Weakly Supervised Semantic Segmentation With Clip