Update src/streamlit_app.py

src/streamlit_app.py

@@ -6,6 +6,7 @@ Version: 3.0 - Production Ready, No Demo Mode
 
 import streamlit as st
 import os
+import tempfile
 
 # Configure page
 st.set_page_config(
@@ -14,6 +15,33 @@ st.set_page_config(
     layout="wide"
 )
 
+# Set up proper cache directories for HF Spaces
+
+
+def setup_cache_directories():
+    """Set up cache directories with proper permissions"""
+    try:
+        # Create cache directories in writable locations
+        cache_dirs = {
+            'HF_HOME': os.path.join(tempfile.gettempdir(), 'huggingface'),
+            'TRANSFORMERS_CACHE': os.path.join(tempfile.gettempdir(), 'huggingface', 'transformers'),
+            'HF_DATASETS_CACHE': os.path.join(tempfile.gettempdir(), 'huggingface', 'datasets'),
+        }
+
+        for env_var, cache_path in cache_dirs.items():
+            os.environ[env_var] = cache_path
+            os.makedirs(cache_path, exist_ok=True)
+            st.info(f"📁 Cache directory set: {env_var} = {cache_path}")
+
+        return True
+    except Exception as e:
+        st.error(f"❌ Failed to set up cache directories: {e}")
+        return False
+
+
+# Set up cache directories
+setup_cache_directories()
+
 # Try to import the fine-tuned model components
 try:
     from fine import ProgrammingEducationAI, ComprehensiveFeedback
@@ -62,6 +90,14 @@ def main():
         st.success("🌐 Using public model - no authentication required")
         st.info(f"📁 Model path: FaroukTomori/codellama-7b-programming-education")
 
+        # Memory optimization info
+        st.markdown("---")
+        st.markdown("### 💾 Memory Optimization")
+        st.info("🔧 8-bit quantization enabled")
+        st.info("📁 Proper cache directories configured")
+        st.info("⚡ Auto device mapping for efficiency")
+        st.warning("⚠️ Model size: ~13GB (quantized to ~7GB)")
+
         # Show if model is loaded in session
         if 'ai_tutor' in st.session_state:
             st.success("✅ Model loaded in session")
@@ -98,11 +134,21 @@ def main():
                             st.info(
                                 f"🔍 Attempting to load model from: {model_path}")
 
+                            # Memory optimization settings for HF Spaces
+                            st.info(
+                                "🔧 Loading with memory optimization for HF Spaces...")
+                            st.info(
+                                "💾 Using 8-bit quantization to reduce memory usage")
+                            st.info(
+                                "📁 Using proper cache directories for permissions")
+
                             ai_tutor = ProgrammingEducationAI(model_path)
                             st.success(
                                 "✅ Model class instantiated successfully")
 
-                            ai_tutor.load_model()
+                            # Load model with memory optimization
+                            ai_tutor.load_model(
+                                load_in_8bit=True, device_map="auto")
                             st.session_state['ai_tutor'] = ai_tutor
                             st.success(
                                 "✅ Fine-tuned model loaded successfully!")