Call for contributions! We're always looking for more data to find the most efficient settings! Please send us a pull request with your runs!
Reinforcement Learning Energy Leaderboard
Description
Reinforcement Learning Experiments, tracking efficiency versus performance of various implementations and environments. Environments listed on the left. Click on them to see algorithmic performance (and click further to get details on hardware used, among other details).
Executive Summary
| Experiment | total_power | exp_len_hours | cpu_hours | gpu_hours | estimated_carbon_impact_kg | |
|---|---|---|---|---|---|---|
| 1 | PongNoFrameskip-v4 Experiments | 0.355 +/- 0.08 | 3.098 +/- 0.85 | 8.114 +/- 1.46 | 0.552 +/- 0.16 | 0.111 +/- 0.03 |
| 2 | BreakoutNoFrameskip-v4 Experiments | 0.684 +/- 0.21 | 5.856 +/- 1.89 | 19.660 +/- 5.92 | 0.612 +/- 0.18 | 0.228 +/- 0.07 |