{"id":51060468,"url":"https://github.com/lancedb/lerobot-lancedb","last_synced_at":"2026-06-23T01:31:07.605Z","repository":{"id":358237052,"uuid":"1236468193","full_name":"lancedb/lerobot-lancedb","owner":"lancedb","description":null,"archived":false,"fork":false,"pushed_at":"2026-05-16T10:45:10.000Z","size":812,"stargazers_count":6,"open_issues_count":0,"forks_count":1,"subscribers_count":0,"default_branch":"main","last_synced_at":"2026-05-16T12:39:41.607Z","etag":null,"topics":[],"latest_commit_sha":null,"homepage":"https://lancedb.github.io/lerobot-lancedb/","language":"Python","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":"other","status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/lancedb.png","metadata":{"files":{"readme":"README.md","changelog":null,"contributing":"CONTRIBUTING.md","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,"zenodo":null,"notice":null,"maintainers":null,"copyright":null,"agents":null,"dco":null,"cla":null}},"created_at":"2026-05-12T09:19:08.000Z","updated_at":"2026-05-16T10:44:56.000Z","dependencies_parsed_at":null,"dependency_job_id":null,"html_url":"https://github.com/lancedb/lerobot-lancedb","commit_stats":null,"previous_names":["lancedb/lerobot-lancedb"],"tags_count":null,"template":false,"template_full_name":null,"purl":"pkg:github/lancedb/lerobot-lancedb","repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/lancedb%2Flerobot-lancedb","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/lancedb%2Flerobot-lancedb/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/lancedb%2Flerobot-lancedb/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/lancedb%2Flerobot-lancedb/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/lancedb","download_url":"https://codeload.github.com/lancedb/lerobot-lancedb/tar.gz/refs/heads/main","sbom_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/lancedb%2Flerobot-lancedb/sbom","scorecard":null,"host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":286080680,"owners_count":34672250,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2026-05-26T15:22:16.424Z","status":"online","status_checked_at":"2026-06-22T02:00:06.391Z","response_time":106,"last_error":null,"robots_txt_status":"success","robots_txt_updated_at":"2025-07-24T06:49:26.215Z","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":"2026-06-23T01:31:07.015Z","updated_at":"2026-06-23T01:31:07.590Z","avatar_url":"https://github.com/lancedb.png","language":"Python","funding_links":[],"categories":[],"sub_categories":[],"readme":"# lerobot-lancedb\n\n📖 **Docs: \u003chttps://lancedb.github.io/lerobot-lancedb/\u003e**\n\nLance-backed datasets for [LeRobot](https://github.com/huggingface/lerobot). Drop-in replacement for `LeRobotDataset` with two storage layouts:\n\n- **`LeRobotLanceDataset`** — per-frame JPEG bytes (lossy, fastest at single-frame access, optional GPU NVJPEG decode).\n- **`LeRobotLanceVideoDataset`** — per-file mp4 bytes stored via Lance blob v2, decoded on the fly with torchcodec. Bit-exact pixels, ~same disk size as upstream.\n\nBoth subclass `LeRobotDataset` so existing trainers / samplers / `isinstance` checks accept them transparently.\n\n## Install\n\n```bash\npip install lerobot-lancedb\n```\n\nFor local development:\n\n```bash\ngit clone https://github.com/lancedb/lerobot-lancedb.git\ncd lerobot-lancedb\npip install -e '.[dev]'\n```\n\n## Quickstart\n\n```bash\n# Convert (recommended path for dtype=video sources)\nlerobot-convert-to-lance-video \\\n    --repo-id=lerobot/aloha_static_cups_open \\\n    --output=./aloha_cups_open_lance_video --overwrite\n```\n\n```python\nfrom lerobot_lancedb import LeRobotLanceVideoDataset\nds = LeRobotLanceVideoDataset(root=\"./aloha_cups_open_lance_video\")\n```\n\nFor the JPEG layout, use `lerobot-convert-to-lance` and `LeRobotLanceDataset` instead. See the [docs](https://lancedb.github.io/lerobot-lancedb/) for the full CLI / API reference.\n\n## Benchmark\n\nRealistic training read pattern (`delta_timestamps`, 8 frames / sample, batch 32, num_workers 4, CPU decode, H100):\n\n| dataset | format | size MB | delta_ts fps | **speedup** |\n|---|---|---:|---:|---:|\n| **pusht** (96×96, 1-cam) | upstream parquet+mp4 | 7.3 | 750 | 1.00× |\n| | `convert_to_lance` (JPEG-95) | 60.0 | 3510 | **4.68×** |\n| | `convert_to_lance --jpeg-quality=100 --jpeg-subsampling=0` | 105.6 | 2909 | 3.88× |\n| | **`convert_to_lance_video`** | **8.0** | 2853 | **3.80×** |\n| **ALOHA cups_open** (480×640, 4-cam) | upstream parquet+mp4 | 485.6 | 18.7 | 1.00× |\n| | `convert_to_lance` (JPEG-95) | 3626.0 | 46.0 | **2.46×** |\n| | `convert_to_lance --jpeg-quality=100 --jpeg-subsampling=0` | 8735.4 | 32.5 | 1.74× |\n| | **`convert_to_lance_video`** | **487.4** | 45.6 | **2.44×** |\n| **Koch lego** (480×640, 2-cam) | upstream parquet+mp4 | 2014.1 | 26.6 | 1.00× |\n| | `convert_to_lance` (JPEG-95) | 8541.0 | 70.8 | **2.66×** |\n| | `convert_to_lance --jpeg-quality=100 --jpeg-subsampling=0` | 17 335.3 | 49.0 | 1.84× |\n| | **`convert_to_lance_video`** | **2015.9** | 53.8 | **2.02×** |\n\nReproducible via [`examples/benchmark_formats.py`](examples/benchmark_formats.py).\n\n## Training parity\n\n`convert_to_lance_video` trains a `DiffusionPolicy` on pusht to **68.4 % gym-pusht success** (seed=42, 500 rollouts) — matches the head-to-head upstream parquet+mp4 result (68.0 %) and the published [`lerobot/diffusion_pusht`](https://huggingface.co/lerobot/diffusion_pusht) (65.4 %).\n\nFull numbers (pusht env-eval + ALOHA cups_open held-out MSE across all storage modes) in [`docs/benchmarks.md`](https://lancedb.github.io/lerobot-lancedb/benchmarks/). Reproducers: [`examples/train_and_eval_lance.py`](examples/train_and_eval_lance.py) and [`examples/aloha_loader_parity.py`](examples/aloha_loader_parity.py).\n\n## Cloud / Hub\n\nBoth readers accept `s3://`, `gs://`, `hf://datasets/...`, `hf://buckets/...` URIs and pick up credentials from the usual env vars (`AWS_*`, `GOOGLE_APPLICATION_CREDENTIALS`, `HF_TOKEN`). Lance does byte-range fetches — no full-dataset download.\n\nPre-converted reference datasets you can paste directly:\n\n```python\nfrom lerobot_lancedb import LeRobotLanceDataset, LeRobotLanceVideoDataset\n\nLeRobotLanceDataset(repo_id=\"lance-format/pusht-lerobot-lancedb\")        # 60 MB JPEG layout\nLeRobotLanceVideoDataset(repo_id=\"lance-format/pusht-lerobot-lancedb-video\")  # 8 MB video-blob layout\n```\n\n\n## License\n\nApache 2.0.\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Flancedb%2Flerobot-lancedb","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Flancedb%2Flerobot-lancedb","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Flancedb%2Flerobot-lancedb/lists"}