Commit Graph

  • fff6bc1a93 chore(relative imports): switching to relative local imports within lerobot.datasets CarolinePascal 2026-05-05 11:56:27 +02:00
  • 141304ac78 fix(arguments order): reverting changes in arguments order in StreamingVideoEncoder CarolinePascal 2026-05-05 11:54:03 +02:00
  • 7296ac97af fix(smolvla2): make silent generation failures visible in REPL Pepijn 2026-05-05 11:47:34 +02:00
  • 9cbbcfb6a2 fix(smolvla2): tokenize lang prompt inline before select_action Pepijn 2026-05-05 11:40:18 +02:00
  • 9b3c752b64 chore(format): formatting code, fixing error messages and variable names CarolinePascal 2026-05-05 11:31:23 +02:00
  • fea41b29f5 fix(datasets): probe parquet for language columns before strict cast Pepijn 2026-05-05 11:31:19 +02:00
  • 7b4d281ef5 fix(smolvla2): build preprocessor fresh, don't round-trip the recipe Pepijn 2026-05-05 11:27:12 +02:00
  • 29bb8bb20e fix(tools): unblock pocket-tts resolution (>=1.0.0,<3.0.0) Pepijn 2026-05-05 11:15:20 +02:00
  • 3fe686ce9f feat(smolvla2): runtime accepts Hub IDs + dataset-driven dry-run Pepijn 2026-05-05 11:08:53 +02:00
  • 3dc73551dd fix(rollout): propagating VideoEncoderConfig to the latest recording modes CarolinePascal 2026-05-05 11:06:45 +02:00
  • a1b8134ef1 fix(smolvla2): train on rendered language batches pepijn 2026-05-05 08:55:56 +00:00
  • 401a217597 chore(ci): increase time stale (#3507) Steven Palma 2026-05-04 22:35:16 +02:00
  • 40094b0464 chore(ci): upgrade docker internal (#3505) Steven Palma 2026-05-04 21:28:52 +02:00
  • ebe6ea34df refactor: clean up import statements Khalil Meftah 2026-05-04 20:20:05 +02:00
  • 2bc273c53b Merge remote-tracking branch 'origin/main' into user/khalil-meftah/2026-02-16-rl-stack-refactor Khalil Meftah 2026-05-04 19:58:03 +02:00
  • 0d60a855be fix(rl): record pre-step observation so (obs, action, next.reward) align in gym_manipulator dataset Khalil Meftah 2026-05-04 19:33:45 +02:00
  • 237bae51e8 feat(default values): applying a consistent naming convention for default RGB cameras video encoder parameters CarolinePascal 2026-05-04 18:05:23 +02:00
  • df8b33fc68 fix(camera_encoder_config): Removing camera_encoder_config from LeRobotDataset, as it's only required in LeRobotDatasetWriter. CarolinePascal 2026-05-04 18:00:14 +02:00
  • 8fa8323c91 fix(annotate): sync language metadata after parquet rewrite pepijn 2026-05-04 15:17:15 +00:00
  • 50e2d7b5f4 chore(doctrings): updating docstrings CarolinePascal 2026-05-04 17:01:11 +02:00
  • 5f7c6ba61d feat(annotate): compact steerable annotation prompts Pepijn 2026-05-04 15:56:52 +02:00
  • fdbfc015a2 fix(peft): fix LoRA resume from Hub (PosixPath + double wrap) (#3485) Jash Shah 2026-05-04 01:52:37 -07:00
  • 223cc8a9e2 feat(smolvla2): inference runtime — select_message + multi-rate REPL Pepijn 2026-04-30 22:04:00 +02:00
  • af6d8ebd5b feat(smolvla2): dual-head forward — flow loss + lm_head text loss Pepijn 2026-04-30 19:54:57 +02:00
  • 37b1eb218a feat(smolvla2): chat-template processor + label mask + predict_actions Pepijn 2026-04-30 19:21:03 +02:00
  • 52e1fd35cb feat(tools): src/lerobot/tools/ — runnable tool registry + SayTool Pepijn 2026-04-30 18:58:04 +02:00
  • 7459dfccb6 feat(policies): scaffold smolvla2 (smolvla + lm_head re-enabled) Pepijn 2026-04-30 18:04:28 +02:00
  • 73740ecf4b feat(annotate): write tool catalog to meta/info.json after annotation Pepijn 2026-04-30 18:51:38 +02:00
  • 1b81e49214 feat(annotate): task rephrasings + video-derived task fallback Pepijn 2026-04-30 16:49:51 +02:00
  • d813c75b76 fix(annotate): align interjections with the actual demo trajectory Pepijn 2026-04-30 16:32:43 +02:00
  • 3434d2ef22 fix(annotate): ground interjections in video + propagate text to plan refresh Pepijn 2026-04-30 16:11:10 +02:00
  • b71e10da6b refactor(annotate): drop dataset-level `tools` parquet column Pepijn 2026-04-30 15:54:37 +02:00
  • 0f6e3230df fix(annotate): decode video frames with PyAV directly Pepijn 2026-04-30 14:19:25 +02:00
