{"data":{"slug":"octo-models-octo","name":"octo","tagline":"Transformer-based robot policy trained on 800k robot trajectories","github_url":"https://github.com/octo-models/octo","owner":"octo-models","repo":"octo","owner_avatar_url":"https://avatars.githubusercontent.com/u/153724612?v=4","primary_language":"Python","stars":1778,"forks":280,"topics":[],"archived":false,"github_pushed_at":"2024-07-31T00:26:15+00:00","maintenance_label":"Dormant","stars_delta_30d":56,"url":"https://www.graphcanon.com/tools/octo-models-octo","markdown_url":"https://www.graphcanon.com/tools/octo-models-octo.md","api_url":"https://www.graphcanon.com/api/graphcanon/tools/octo-models-octo","graph_url":"https://www.graphcanon.com/api/graphcanon/graph?tool=octo-models-octo","description":"Octo is a transformer-based robot policy trained on a diverse mix of 800k robot trajectories.","homepage_url":"https://octo-models.github.io/","license":"MIT","open_issues":96,"watchers":19,"ai_summary":"Octo is a Python-based repository that provides a transformer-based robot policy trained on a large dataset of robot trajectories. It supports installation on CPU, GPU, and TPU, with detailed instructions for setting up the environment and dependencies, including Jax for GPU and TPU support.","readme_excerpt":"## Installation\n```bash\nconda create -n octo python=3.10\nconda activate octo\npip install -e .\npip install -r requirements.txt\n```\nFor GPU:\n```bash\npip install --upgrade \"jax[cuda11_pip]==0.4.20\" -f https://storage.googleapis.com/jax-releases/jax_cuda_releases.html\n```\n\nFor TPU\n```bash\npip install --upgrade \"jax[tpu]==0.4.20\" -f https://storage.googleapis.com/jax-releases/libtpu_releases.html\n```\nSee the [Jax Github page](https://github.com/google/jax) for more details on installing Jax.\n\nTest the installation by finetuning on the debug dataset:\n```bash\npython scripts/finetune.py --config.pretrained_path=hf://rail-berkeley/octo-small-1.5 --debug\n```","github_created_at":"2023-12-13T09:58:56+00:00","created_at":"2026-07-11T23:09:55.456023+00:00","updated_at":"2026-09-20T04:29:29.478948+00:00","categories":[{"slug":"model-training","name":"Model Training","url":"https://www.graphcanon.com/categories/model-training","markdown_url":"https://www.graphcanon.com/categories/model-training.md","api_url":"https://www.graphcanon.com/api/graphcanon/categories/model-training"}],"tags":[{"slug":"finetuning","name":"finetuning"},{"slug":"gpu","name":"gpu"},{"slug":"jax","name":"jax"},{"slug":"robotics","name":"robotics"},{"slug":"tpu","name":"tpu"},{"slug":"transformer","name":"transformer"}],"trust":{"provenance":{"is_fork":false,"github_id":731075982,"owner_type":"Organization","methodology":"github_public_v1","parent_repo":null,"near_duplicate_slugs":[]},"computed_at":"2026-09-18T08:16:08.407Z","maintenance":{"label":"Dormant","score":18,"methodology":"github_public_v1","releases_90d":0,"days_since_push":779,"last_release_at":"2024-05-24T00:28:01Z","stars_delta_30d":56,"open_issues_delta_30d":0},"security_summary":{"status":"findings","scanner":"osv@v1","low_count":54,"high_count":0,"last_scan_at":"2026-09-18T08:16:11.043Z","medium_count":0,"scan_profile":"deps","critical_count":0}},"capability_facts":{"scan":{"source":"repo_scan","observed_at":"2026-09-18T08:16:07.280Z"},"languages":{"value":["python"],"source":"github.language+pyproject.toml","observed_at":"2026-09-18T08:16:07.280Z"},"license_spdx":{"value":"MIT","source":"github.license","observed_at":"2026-09-18T08:16:07.280Z"}},"decision_facts":{"hosting":null,"pricing":null,"requirements":{"notes":["Installation requires setting up a Python 3.10 environment with Conda.","For GPU support, Jax with CUDA 11 is required.","For TPU support, Jax with TPU support is required."],"min_ram_gb":null,"requires_docker":false},"constraints":{"min_ram_gb":null,"requires_docker":false},"when_to_use":["Use Octo when you need a transformer-based policy for robot control tasks that can leverage a large dataset of 800k robot trajectories.","Consider Octo if you are working on projects that require high-performance computing resources, as it supports GPU and TPU installations via Jax.","Octo is suitable for researchers and developers who prefer Python and are comfortable with setting up complex environments using Conda and Jax."],"when_not_to_use":["Avoid Octo if your project does not require the scale of data it was trained on, as the model might be overkill for smaller datasets.","Do not use Octo if you are working in an environment that does not support Python or the specific versions of Jax required for GPU and TPU support.","If your project has strict licensing requirements that do not allow for MIT-licensed software, Octo would not be a suitable choice."],"source":"enrich:decision_facts","observed_at":"2026-09-18T09:38:15.790Z"},"constraint_facets":{"min_ram_gb":null,"requires_docker":false},"decision_summary":[{"label":"Requirements","value":"Installation requires setting up a Python 3.10 environment with Conda.; For GPU support, Jax with CUDA 11 is required.; For TPU support, Jax with TPU support is required."},{"label":"Adopt for","value":"Octo is a transformer-based robot policy trained on 800k robot trajectories, offering support for CPU, GPU, and TPU installations."}]}}