@ -183,6 +172,6 @@ with gr.Blocks(css=CSS, theme="Ajaxon6255/Emerald_Isle") as demo:
     )
     gr.HTML(footer)
 
-# Launch the application
+
 if __name__ == "__main__":
-    demo.launch()
+    demo.launch()