  • 2f2e42c4aa log(annotate): warn loudly on first video decode failure Pepijn 2026-04-30 13:48:02 +02:00
  • 5ee0104739 log(annotate): surface resolved frame-provider cameras at startup Pepijn 2026-04-30 13:34:00 +02:00
  • e064cfcb04 fix(annotate): seed Module 3 cameras from camera_keys + camera_key fallback Pepijn 2026-04-30 13:05:08 +02:00
  • b3d9494831 docs(annotate): add HF Jobs runner example for lerobot-annotate Pepijn 2026-04-30 11:00:28 +02:00
  • 1217fdb6f0 feat(annotate): emit VQA per-camera and propagate camera field Pepijn 2026-04-30 10:48:33 +02:00
  • d0388e1142 fix(annotate): transcode subclips to H.264 instead of stream-copy Pepijn 2026-04-29 17:00:39 +02:00
  • 524aa59faa feat(annotate): pack multiple vllm replicas per GPU via num_gpus Pepijn 2026-04-29 16:11:51 +02:00
  • 27f7829b09 feat(annotate): forward chat_template_kwargs to OpenAI extra_body Pepijn 2026-04-29 15:00:23 +02:00
  • 7f8bf108e8 fix(annotate): include prompt .txt files in wheel Pepijn 2026-04-29 11:47:03 +02:00
  • 855ff027f8 refactor(annotate): drop HF Inference Providers code path Pepijn 2026-04-29 00:53:08 +02:00
  • 3b797bb118 feat(annotate): --vlm.push_to_hub uploads the annotated dataset Pepijn 2026-04-29 00:28:38 +02:00
  • aea04721ae feat(annotate): parallelize episodes within each module phase Pepijn 2026-04-28 23:59:02 +02:00
  • ab5479129a fix(annotate): probe /v1/models for spawn-helper readiness Pepijn 2026-04-28 23:47:52 +02:00
  • e6d4ac6f02 fix(annotate): lock-protect per-line writes for parallel server streams Pepijn 2026-04-28 23:19:37 +02:00
  • 5722d365c5 feat(annotate): client_concurrency for parallel in-flight requests Pepijn 2026-04-28 23:07:48 +02:00
  • 3d7e60cee4 feat(annotate): parallel_servers spawns N independent vllm replicas Pepijn 2026-04-28 23:06:20 +02:00
  • 7b767d4d60 feat(annotate): per-episode progress logs in executor Pepijn 2026-04-28 22:56:03 +02:00
  • f1e3ab7794 fix(annotate): don't crash pipeline on persistent JSON parse failure Pepijn 2026-04-28 22:33:54 +02:00
  • 585341ba9f fix(annotate): robust JSON extraction (think tags + first balanced object) Pepijn 2026-04-28 22:15:25 +02:00
  • 23ff346027 fix(annotate): stream child stdout char-by-char so tqdm \\r progress flushes Pepijn 2026-04-28 21:58:12 +02:00
  • 3c5cbe7af4 test(annotate): adjust video-block test for fps-based frame sampling Pepijn 2026-04-28 19:49:08 +02:00
  • f2cbd97635 feat(annotate): Module 1 samples image frames at fps rate Pepijn 2026-04-28 19:48:25 +02:00
  • c06c8d594a feat(annotate): use cached HF token from huggingface-cli login Pepijn 2026-04-28 19:36:24 +02:00
  • cd495a3a9d feat(annotate): default to HF Inference Providers, no local GPU needed Pepijn 2026-04-28 19:33:34 +02:00
  • c99ac45cd1 feat(annotate): one-flag HF Inference Providers backend Pepijn 2026-04-28 19:28:13 +02:00
  • 13aaafeae0 fix(annotate): omit mm_processor_kwargs by default; transformers serve rejects it Pepijn 2026-04-28 19:11:58 +02:00
  • 2129648bf4 fix(annotate): mm_processor_kwargs in extra_body; inline file URLs as data URLs Pepijn 2026-04-28 18:53:43 +02:00
  • f5cd3f6e4e fix(annotate): detect server ready via stdout banner, not /v1/models polls Pepijn 2026-04-28 17:43:47 +02:00
  • ecf5766301 fix(annotate): visible auto_serve via stdout prints + live server log stream Pepijn 2026-04-28 17:34:36 +02:00
  • 11597d4f71 fix(annotate): auto_serve defaults to True; probe before spawning Pepijn 2026-04-28 17:14:48 +02:00
  • 8b9c598cf4 feat(annotate): auto_serve mode spawns and tears down inference server Pepijn 2026-04-28 16:39:50 +02:00
  • b325475b38 feat(annotate): video_url block for openai backend Pepijn 2026-04-28 16:26:19 +02:00
  • ef137ff86a feat(annotate): openai-compatible backend for transformers/ktransformers serve Pepijn 2026-04-28 16:24:11 +02:00
  • c5df821a96 fix(annotate): use vllm.chat() API for multimodal prompts Pepijn 2026-04-28 16:20:50 +02:00
