About contrastive loss and learning rate

[song-leap]

Thank you so much for sharing your work! The model crashed when I tried to port graph-based contrastive loss to other work. It was initially determined that the learning rate was set too high. Did you encounter the same problem when transferring your ideas to other models?