Release Blog

https://huggingface.co/blog/lerobot/lerobot-release-v070

⚠️ Breaking changes (v0.6.1 -> v0.7.0)

v0.7.0 rebuilds installation, training, inference and the hardware layer. A few changes need your attention when upgrading:

Installation

• On Linux, torch and torchvision now come from PyPI as CUDA 13 builds (NVIDIA driver 580.65 or newer); the cu128 index is gone. For older drivers, install torch from a CUDA 12 index first (see the installation guide). macOS needs 14+ (15+ for MPS).
• Pins moved: transformers 5.22, huggingface-hub>=1.31,<3, motorbridge 0.5, datasets<6, Pillow<14, torchcodec<0.18.

Training

• Weights & Biases is replaced by Trackio (the training extra installs trackio): --wandb.* is now --trackio.*, entity/mode have no equivalent, remote jobs with tracking need --trackio.space_id, and checkpoints are no longer uploaded as artifacts by default.
• FSDP1 is replaced by FSDP2/HSDP (--parallelism.dp_shard, --parallelism.dp_replicate, --accelerator.*); mixed precision is set with --accelerator.mixed_precision. FSDP checkpoints saved with lerobot ≤ 0.6.x cannot be resumed (their model.safetensors still loads), and Accelerate YAML/env configuration is rejected unless LEROBOT_ALLOW_ACCELERATE_ENV=1.
• --policy.use_peft=true alone now trains LoRA instead of silently fully fine-tuning.
• Image transforms are batched: GaussianNoise, MotionBlur, GaussianPatchBrightness, RandomShadow, CoarseDropout, GammaCorrection and PlanckianJitter were silent no-ops on uint8 frames and now apply. ImageTransforms takes tensors only; the per-sample classes are deprecated.

Inference & rollout

• lerobot.async_inference and lerobot[async] are removed: use lerobot[remote], lerobot-policy-server and lerobot-rollout --inference.type=remote. lerobot.rollout.inference moved to lerobot.inference.
• select_action / predict_action_chunk may return a PolicyOutput ({"action", "prediction"}) instead of a bare tensor; use policy_output_to_transition when calling them directly.
• With --interpolation_multiplier>1, rollout strategies record one frame per policy action instead of one per tick.

Simulation & environments

• lerobot.envs.<sim> moved to lerobot.sim.integrations.<sim>, env configs are named *EnvConfig (LiberoEnv -> LiberoEnvConfig), and in-process envs return canonical observations (HWC uint8 images, flat float32 state; map camera names with --rename_map). --env.type is unchanged.
• Removed env fields: Aloha/PushT observation_height/observation_width, MetaWorld episode_length/multitask_eval, RoboTwin obs_type/render_mode. MetaWorld's default task is metaworld-push-v3.
• EnvHub make_env is called without cfg and must return an unwrapped SyncVectorEnv/AsyncVectorEnv unless --env.use_native_batch=true.

Policies

• dtype is the single precision key of every policy config; torch_dtype, vlm_dtype, model_params_fp32 and model_dtype are ignored or removed. In Python pass torch.bfloat16, not a string.
• MolmoAct2: enable_lora_vlm, enable_lora_action_expert and train_action_expert_only are replaced by --policy.train_mode_vlm={fft,lora,freeze}; action_mode defaults to continuous.
• EO1: chunk_size and n_action_steps default to 16.
• VLA-JEPA: checkpoints trained before the RoPE fix need --policy.world_model_rope_legacy=true; binarize_gripper_action and pre_snap_gripper_action default to false.

Datasets

• --dataset.streaming=true uses a new disk-less episode-pool reader with --dataset.streaming_* options; Backtrackable is removed.

Hardware

• Feetech and Dynamixel run on one rustypot-backed SerialMotorsBus (lerobot[feetech]/[dynamixel] install rustypot instead of the vendor SDKs). FeetechMotorsBus/DynamixelMotorsBus are deprecated aliases, OperatingMode/TorqueMode give way to bus.set_operating_mode("position"), and the control tables, encoding_utils and the SDK-level helpers (broadcast_ping, scan_port, ...) are removed. Calibration files are unchanged.
• reBot B601: the config is driven by motor_family (dm, rs); gripper_mit_kp/gripper_mit_kd are removed and per-joint gains take a dict instead of a list.

