removed gpu

start.sh

@@ -6,7 +6,7 @@ export MKL_NUM_THREADS=4  # Optimize Intel MKL threading
 export CUDA_VISIBLE_DEVICES=0  # Use the first GPU if available
 
 # Start Ollama in the background with optimized settings
-ollama serve --gpu 0 &  # Use GPU 0 if available
+ollama serve &  # Use GPU 0 if available
 
 # Pull the model if not already present
 if ! ollama list | grep -q "tinyllama"; then