We study peer effects in online training participation using unique data from a large-scale online teacher training program. The platform data allow us to observe the accurate duration of attendance for every individual-lecture pair. We classify peer groups as close peers, local peers, and global peers based on their relationships. By controlling for unobserved heterogeneity, we find positive effects of close and local peer appearance on trainees' joining a lecture and on their length of stay in the lecture. However, global peers generate a negative but economically insignificant impact. Peer effects differ by group and increase with the relationship closeness. Using the survey data, we investigate the mechanisms of peer influences and find that social interactions facilitate online peer effects. Peer pressure and reputation concerns also help explain our findings. Our results shed new light on how peer effects can be utilized to improve the effectiveness of online learning.