Update app.py

app.py

@@ -5,6 +5,7 @@ from qwen_vl_utils import process_vision_info
 import logging
 from typing import List, Dict
 import gc
+import os
 
 # Setup logging
 logging.basicConfig(level=logging.INFO)
@@ -23,39 +24,37 @@ class HealthAssistant:
     def initialize_model(self):
         try:
             logger.info("Loading Qwen2-VL model...")
+            # Initialize model with default settings
             self.model = Qwen2VLForConditionalGeneration.from_pretrained(
                 self.model_name,
-                torch_dtype=torch.bfloat16,
-                attn_implementation="flash_attention_2",
-                device_map="auto"
-            )
-            self.tokenizer = AutoTokenizer.from_pretrained(self.model_name)
-            self.processor = AutoProcessor.from_pretrained(
-                self.model_name,
-                min_pixels=256*28*28,
-                max_pixels=1280*28*28
+                torch_dtype="auto",
+                device_map="auto",
+                trust_remote_code=True
             )
+            
+            # Initialize processor
+            self.processor = AutoProcessor.from_pretrained(self.model_name)
             logger.info("Model loaded successfully")
         except Exception as e:
-            logger.error(f"Error loading model: {e}")
+            logger.error(f"Error initializing model: {e}")
             raise
 
     def generate_response(self, message: str, history: List = None) -> str:
         try:
-            # Format conversation with health context
+            # Format messages for Qwen2-VL
             messages = self._format_messages(message, history)
             
-            # Prepare for inference
+            # Prepare for inference using qwen_vl_utils
             text = self.processor.apply_chat_template(
                 messages,
                 tokenize=False,
                 add_generation_prompt=True
             )
             
-            # Since we're not using images in this case
-            image_inputs, video_inputs = [], []
+            # Process vision info (empty for text-only)
+            image_inputs, video_inputs = process_vision_info(messages)
             
-            # Process inputs
+            # Prepare inputs
             inputs = self.processor(
                 text=[text],
                 images=image_inputs,
@@ -63,21 +62,24 @@ class HealthAssistant:
                 padding=True,
                 return_tensors="pt"
             )
+            
+            # Move to appropriate device
             inputs = inputs.to(self.model.device)
 
             # Generate response
             generated_ids = self.model.generate(
                 **inputs,
-                max_new_tokens=256,
+                max_new_tokens=128,
                 do_sample=True,
                 temperature=0.7,
                 top_p=0.9
             )
 
-            # Decode response
+            # Trim and decode response
             generated_ids_trimmed = [
                 out_ids[len(in_ids):] for in_ids, out_ids in zip(inputs.input_ids, generated_ids)
             ]
+            
             output_text = self.processor.batch_decode(
                 generated_ids_trimmed,
                 skip_special_tokens=True,
@@ -96,24 +98,29 @@ class HealthAssistant:
             return "I apologize, but I encountered an error. Please try again."
 
     def _format_messages(self, message: str, history: List = None) -> List[Dict]:
-        """Format messages for the Qwen2-VL model"""
-        # Add system context
+        """Format messages for Qwen2-VL"""
         messages = []
-        
-        # Add health context
+
+        # Add health context as system message
         health_context = self._get_health_context()
         if health_context:
             messages.append({
                 "role": "system",
-                "content": [{"type": "text", "text": f"Current health information:\n{health_context}"}]
+                "content": [{"type": "text", "text": f"Health Context:\n{health_context}"}]
             })
 
         # Add conversation history
         if history:
-            for user_msg, assistant_msg in history[-3:]:  # Last 3 exchanges
+            for user_msg, assistant_msg in history[-3:]:
                 messages.extend([
-                    {"role": "user", "content": [{"type": "text", "text": user_msg}]},
-                    {"role": "assistant", "content": [{"type": "text", "text": assistant_msg}]}
+                    {
+                        "role": "user",
+                        "content": [{"type": "text", "text": user_msg}]
+                    },
+                    {
+                        "role": "assistant",
+                        "content": [{"type": "text", "text": assistant_msg}]
+                    }
                 ])
 
         # Add current message
@@ -125,7 +132,6 @@ class HealthAssistant:
         return messages
 
     def _get_health_context(self) -> str:
-        """Get health metrics and medications context"""
         context_parts = []
         
         if self.metrics:
@@ -185,7 +191,7 @@ class GradioInterface:
         return "", history
 
     def add_health_metrics(self, weight: float, steps: int, sleep: float) -> str:
-        if not all([weight, steps, sleep]):
+        if not all([weight is not None, steps is not None, sleep is not None]):
             return "⚠️ Please fill in all metrics."
         
         if self.assistant.add_metrics(weight, steps, sleep):
@@ -202,19 +208,14 @@ class GradioInterface:
 
     def create_interface(self):
         with gr.Blocks(title="Health Assistant", theme=gr.themes.Soft()) as demo:
-            gr.Markdown(
-                """
-                # 🏥 AI Health Assistant
-                Powered by Qwen2-VL for intelligent health guidance and monitoring.
-                """
-            )
+            gr.Markdown("# 🏥 AI Health Assistant")
             
             with gr.Tabs():
                 # Chat Interface
                 with gr.Tab("💬 Health Chat"):
                     chatbot = gr.Chatbot(
                         height=450,
-                        show_label=False
+                        show_label=False,
                     )
                     with gr.Row():
                         msg = gr.Textbox(
@@ -262,14 +263,6 @@ class GradioInterface:
                 outputs=[med_status]
             )
 
-            gr.Markdown(
-                """
-                ### ⚠️ Important Note
-                This AI assistant provides general health information only.
-                Always consult healthcare professionals for medical advice.
-                """
-            )
-
         return demo
 
 def main():
@@ -278,8 +271,9 @@ def main():
         demo = interface.create_interface()
         demo.launch(
             share=False,
-            enable_queue=True,
-            max_threads=4
+            server_name="0.0.0.0",
+            server_port=7860,
+            enable_queue=True
         )
     except Exception as e:
         logger.error(f"Error starting application: {e}")