Update app.py

app.py

@@ -158,7 +158,7 @@ from llama_cpp import Llama
 # set GPU_LAYERS to 15 if you have a 8GB GPU so both models can fit in
 # else 35 full layers + XTTS works fine on T4 16GB
 # 5gb per llm, 4gb XTTS -> full layers should fit T4 16GB , 2LLM + XTTS
-GPU_LAYERS=int(os.environ.get("GPU_LAYERS", 35))
+GPU_LAYERS=int(os.environ.get("GPU_LAYERS", 30))
 
 LLAMA_VERBOSE=False
 print("Running LLM Mistral")