What's Changed

• chore(dependencies): Bump lerobot to 0.6.2 by @imstevenpmwork in #4313
• add third party hardware section by @NikodemBartnik in #4308
• fix(lekiwi): relative target keyerror by @zuoxingdong in #4281
• feat(data): add recipe-driven language supervision by @pkooij in #4182
• feat(train): parallel training framework — FSDP2, HSDP, gradient accumulation, and DCP checkpoints by @HaomingSong in #4010
• feat(train): split phase timing metrics by @pkooij in #4344
• fix(dataset): use conservative bounds for quantile aggregation instead of incorrect weighted mean by @aHiroakiIshikawa in #3804
• fix(train): repair the two CI failures introduced by #4010 by @HaomingSong in #4347
• feat(train): add opt-in EMA of the policy weights (--ema.enable=true) by @alejodosr in #4323
• test(ema): Rename test that trips TruffleHog's Lob detector by @pkooij in #4371
• fix(tests): port the multi-GPU FSDP test off the superseded accelerate YAML flow by @pkooij in #4370
• docs: add API documentation infrastructure by @pkooij in #4348
• chore: sort imports in vla_jepa tests by @pkooij in #4354
• fix(datasets): resolve only recipe-referenced bindings by @pkooij in #4382
• fix(ci): benchmark jobs skipped for forks by @imstevenpmwork in #4411
• fix(rollout): align state order with checkpoint by @pkooij in #4416
• feat(pi05): add optional visual and proprioceptive MEM by @pkooij in #4076
• perf(policies): move image normalization to device in prepare_observation_for_inference by @imstevenpmwork in #4433
• fix(legacy depth markers): unifying legacy depth markers across the codebase by @CarolinePascal in #4428
• Warn when freezing VLA-JEPA Qwen conditioning by @WOLIKIMCHENG in #4421
• fix(control): unify control-loop pacing and cadence reporting by @imstevenpmwork in #4423
• fix: use <= instead of < for timestamp tolerance comparison in video_utils by @walle250ai in #3423
• fix(vla_jepa): add opt-in causal world-model context to stop future-frame leakage by @ravediamond in #4381
• feat(rollout): interactive sessions — chat-style CLI and programmatic rollout control by @imstevenpmwork in #4380
• (add) SONIC decoder by @nepyope in #4090
• feat(policies): vla-jepa refactoring by @Maximellerbach in #4425
• feat(pi05): add optional training-time RTC by @pkooij in #4056
• fix(molmoact2): align continuous training with official recipe by @hq-fang in #4249
• perf(fastwam): add compiled cached action inference by @darthjaja6 in #4491
• fix(rl): guard learner checkpoint saving against non-positive save_freq by @HaomingSong in #4479
• fix(tests): fix peft cli test by @HaomingSong in #4477
• fix(datasets): keep small splits from dropping episodes in split_dataset by @VihaanAgarwal in #4001
• fix(datasets): reject mismatched aggregate roots by @tandede in #4488
• docs: correct calibration guide grammar by @POPOXR in #4508
• feat(datasets): native support for LeRobot datasets stored as LanceDB datasets by @AyushExel in #4363
• chore: update full_tests.yml by @hf-security-analysis[bot] in #4531
• add upper bounds for lancedb dependencies by @AyushExel in #4532
• support non-default backends in foxglove playback by @AyushExel in #4542
• Fix reset config joint positions encoding by @s1lent4gnt in #4545
• add dataset reader registry for storage backends by @AyushExel in #4541
• make eval_split format aware by @AyushExel in #4543
• fix(datasets): answer delta queries through cached single-column views by @Rick0525 in #4314
• fix(kinematics): iterate inverse_kinematics to convergence by @0o8o0-blip in #3613
• feat(rollout): bring your own strategy by @imstevenpmwork in #4559
• dataset-via keep --root as a st...

