Upload 25 files

app.py

@@ -2,39 +2,73 @@ import os
 import io
 import sys
 
+# Set up logging early
+import logging
+logging.basicConfig(
+    level=logging.INFO, 
+    format='%(asctime)s - %(levelname)s - %(message)s',
+    handlers=[
+        logging.StreamHandler(sys.stdout)
+    ]
+)
+logger = logging.getLogger(__name__)
+
+# Add current directory to Python path
+sys.path.insert(0, '/app/EmoVIT')
+sys.path.insert(0, '/app/EmoVIT/lib')
+
 # Set cache directories before importing any ML libraries
 os.environ['TRANSFORMERS_CACHE'] = os.environ.get('TRANSFORMERS_CACHE', '/app/.cache/transformers')
 os.environ['HF_HOME'] = os.environ.get('HF_HOME', '/app/.cache/huggingface') 
 os.environ['TORCH_HOME'] = os.environ.get('TORCH_HOME', '/app/.cache/torch')
 os.environ['HF_DATASETS_CACHE'] = os.environ.get('HF_DATASETS_CACHE', '/app/.cache/datasets')
+os.environ['PYTHONUNBUFFERED'] = '1'
 
 # Create cache directories if they don't exist
 for cache_dir in ['/app/.cache/transformers', '/app/.cache/huggingface', '/app/.cache/torch', '/app/.cache/datasets']:
     os.makedirs(cache_dir, exist_ok=True)
 
-import torch
-from flask import Flask, render_template, request, jsonify, url_for
-from PIL import Image
-import base64
-import logging
+logger.info("🔧 Environment setup complete")
+logger.info(f"PYTHONPATH: {sys.path[:3]}")
 
-# Safe import with error handling
+# Import basic dependencies
 try:
+    import torch
+    from flask import Flask, render_template, request, jsonify, url_for
+    from PIL import Image
+    import base64
     import numpy as np
+    logger.info("✅ Basic dependencies loaded successfully")
+except ImportError as e:
+    logger.error(f"❌ Failed to import basic dependencies: {e}")
+    sys.exit(1)
+
+# Safe import with error handling for LAVIS
+try:
     # Check numpy version compatibility
     numpy_version = np.__version__
-    logging.info(f"NumPy version: {numpy_version}")
+    logger.info(f"NumPy version: {numpy_version}")
     
     from transformers import AutoTokenizer
+    logger.info("✅ Transformers imported successfully")
+    
+    # Try to import LAVIS components
+    import lavis
+    logger.info("✅ LAVIS base imported successfully")
+    
     from blip2_vicuna_instruct import Blip2VicunaInstruct
+    logger.info("✅ Blip2VicunaInstruct imported successfully")
+    
     MODEL_AVAILABLE = True
-    logging.info("All imports successful")
+    logger.info("✅ All imports successful - Full model mode enabled")
 except ImportError as e:
-    logging.error(f"Model import failed: {e}")
+    logger.error(f"❌ Model import failed: {e}")
+    logger.info("🔄 Running in demo mode without full model capabilities")
     MODEL_AVAILABLE = False
     Blip2VicunaInstruct = None
 except Exception as e:
-    logging.error(f"Unexpected error during import: {e}")
+    logger.error(f"❌ Unexpected error during import: {e}")
+    logger.info("🔄 Running in demo mode without full model capabilities")
     MODEL_AVAILABLE = False
     Blip2VicunaInstruct = None
 
@@ -50,12 +84,23 @@ def load_model():
     global model, device
     
     if not MODEL_AVAILABLE:
-        logging.error("Model is not available due to import errors")
+        logger.warning("⚠️ Model is not available due to import errors")
         return
     
     try:
         device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
-        logging.info(f"Using device: {device}")
+        logger.info(f"🔧 Using device: {device}")
+        
+        # Check if we have CUDA support
+        if torch.cuda.is_available():
+            logger.info(f"🎮 CUDA available: {torch.cuda.get_device_name(0)}")
+        else:
+            logger.info("🖥️ Running on CPU")
+        
+        # For demo purposes, we'll skip actual model loading if LAVIS isn't available
+        if Blip2VicunaInstruct is None:
+            logger.warning("⚠️ Blip2VicunaInstruct class not available - skipping model load")
+            return
         
         # Cấu hình model - có thể cần điều chỉnh theo config thực tế
         model_config = {
@@ -74,15 +119,17 @@ def load_model():
             "qformer_text_input": True,
         }
         
+        logger.info("🔄 Initializing model...")
         # Khởi tạo model
         model = Blip2VicunaInstruct(**model_config)
         model.to(device)
         model.eval()
         
-        logging.info("Model loaded successfully!")
+        logger.info("✅ Model loaded successfully!")
         
     except Exception as e:
-        logging.error(f"Error loading model: {str(e)}")
+        logger.error(f"❌ Error loading model: {str(e)}")
+        logger.info("🔄 Continuing in demo mode...")
         model = None
 
 def preprocess_image(image):
@@ -107,7 +154,7 @@ def preprocess_image(image):
         return image_tensor
         
     except Exception as e:
-        logging.error(f"Error preprocessing image: {str(e)}")
+        logger.error(f"❌ Error preprocessing image: {str(e)}")
         return None
 
 def predict_emotion(image_tensor, prompt="What emotion is shown in this image?"):
@@ -139,7 +186,7 @@ def predict_emotion(image_tensor, prompt="What emotion is shown in this image?")
             return result[0] if result else "Unable to predict emotion"
             
     except Exception as e:
-        logging.error(f"Error predicting emotion: {str(e)}")
+        logger.error(f"❌ Error predicting emotion: {str(e)}")
         return f"Error: {str(e)}"
 
