Update app.py

app.py

@@ -2,85 +2,71 @@ import os
 import time
 import spaces
 import torch
-from transformers import AutoModelForCausalLM, AutoTokenizer, TextIteratorStreamer
+from transformers import AutoModelForCausalLM, AutoTokenizer, TextIteratorStreamer, BitsAndBytesConfig
 import gradio as gr
 from threading import Thread
 
 HF_TOKEN = os.environ.get("HF_TOKEN", None)
 MODEL = "AGI-0/Art-v0-3B"
 
-device = "cuda"  # Use "cpu" if no GPU available
+TITLE = """<h2>Link to the model: <a href="https://huggingface.co/AGI-0/Art-v0-3B">click here</a></h2>"""
 
-tokenizer = AutoTokenizer.from_pretrained(MODEL)
-model = AutoModelForCausalLM.from_pretrained(
-    MODEL, torch_dtype=torch.bfloat16, device_map="auto"
-)
-end_of_sentence = tokenizer.convert_tokens_to_ids("<|im_end|>")
-
-class ConversationManager:
-    def __init__(self):
-        self.user_history = []  # User-facing history with formatting
-        self.model_history = []  # Model-facing history without formatting
-
-    def add_exchange(self, user_message, model_response):
-        formatted_response = self.format_response(model_response)
-        self.model_history.append((user_message, model_response))
-        self.user_history.append((user_message, formatted_response))
-        print(f"\nModel History Updated: {self.model_history}")
-        print(f"\nUser History Updated: {self.user_history}")
-
-    def format_response(self, response):
-        """Format response for UI while keeping raw text for model."""
-        if "<|end_reasoning|>" in response:
-            parts = response.split("<|end_reasoning|>")
-            reasoning, rest = parts[0], parts[1] if len(parts) > 1 else ""
-            return f"<details><summary>Click to see reasoning</summary>\n\n{reasoning}\n\n</details>\n\n{rest}"
-        return response
+PLACEHOLDER = """
+<center>
+<p>Hi! How can I help you today?</p>
+</center>
+"""
 
-    def get_user_history(self):
-        return self.user_history
+CSS = """
+.duplicate-button {
+    margin: auto !important;
+    color: white !important;
+    background: black !important;
+    border-radius: 100vh !important;
+}
+h3 {
+    text-align: center;
+}
+"""
 
-    def get_model_history(self):
-        return self.model_history
+device = "cuda"  # for GPU usage or "cpu" for CPU usage
 
-conversation_manager = ConversationManager()
+tokenizer = AutoTokenizer.from_pretrained(MODEL)
+model = AutoModelForCausalLM.from_pretrained(
+    MODEL,
+    torch_dtype=torch.bfloat16,
+    device_map="auto")
+end_of_sentence = tokenizer.convert_tokens_to_ids("<|im_end|>")
 
 @spaces.GPU()
 def stream_chat(
-    message: str,
+    message: str, 
     history: list,
-    system_prompt: str,
-    temperature: float = 0.2,
-    max_new_tokens: int = 4096,
-    top_p: float = 1.0,
-    top_k: int = 1,
+    system_prompt: str = "",  # Include system prompt
+    temperature: float = 0.2, 
+    max_new_tokens: int = 4096, 
+    top_p: float = 1.0, 
+    top_k: int = 1, 
     penalty: float = 1.1,
 ):
-    print(f'User Message: {message}')
-    
-    model_history = conversation_manager.get_model_history()
-    print(f'Model History: {model_history}')
-    
+    print(f'message: {message}')
+    print(f'history: {history}')
+
     conversation = []
-    for prompt, answer in model_history:
+    for prompt, answer in history:
         conversation.extend([
-            {"role": "user", "content": prompt},
+            {"role": "user", "content": prompt}, 
             {"role": "assistant", "content": answer},
         ])
+
     conversation.append({"role": "user", "content": message})
 
-    print(f'Formatted Conversation for Model: {conversation}')
+    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)
     
-    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,
+        input_ids=input_ids, 
         max_new_tokens=max_new_tokens,
         do_sample=False if temperature == 0 else True,
         top_p=top_p,
@@ -91,35 +77,43 @@ def stream_chat(
         streamer=streamer,
     )
 
-    buffer = ""
-    original_response = ""
-
     with torch.no_grad():
         thread = Thread(target=model.generate, kwargs=generate_kwargs)
         thread.start()
+    
+    buffer = ""
+    user_buffer = ""
+    found_token = False
+    
+    for new_text in streamer:
+        buffer += new_text
+        user_buffer += new_text
+
+        if "<|end_reasoning|>" in user_buffer and not found_token:
+            parts = user_buffer.split("<|end_reasoning|>")
+            reasoning = parts[0]
+            rest = parts[1] if len(parts) > 1 else ""
+            user_buffer = f"<details><summary>Click to see reasoning</summary>\n\n{reasoning}\n\n</details>\n\n{rest}"
+            found_token = True
+            
+        yield user_buffer
 
-        for new_text in streamer:
-            buffer += new_text
-            original_response += new_text
-            print(f'Streaming: {buffer}')
-            yield conversation_manager.format_response(buffer)
+    history.append((message, buffer)) # Crucial: Append the original buffer
 
-    conversation_manager.add_exchange(message, original_response)
 
-chatbot = gr.Chatbot(height=600, placeholder="<center><p>Hi! How can I help you today?</p></center>")
+chatbot = gr.Chatbot(height=600, placeholder=PLACEHOLDER)
 
-demo = gr.Blocks()
-with demo:
-    gr.HTML("<h2>Link to the model: <a href='https://huggingface.co/AGI-0/Art-v0-3B'>click here</a></h2>")
+with gr.Blocks(css=CSS, theme="soft") as demo:
+    gr.HTML(TITLE)
     gr.DuplicateButton(value="Duplicate Space for private use", elem_classes="duplicate-button")
     gr.ChatInterface(
         fn=stream_chat,
         chatbot=chatbot,
         fill_height=True,
-        additional_inputs_accordion=gr.Accordion(label="⚙️ Parameters", open=False, render=False),
+        additional_inputs_accordion=gr.Accordion(label="⚙️ Parameters", open=False, render=False), # Added system prompt textbox here
         additional_inputs=[
-            gr.Textbox(value="", label="System Prompt", render=False),
-            gr.Slider(minimum=0, maximum=1, step=0.1, value=0.2, label="Temperature", render=False),
+           gr.Textbox(value="", label="System Prompt", lines=2, render=False), # Added system prompt textbox
+           gr.Slider(minimum=0, maximum=1, step=0.1, value=0.2, label="Temperature", render=False),
             gr.Slider(minimum=128, maximum=8192, step=1, value=4096, label="Max new tokens", render=False),
             gr.Slider(minimum=0.0, maximum=1.0, step=0.1, value=1.0, label="top_p", render=False),
             gr.Slider(minimum=1, maximum=50, step=1, value=1, label="top_k", render=False),
@@ -135,4 +129,4 @@ with demo:
     )
 
 if __name__ == "__main__":
-    demo.launch()
+    demo.launch()