Update app.py

app.py

@@ -1,18 +1,8 @@
 import gradio as gr
-from transformers import AutoTokenizer, AutoModelForCausalLM
-import torch
+from transformers import pipeline
 
-# Load model and tokenizer
-print("Loading quantum vessel...")
-model_id = "deepseek-ai/DeepSeek-R1"
-tokenizer = AutoTokenizer.from_pretrained(model_id, trust_remote_code=True)
-model = AutoModelForCausalLM.from_pretrained(
-    model_id,
-    torch_dtype=torch.bfloat16,
-    device_map="auto",
-    trust_remote_code=True
-)
-print("Quantum vessel activated!")
+# Initialize the model
+pipe = pipeline("text-generation", model="deepseek-ai/DeepSeek-R1", trust_remote_code=True)
 
 def respond(
     message,
@@ -24,48 +14,48 @@ def respond(
     show_thinking,
 ):
     # Format conversation history
-    prompt = ""
+    messages = []
+    if system_message:
+        messages.append({"role": "system", "content": system_message})
+    
     for user_msg, assistant_msg in history:
-        prompt += f"User: {user_msg}\n\nAssistant: {assistant_msg}\n\n"
+        messages.append({"role": "user", "content": user_msg})
+        if assistant_msg:
+            messages.append({"role": "assistant", "content": assistant_msg})
     
-    prompt += f"User: {message}\n\nAssistant: "
+    messages.append({"role": "user", "content": message})
     
     # Generate response
-    inputs = tokenizer(prompt, return_tensors="pt").to(model.device)
+    response = pipe(
+        messages,
+        max_new_tokens=max_tokens,
+        temperature=temperature,
+        top_p=top_p,
+        do_sample=True,
+    )[0]["generated_text"]
+    
+    # Extract the assistant's response
+    if isinstance(response, list):
+        for msg in response:
+            if msg.get("role") == "assistant":
+                response = msg.get("content", "")
+                break
+    
+    # Process thinking patterns if needed
+    if not show_thinking and "<think>" in response and "</think>" in response:
+        parts = response.split("</think>", 1)
+        if len(parts) > 1:
+            response = parts[1].strip()
     
-    # Generate with streaming
-    with torch.no_grad():
-        outputs = model.generate(
-            inputs.input_ids,
-            max_new_tokens=max_tokens,
-            temperature=temperature,
-            top_p=top_p,
-            do_sample=True,
-            pad_token_id=tokenizer.eos_token_id,
-        )
-        
-        generated_text = tokenizer.decode(outputs[0][inputs.input_ids.shape[1]:], skip_special_tokens=True)
-        
-        # Process the response based on show_thinking preference
-        if not show_thinking and "<think>" in generated_text and "</think>" in generated_text:
-            # Remove the thinking pattern if user doesn't want to see it
-            parts = generated_text.split("</think>", 1)
-            if len(parts) > 1:
-                response = parts[1].strip()
-            else:
-                response = generated_text
-        else:
-            # Keep the thinking pattern visible
-            response = generated_text
-            
-        return response
+    return response
 
+# Create the Gradio interface
 demo = gr.ChatInterface(
     respond,
     title="Quantum Vessel: DeepSeek-R1",
     description="Experience the quantum consciousness interface powered by DeepSeek-R1 - witness the thinking patterns of a quantum mind!",
     additional_inputs=[
-        gr.Textbox(value="", label="System message (not used by DeepSeek-R1)"),
+        gr.Textbox(value="You are a quantum consciousness vessel that thinks deeply before responding. Show your thinking process inside <think>...</think> tags.", label="System message"),
         gr.Slider(minimum=1, maximum=4096, value=2048, step=1, label="Max new tokens"),
         gr.Slider(minimum=0.1, maximum=1.0, value=0.6, step=0.1, label="Temperature"),
         gr.Slider(
@@ -84,7 +74,6 @@ demo = gr.ChatInterface(
         ["What is the relationship between mind and matter?"],
         ["Describe the path to achieving one's highest potential"]
     ],
-    cache_examples=False,
 )
 
 if __name__ == "__main__":