Update app.py

app.py

@@ -2,7 +2,7 @@ import os
 import time
 import spaces
 import torch
-from transformers import AutoModelForCausalLM, AutoTokenizer, TextIteratorStreamer, BitsAndBytesConfig
+from transformers import AutoModelForCausalLM, AutoTokenizer, TextIteratorStreamer
 import gradio as gr
 from threading import Thread
 
@@ -10,12 +10,15 @@ 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 - I'm MisMath,Your Math advisor</p>
 </center>
 """
 
+
 CSS = """
 .duplicate-button {
     margin: auto !important;
@@ -23,36 +26,25 @@ 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
 
-quantization_config = BitsAndBytesConfig(
-    load_in_4bit=True,
-    bnb_4bit_compute_dtype=torch.bfloat16,
-    bnb_4bit_use_double_quant=True,
-    bnb_4bit_quant_type="nf4")
-
 tokenizer = AutoTokenizer.from_pretrained(MODEL)
 model = AutoModelForCausalLM.from_pretrained(
     MODEL,
     torch_dtype=torch.bfloat16,
     device_map="auto",
-    quantization_config=quantization_config)
+    ignore_mismatched_sizes=True)
 
 @spaces.GPU()
 def stream_chat(
     message: str, 
-    history: list,
-    system_prompt: str,
-    temperature: float = 0.8, 
+    history: list, 
+    temperature: float = 0.3, 
     max_new_tokens: int = 1024, 
     top_p: float = 1.0, 
     top_k: int = 20, 
@@ -61,27 +53,28 @@ def stream_chat(
     print(f'message: {message}')
     print(f'history: {history}')
 
-    # Prepare the conversation as plain text
-    conversation_text = system_prompt + "\n"
+    conversation = []
     for prompt, answer in history:
-        conversation_text += f"User: {prompt}\nAssistant: {answer}\n"
-
-    conversation_text += f"User: {message}\n"
+        conversation.extend([
+            {"role": "user", "content": prompt}, 
+            {"role": "assistant", "content": answer},
+        ])
 
-    # Tokenize the conversation text
-    input_ids = tokenizer(conversation_text, return_tensors="pt").input_ids.to(model.device)
+    conversation.append({"role": "user", "content": message})
 
+    input_text=tokenizer.apply_chat_template(conversation, tokenize=False)
+    inputs = tokenizer.encode(input_text, return_tensors="pt").to(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=inputs, 
+        max_new_tokens = max_new_tokens,
+        do_sample = False if temperature == 0 else True,
+        top_p = top_p,
+        top_k = top_k,
+        temperature = temperature,
         streamer=streamer,
+        pad_token_id = 10,
     )
 
     with torch.no_grad():
@@ -92,8 +85,9 @@ def stream_chat(
     for new_text in streamer:
         buffer += new_text
         yield buffer
+
             
-chatbot = gr.Chatbot(height=500, placeholder=PLACEHOLDER)
+chatbot = gr.Chatbot(height=600, placeholder=PLACEHOLDER)
 
 footer = """
 <div style="text-align: center; margin-top: 20px;">
@@ -114,16 +108,11 @@ with gr.Blocks(css=CSS, theme="Ajaxon6255/Emerald_Isle") as demo:
         fill_height=True,
         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",
-                label="System Prompt",
-                render=False,
-            ),
             gr.Slider(
                 minimum=0,
                 maximum=1,
                 step=0.1,
-                value=0.8,
+                value=0.3,
                 label="Temperature",
                 render=False,
             ),
@@ -168,7 +157,6 @@ with gr.Blocks(css=CSS, theme="Ajaxon6255/Emerald_Isle") as demo:
         ],
         cache_examples=False,
     )
-    gr.HTML(footer)
 
 
 if __name__ == "__main__":