{"id":22771717,"url":"https://github.com/stonet2000/rl-robotics-speedrun","last_synced_at":"2025-07-24T00:33:24.237Z","repository":{"id":260522435,"uuid":"881287644","full_name":"StoneT2000/rl-robotics-speedrun","owner":"StoneT2000","description":"speed-running solving robot manipulation tasks","archived":false,"fork":false,"pushed_at":"2024-10-31T22:58:19.000Z","size":2109,"stargazers_count":21,"open_issues_count":0,"forks_count":2,"subscribers_count":2,"default_branch":"main","last_synced_at":"2025-04-13T06:55:12.220Z","etag":null,"topics":[],"latest_commit_sha":null,"homepage":null,"language":"Python","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":"mit","status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/StoneT2000.png","metadata":{"files":{"readme":"README.md","changelog":null,"contributing":null,"funding":null,"license":"LICENSE","code_of_conduct":null,"threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":null,"support":null,"governance":null,"roadmap":null,"authors":null,"dei":null,"publiccode":null,"codemeta":null}},"created_at":"2024-10-31T09:16:41.000Z","updated_at":"2025-03-07T03:33:18.000Z","dependencies_parsed_at":"2025-02-05T13:52:03.662Z","dependency_job_id":"6da8a6f2-5667-4637-8eee-dac57020eb08","html_url":"https://github.com/StoneT2000/rl-robotics-speedrun","commit_stats":null,"previous_names":["stonet2000/rl-robotics-speedrun"],"tags_count":0,"template":false,"template_full_name":null,"purl":"pkg:github/StoneT2000/rl-robotics-speedrun","repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/StoneT2000%2Frl-robotics-speedrun","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/StoneT2000%2Frl-robotics-speedrun/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/StoneT2000%2Frl-robotics-speedrun/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/StoneT2000%2Frl-robotics-speedrun/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/StoneT2000","download_url":"https://codeload.github.com/StoneT2000/rl-robotics-speedrun/tar.gz/refs/heads/main","sbom_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/StoneT2000%2Frl-robotics-speedrun/sbom","host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":266774778,"owners_count":23982246,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2022-07-04T15:15:14.044Z","status":"online","status_checked_at":"2025-07-23T02:00:09.312Z","response_time":66,"last_error":null,"robots_txt_status":null,"robots_txt_updated_at":null,"robots_txt_url":"https://github.com/robots.txt","online":true,"can_crawl_api":true,"host_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub","repositories_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories","repository_names_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repository_names","owners_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners"}},"keywords":[],"created_at":"2024-12-11T16:16:52.375Z","updated_at":"2025-07-24T00:33:24.213Z","avatar_url":"https://github.com/StoneT2000.png","language":"Python","funding_links":[],"categories":[],"sub_categories":[],"readme":"# RL Robotics Speedrun\n\nSpeed-running solving robot manipulation tasks in [ManiSkill](https://github.com/haosulab/ManiSkill). Goal is to simply solve a list of tasks as fast as possible with RL + fixed dense rewards starting from scratch.\n\n\nInspired by the [great speedrunning work done for LLMs by Jordan et. al](https://github.com/KellerJordan/modded-nanogpt)\n\n## Getting Started\n\nYou can run the speedrun/benchmark via your local machine or [google colab](https://colab.research.google.com/github/StoneT2000/rl-robotics-speedrun/blob/main/rl_robotics_speedrun_colab.ipynb). To set up locally, we recommend using conda/mamba to manage dependencies:\n\n```bash\nconda create -n rl-robotics-speedrun python=3.11\nconda activate rl-robotics-speedrun\ngit clone https://github.com/StoneT2000/rl-robotics-speedrun\ncd rl-robotics-speedrun\npip install -e .\n```\n\n## Benchmarking \n\nTo run the benchmark `cd` into one of the folders and run the script.sh file.\n\nThis by default logs results to tensorboard and wandb. You will need to setup a [wandb account](https://wandb.ai/). Wandb helps better display aggregated results. \n\n```bash\ncd records/baseline \u0026\u0026 bash script.sh\n```\n\nThe current standard is to run PPO initialized with random weights on 10 different seeds. By default this does not record evaluation videos. Remove `--no-capture_video` to record videos. Remove `--track` to not use wandb. Each of the 10 runs takes about 2 minutes to run on a 4090.\n\nCurrent best is `records/10312024_cudagraphs` which is standard PPO + GPU Simulation, very few steps per environment during rollouts, and cudagraphs enabled based on [leanrl](https://github.com/pytorch-labs/LeanRL/). Achieves \u003e= 95% success rate after ~80s\n\n\n| Environment | Best Time | Wandb Results |\n|------------|-----------|---------------|\n| PickCube-v1 (state) | 80s | [Link](https://wandb.ai/stonet2000/PPO-ManiSkill-GPU-SpeedRun/workspace?nw=qgul0t4vstq) |\n| PickCube-v1 (visual) | ~ | ~ |\n| PushT-v1 (state) | ~ | ~ |\n| PushT-v1 (visual) | ~ | ~ |\n\nAll results on 4090: [Wandb Link](https://wandb.ai/stonet2000/PPO-ManiSkill-GPU-SpeedRun/workspace?nw=qgul0t4vstq)\n\nAll results on L4 (Google Colab GPU): [Wandb Link](https://wandb.ai/stonet2000/PPO-ManiSkill-GPU-SpeedRun/workspace?nw=i9kpqaqywjd)\n\nFigure of results on 4090 for PickCube-v1 across 10 seeds. Shaded area is the standard error:\n![](./assets/PickCube-v1_eval_return_4090.png)\n![](./assets/PickCube-v1_eval_success_once_4090.png)","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fstonet2000%2Frl-robotics-speedrun","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fstonet2000%2Frl-robotics-speedrun","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fstonet2000%2Frl-robotics-speedrun/lists"}