Update start.sh

start.sh

@@ -9,8 +9,8 @@ export CUDA_VISIBLE_DEVICES=0  # Use the first GPU if available
 ollama serve &  # Use GPU 0 if available
 
 # Pull the model if not already present
-if ! ollama list | grep -q "tinyllama"; then
-    ollama pull tinyllama
+if ! ollama list | grep -q "llama3.2:1b"; then
+    ollama pull llama3.2:1b
 fi
 
 # Wait for Ollama to start up (use a more robust check)