Optimize model loading to fix storage limit issues - use streaming and memory cleanup

app.py

@@ -8,6 +8,7 @@ import numpy as np
 import requests
 from io import BytesIO
 import os
+import gc  # For garbage collection
 
 # CIFAR-100 class names
 CIFAR100_CLASSES = [
@@ -91,10 +92,10 @@ model_loaded = False
 model_status = "Not loaded"
 
 def load_model_with_fallbacks():
-    """Try multiple methods to load the model"""
+    """Try multiple methods to load the model with optimized memory usage"""
     global model, model_loaded, model_status
     
-    # Method 1: Try GitHub releases
+    # Method 1: Try GitHub releases with streaming
     try:
         print("🔄 Attempting to load model from GitHub releases...")
         model_url = "https://github.com/santhoshv6/era_v4_s8_assignment/releases/download/v1.0/model_best.pth"
@@ -104,19 +105,35 @@ def load_model_with_fallbacks():
             'User-Agent': 'Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36'
         }
         
-        response = requests.get(model_url, headers=headers, timeout=30)
+        # Stream the download to avoid memory issues
+        response = requests.get(model_url, headers=headers, timeout=60, stream=True)
         response.raise_for_status()
         
-        print(f"✅ Downloaded model: {len(response.content)} bytes")
+        # Load directly from stream to minimize memory usage
+        model_data = BytesIO()
+        for chunk in response.iter_content(chunk_size=8192):
+            model_data.write(chunk)
         
-        # Load the model state dict
-        checkpoint = torch.load(BytesIO(response.content), map_location='cpu')
+        model_data.seek(0)
+        print(f"✅ Downloaded model: {model_data.getbuffer().nbytes} bytes")
+        
+        # Load the model state dict with memory optimization
+        checkpoint = torch.load(model_data, map_location='cpu')
+        
+        # Clear the downloaded data immediately
+        model_data.close()
+        del model_data
         
         if 'state_dict' in checkpoint:
             model.load_state_dict(checkpoint['state_dict'])
             model.eval()
+            
+            # Clear checkpoint data to free memory
+            accuracy = checkpoint.get('test_acc', 77.45)
+            del checkpoint
+            gc.collect()  # Force garbage collection
+            
             model_loaded = True
-            accuracy = checkpoint.get('test_acc', 77.45)  # Default to known accuracy
             model_status = f"✅ Loaded from GitHub (Accuracy: {accuracy:.2f}%)"
             print(f"✅ Model loaded successfully! Accuracy: {accuracy:.2f}%")
             return True