 @app.route('/')
@@ -152,7 +199,10 @@ def predict():
     """Handle image upload and prediction"""
     try:
         if not MODEL_AVAILABLE:
-            return jsonify({'error': 'Model is not available due to import errors. Please check dependencies.'}), 500
+            return jsonify({
+                'error': 'Model is not available due to import errors. Please check dependencies.',
+                'details': 'The application is running in demo mode. Full model functionality requires proper LAVIS installation.'
+            }), 500
             
         if 'image' not in request.files:
             return jsonify({'error': 'No image file provided'}), 400
@@ -161,6 +211,8 @@ def predict():
         if file.filename == '':
             return jsonify({'error': 'No image selected'}), 400
         
+        logger.info(f"📷 Processing image: {file.filename}")
+        
         # Đọc và xử lý image
         image = Image.open(io.BytesIO(file.read()))
         
@@ -174,8 +226,10 @@ def predict():
         
         # If model is not loaded, return a fallback response
         if model is None:
-            emotion_result = "Model not loaded - unable to analyze emotion. This might be due to missing model weights or configuration issues."
+            emotion_result = "Model not loaded - unable to analyze emotion. This might be due to missing model weights or configuration issues. Running in demo mode."
+            logger.warning("⚠️ Model not available, returning demo response")
         else:
+            logger.info("🔄 Running model inference...")
             # Preprocess image
             image_tensor = preprocess_image(image)
             if image_tensor is None:
@@ -183,16 +237,19 @@ def predict():
             
             # Predict emotion
             emotion_result = predict_emotion(image_tensor, custom_prompt)
+            logger.info(f"✅ Prediction complete: {emotion_result[:50]}...")
         
         return jsonify({
             'success': True,
             'emotion': emotion_result,
             'image': img_str,
-            'prompt': custom_prompt
+            'prompt': custom_prompt,
+            'model_available': MODEL_AVAILABLE,
+            'model_loaded': model is not None
         })
         
     except Exception as e:
-        logging.error(f"Error in prediction: {str(e)}")
+        logger.error(f"❌ Error in prediction: {str(e)}")
         return jsonify({'error': f'Prediction failed: {str(e)}'}), 500
 
 @app.route('/health')
@@ -206,15 +263,21 @@ def health():
     })
 
 if __name__ == '__main__':
-    # Setup logging
-    logging.basicConfig(level=logging.INFO)
+    # Setup logging (already done above, but ensure it's configured)
+    logger.info("🚀 Starting EmoVIT Flask application...")
     
     # Load model
-    logging.info("Loading model...")
+    logger.info("📝 Loading model...")
     load_model()
     
+    if MODEL_AVAILABLE and model is not None:
+        logger.info("✅ Model loaded successfully - Full functionality available")
+    else:
+        logger.warning("⚠️ Model not available - Running in demo mode")
+    
     # Determine port for Hugging Face Spaces
     port = int(os.environ.get("PORT", 7860))
+    logger.info(f"🌐 Starting server on port {port}")
     
-    # Run app
+    # Run app with proper logging
     app.run(host="0.0.0.0", port=port, debug=False)

start.sh

@@ -19,12 +19,15 @@ export HF_HOME=${HF_HOME:-/app/.cache/huggingface}
 export TORCH_HOME=${TORCH_HOME:-/app/.cache/torch}
 export HF_DATASETS_CACHE=${HF_DATASETS_CACHE:-/app/.cache/datasets}
 export HUGGINGFACE_HUB_CACHE=${HUGGINGFACE_HUB_CACHE:-/app/.cache/huggingface/hub}
+export PYTHONPATH=/app/EmoVIT:/app/EmoVIT/lib:$PYTHONPATH
+export PYTHONUNBUFFERED=1
 
 echo "✅ Cache directories set up"
 echo "🔧 Environment variables:"
 echo "  TRANSFORMERS_CACHE=$TRANSFORMERS_CACHE"
 echo "  HF_HOME=$HF_HOME" 
 echo "  TORCH_HOME=$TORCH_HOME"
+echo "  PYTHONPATH=$PYTHONPATH"
 
 # Test dependencies
 echo "🔍 Testing dependencies..."
@@ -38,25 +41,37 @@ try:
     print(f'✅ NumPy {numpy.__version__}')
 except Exception as e:
     print(f'❌ NumPy: {e}')
+    sys.exit(1)
 
 try:
     import cv2
     print(f'✅ OpenCV {cv2.__version__}')
 except Exception as e:
     print(f'❌ OpenCV: {e}')
+    sys.exit(1)
 
 try:
     import torch
     print(f'✅ PyTorch {torch.__version__}')
 except Exception as e:
     print(f'❌ PyTorch: {e}')
+    sys.exit(1)
 
 try:
     import transformers
     print(f'✅ Transformers {transformers.__version__}')
 except Exception as e:
     print(f'❌ Transformers: {e}')
+    sys.exit(1)
+
+try:
+    import lavis
+    print(f'✅ LAVIS imported successfully')
+except Exception as e:
+    print(f'❌ LAVIS: {e}')
+    print('Warning: LAVIS not available, app will run in demo mode')
 "
 
 echo "🌐 Starting Flask application..."
-exec python3 app.py
+# Use unbuffered Python output for better logging in containers
+exec python3 -u app.py