⚠️ Breaking changes (v0.6.0 -> v0.6.1)

• lerobot.types module was renamed to lerobot.lerobot_types: #4232

What's Changed

• chore(dependencies): Bump lerobot to 0.6.1 by @imstevenpmwork in #3957
• Make doc builds faster by @mishig25 in #3958
• docs(GR00T): Update GR00T 1.7 LIBERO checkpoints by @liorbenhorin in #3961
• feat(annotate): improve VLM subtask annotation (legible contact sheets, seeded relabeling, self-hosted vLLM recipe) by @pkooij in #3896
• Fix readme when using annotations by @NikodemBartnik in #4029
• refactor(evo1): use transformers flash attention probe by @imstevenpmwork in #4013
• refactor(hub): load safetensors directly on target device by @imstevenpmwork in #4012
• Aggregate policy sub-losses through MetricsTracker by @Maximellerbach in #4024
• Fix annotation example by @NikodemBartnik in #4041
• Move annotation dependencies to module scope by @pkooij in #4040
• fix(safetensors): expand bare "cuda" to current device for safetensors loads by @Maximellerbach in #4042
• refactor(policies): resolve policy components by convention by @imstevenpmwork in #4015
• refactor(processors): share policy pipeline builders by @imstevenpmwork in #4016
• refactor(policies): extract shared model components by @imstevenpmwork in #4054
• refactor(wall-x): subclass native Transformers Qwen2.5-VL instead of vendoring it by @imstevenpmwork in #4035
• refactor(pi05): use shared VLA components by @imstevenpmwork in #4063
• refactor(pi0): use shared VLA components by @imstevenpmwork in #4062
• refactor(eo1): reuse shared VLA components by @imstevenpmwork in #4061
• refactor(pi0_fast): reuse shared VLA components by @imstevenpmwork in #4055
• refactor(xvla): subclass native Transformers Florence2 instead of vendoring it by @imstevenpmwork in #4089
• chore(ci): upgrade claude workflow by @imstevenpmwork in #4096
• chore(docs): update security policy to adopt HF standards by @imstevenpmwork in #4098
• refactor(smolvla): reuse shared VLA components by @imstevenpmwork in #4064
• feat(annotate): run lerobot-annotate on HF Jobs via --job.target by @pkooij in #4095
• refactor(pi05): remove unused variables in embed_suffix method by @dudududukim in #3263
• Feature omx docs by @Eunsung-kespion in #3421
• docs(readme): fix missing periods at end of sentences by @ykdojo in #3473
• fix(scripts): register third-party plugins in lerobot_setup_motors by @imstevenpmwork in #4123
• fix(gamepad): Gamepad on macos often does not need fallback by @imstevenpmwork in #4125
• feat(diffusion): add gradient checkpointing for memory optimization by @imstevenpmwork in #4127
• fix(openarms): stop set_zero_position()ing on connect by @nepyope in #4058
• fix(robots): add retries while recording motor ranges by @imstevenpmwork in #4126
• fix(envs): set LiberoEnvConfig.fps default to 20 to match robosuite by @imstevenpmwork in #4124
• feat(dataset): add slice support to LeRobotDataset.getitem by @imstevenpmwork in #4129
• chore(agents): adding additional infos to AGENTS.md and bring-your-own-policies.mdx by @Maximellerbach in #3904
• feat(robot): Make SO follower P coefficient configurable by @imstevenpmwork in #4142
• chore(dependencies): update uv.lock by @imstevenpmwork in #3963
• chore(deps): bump diffusers to >=0.38.0,<0.40.0 by @imstevenpmwork in #4145
