This makes sure that the GPU_batch_init will not crash on an assert where the data of vertex buffer needs to be allocated.