  • 7ec3d7999c fix(annotate): drop guided_decoding=dict (api differs across vllm) Pepijn 2026-04-28 16:14:31 +02:00
  • 712d63abbd fix(annotate): tolerate decoder returning fewer frames than requested Pepijn 2026-04-28 14:21:38 +02:00
  • 6653999983 fix(annotate): default video decode backend to pyav Pepijn 2026-04-28 14:10:57 +02:00
  • 4bdbedc9a0 fix(annotate): default trust_remote_code=False for HF loaders Pepijn 2026-04-28 13:39:55 +02:00
  • e240305e8e fix(annotate): default transformers backend to manual GPU placement Pepijn 2026-04-28 13:27:19 +02:00
  • ccd189b264 fix(annotate): LEROBOT_DISABLE_CUDNN escape hatch for conv3d crash Pepijn 2026-04-28 12:05:00 +02:00
  • ef1242bbd4 fix(annotate): expose gpu_memory_utilization and max_model_len for vllm Pepijn 2026-04-28 11:29:22 +02:00
  • ebf4a04d41 fix(annotate): pass trust_remote_code=True to HF auto-classes Pepijn 2026-04-28 09:44:15 +02:00
  • 4419b4ef1b fix(annotate): low_cpu_mem_usage=True on transformers load path Pepijn 2026-04-27 23:29:20 +02:00
  • ff06ca82d2 fix(annotate): use device_map='auto' for transformers backend Pepijn 2026-04-27 23:05:14 +02:00
  • fcb01e73eb fix(annotate): try AutoModelForImageTextToText first, fall back to AutoModelForVision2Seq Pepijn 2026-04-27 22:31:56 +02:00
  • 268f8d1f53 fix(annotate): replace Literal types with str for older draccus Pepijn 2026-04-27 18:26:57 +02:00
  • 663fff0ae2 feat(annotate): Module 1 sees the whole episode as one video block Pepijn 2026-04-27 17:08:36 +02:00
  • 9d6af804bf feat(annotate): attach camera keyframes to module prompts; default to Qwen3.6-27B-FP8 Pepijn 2026-04-27 16:58:45 +02:00
  • f763f85213 feat: language annotation pipeline (PR 2/3) Pepijn 2026-04-27 16:22:51 +02:00
  • e3e9374e2c feat(language): tool catalog in meta/info.json + LeRobotDatasetMetadata.tools Pepijn 2026-04-30 18:44:58 +02:00
  • c1a0c601e2 feat(language): task_aug style + automatic ${task} rephrasing rotation Pepijn 2026-04-30 16:45:39 +02:00
  • d4a568ee6c fix(rl): update gripper position key and handle action absence during reset Khalil Meftah 2026-04-30 14:56:40 +02:00
  • 016799dfa1 chore(format): formatting code CarolinePascal 2026-04-30 14:42:37 +02:00
  • 51b9038458 chore(PyAV): cleaning up PyAV utils and encoding parameters checks to stick to the minimun required tooling. CarolinePascal 2026-04-30 14:23:48 +02:00
  • d656da8ccc fix(pi): keep training sampling outside compiled forwards (#3487) Haoming Song 2026-04-30 19:21:17 +08:00
  • 1ca38d9748 fix(language): drop motion from VIEW_DEPENDENT_STYLES Pepijn 2026-04-30 10:54:12 +02:00
  • 5a6aa64570 feat(language): per-camera tagging on view-dependent styles Pepijn 2026-04-30 10:48:17 +02:00
  • cc9a2e5c99 chore(format): fixing formatting issues CarolinePascal 2026-04-24 17:19:19 +02:00
  • a2376389f9 test(new): adding new tests for encoding related features CarolinePascal 2026-04-24 17:15:41 +02:00
  • 57a619ab02 test(existing): adapting existing tests CarolinePascal 2026-04-24 17:15:24 +02:00
  • 7f624adcc5 chore(duplicate): removing duplicate get_codec_options definition CarolinePascal 2026-04-24 17:10:30 +02:00
  • 375cf1fdf3 feat(pyav checks): making pyav parameters checks more robust CarolinePascal 2026-04-24 16:46:13 +02:00
  • b2c2bb7641 feat(VideoEncoderConfig init): making VideoEncoderConfig more robust and adaptable to multiple backends CarolinePascal 2026-04-24 16:43:54 +02:00
  • 4a87ee1537 fix(concatenation compatibility): adding compatibility check when concatenating video files CarolinePascal 2026-04-23 20:09:15 +02:00
  • e44f86e516 feat(metadata): adding encoding parameters in dataset metadata CarolinePascal 2026-04-22 21:49:50 +02:00
  • a0e3acdb67 chore(docs): updating the docs CarolinePascal 2026-04-22 21:18:58 +02:00
  • 38ff579bcc feat(VideoEncoderConfig): propagating the VideoEncoderConfig in the codebase CarolinePascal 2026-04-22 21:18:29 +02:00