PRED: Pre-training via Semantic Rendering on LiDAR Point Clouds | Read Paper on Bytez