How to use cuda to accelerate when using quick_train, does the method quick_train provide some parameter to specify the device