Running 1 Estimated Performance Frontiers of Open LLM Leaderboard Tasks ๐ Estimate LLM task performance from pretraining compute