• (depth image processing): excluding depth frames from the RGB to BGR image processing by @CarolinePascal in #4135
• feat(dataset): accept token argument for private HF Hub datasets by @imstevenpmwork in #4136
• fix conversion error when tag is not exists. by @k1000dai in #3656
• chore: enable Dependabot weekly GitHub Actions bumps by @hf-dependantbot-rollout[bot] in #3677
• fix(datasets): clear video frame staging on episode reset by @MihaiAnca13 in #3683
• fix(scripts): restore policy training mode after eval_policy() in lerobot-eval by @imstevenpmwork in #4162
• fix(record): connect teleoperator before robot to avoid watchdog jump by @imstevenpmwork in #4166
• fix(processor): avoid Hub fallback for missing local processor state by @WOLIKIMCHENG in #3715
• fix(rollout): account for policy revision loading by @imstevenpmwork in #4161
• fix(envs): make MetaWorld evaluation reproducible with seeded object positions by @landegt in #3727
• (fix) Add LeKiwi support to lerobot-replay CLI by @zuoxingdong in #3739
• feat(config): add multiprocessing option to DataLoader context and sets spawn as default by @imstevenpmwork in #4139
• fix(processors): wrong feature key dropped in delta-action transform_features by @imstevenpmwork in #4165
• feat(rl): add multiprocessing option to training pipeline and sets spawn as default + guard by @imstevenpmwork in #4140
• feat(lekiwi): support LeKiwi in the rollout/eval CLI by @zuoxingdong in #3742
• fix(cameras): release device handle when connect() setup fails by @imstevenpmwork in #4187
• feat(camera): add manual exposure, gain, and white balance options for RealSense cameras by @imstevenpmwork in #4130
• feat(record): add --dataset.no_stamp to opt out of repo_id timestamping by @imstevenpmwork in #4190
• feat(motors): add XH540-W150, XC330-T288, XC330-T181 to Dynamixel tables by @charlie8612 in #3815
• refactor(logging): standardize logging with getLogger(name) in scripts by @imstevenpmwork in #4192
• fix(dependencies): protect peft import by @imstevenpmwork in #4188
• fix(policy): honor revisions when loading PEFT checkpoints by @s1lent4gnt in #4189
• Handle resuming empty local datasets by @AlexandreEDMOND in #3859
• chore(mypy): cover annotations and transforms by @AlexandreEDMOND in #3860
• fix(utils): handle missing/unresponsive TTS on Linux by @imstevenpmwork in #4199
• fix(env): eval env lifecycle by @imstevenpmwork in #4194
• fix peft factory test mocking by @s1lent4gnt in #4201
• fix(eval): prevent eval_policy crash when start_seed is None and num_envs>1 by @imstevenpmwork in #4203
• fix(datasets): bound memory of augment_dataset_quantile_stats by sampling frames by @pepa99 in #3749
• perf(docker): split dependency install from source copy for CI layer caching by @imstevenpmwork in #4208
• fix(robots): retry SO follower/leader bus reads on transient Feetech errors by @imstevenpmwork in #4207
• docs: fix dataset split fraction example by @saime428 in #3936
• docs: fix wording in backward compatibility guide by @Old-Ding in #3938
• docs: fix wording in guides by @Old-Ding in #3939
• chore(datasets): add typing to aggregate helpers by @imstevenpmwork in #4211
• fix(evo1): re-pad normalizer stats when loading from checkpoint by @nepyope in #3945
• fix(robot): type FK-to-EE action features as ACTION not STATE by @imstevenpmwork in #4213
• ...

