S2T is a method from paper "Self-Supervised Temporal Graph Learning with Temporal and Structural Intensity Alignment", you can find it via arXiv: https://arxiv.org/abs/2302.07491
We have uploaded the runnable code version, which will be further improved after the acceptance of the paper.
Our code is improved based on TREND, which can be referenced: https://github.com/WenZhihao666/TREND