Release Blog

https://huggingface.co/blog/lerobot-release-v060

⚠️ Breaking changes (v0.5.1 -> v0.6.0)

v0.6.0 cleans house, and a few changes need your attention when upgrading:

• pip install lerobot no longer includes dataset or training dependencies; add the extra you need (e.g. lerobot[training]).
• Some import paths have changed. Please use the canonical public entry points rather than internal submodule paths.
• GR00T N1.5 is replaced by N1.7 (pin lerobot==0.5.1 if you need N1.5).
• The minimum PyTorch version is now 2.7.
• eval_freq was renamed to env_eval_freq in the train config.
• The RL stack was rebuilt: the sac policy type is now gaussian_actor under the new modular RL API.
• Legacy per-frame subtask_index annotations are superseded by the new language columns.
• --dataset.vcodec was renamed to --dataset.rgb_encoder.vcodec, such that RGB and depth cameras video codecs may be set separately.

What's Changed

• chore(dependencies): Bump lerobot to 0.5.2 by @imstevenpmwork in #3307
• refactor(envs): move benchmark dispatch into EnvConfig subclasses by @pkooij in #3272
• feat(envs): lazy env init + AsyncVectorEnv as default for n_envs > 1 by @pkooij in #3274
• feat(ci): add agent assitance workflow by @imstevenpmwork in #3332
• chore(security): update claude.yml by @hf-security-analysis[bot] in #3333
• fix(ci): pin claude-code-action to v1.0.88 by @imstevenpmwork in #3336
• chore(ci): proper claude args workflow by @imstevenpmwork in #3338
• fix(dataset): adding metadata loading when reading from a dataset after writing by @CarolinePascal in #3305
• chore(dependencies): update uv.lock by @imstevenpmwork in #3308
• chore(dependencies): update uv.lock by @imstevenpmwork in #3361
• refactor(deps): minimal default install with facade pattern and dependency isolation by @imstevenpmwork in #3362
• fix(rl): ensure queue and process cleanup on abnormal exit by @jashshah999 in #3063
• feat(ci): benchmark smoke tests with isolated Docker images (LIBERO + MetaWorld) by @pkooij in #3319
• fix(feetech): motor position readings overflow by @Maximellerbach in #3373
• Update reward classifier training config by @EPronovost in #3147
• fix(test): add missing device placement in multi-task DiT tests by @s1lent4gnt in #3349
• fix(groot): compatibility fixes for gr00t in v0.5 by @mtiezzi in #3182
• Add thread synchronization to ReplayBuffer to prevent race condition between add() and sample() by @s1lent4gnt in #3372
• fix(rl): preserve teleop intervention flag in step_env_and_process_transition by @remete618 in #3273
• fix: catch KeyboardInterrupt in safe_stop_image_writer to prevent corrupted frames by @s1lent4gnt in #3381
• refactor(imports): enforce guard pattern by @imstevenpmwork in #3382
• chore(docs): update contributing by @imstevenpmwork in #3387
• chore: bump doc-builder SHA for PR upload workflow by @rtrompier in #3386
• chore(docs): adding a jupyter notebook that gives you ready-to-paste commands by @Maximellerbach in #3395
• feat(ci): send slack notification if latest dependency test is broken by @imstevenpmwork in #3398
• fix(ci): not fail when skipped by @imstevenpmwork in #3399
• fix(record): pass rename_map to make_policy in lerobot-record by @wadeKeith in #3240
• Improve loading performance in _absolute_to_relative_idx when remapping indices by @hello3x3 in #3279
• chore(dependencies): update uv.lock by @imstevenpmwork in #3365
• feat(dataset): 2x faster dataloader via parallel decode, uint8 transport, and persistent workers by @imstevenpmwork in #3406
• fix(dataset): cleanup_interrupted_episode wipes image temp dirs by @whats2000 in #3405
• episods→episodes by @k1000dai in #3410
• fix(policies): replace deprecated torch.cuda.amp.autocast with torch.… by @Defalt-Meh in #3167
• ci: skip Docker Hub login step on fork PRs by @pkooij in #3417
• fix(envs): preserve AsyncVectorEnv metadata/unwrapped in lazy eval envs by @HaomingSong in #3416
• feat(envs): add RoboCasa365 benchmark integration by @pkooij in #3375
• feat(envs): add RoboTwin 2.0 benchmark by @pkooij in #3315
• feat(envs): add RoboCerebra long-horizon manipulation benchmark by @pkooij in #3314
• feat(envs): add RoboMME benchmark by @pkooij in #3311
• feat(envs): add LIBERO-plus robustness benchmark by @pkooij in #3313
• feat(sim): VLABench benchmark integration by @pkooij in #3396
• fix(robotwin): pin compatible curobo in benchmark image by @pkooij in #3427
• docs(agents): add AGENT_GUIDE.md for user facing agent by @pkooij in #3430
• chore(dependencies): update uv.lock by @imstevenpmwork in #3408
• fix(smolvla): correct loss normalization for padded actions by @whats2000 in #3434
• docs(xvla): fix broken local file links in documentation by @kuafou in #3291
• fix(policy): loss normalization for padded actions in ACT, Diffusion, and MultiTaskDiT by @imstevenpmwork in #3442
• fix(sarm): handle BaseModelOutputWithPooling from transformers 5.x in… by @masato-ka in #3419
• fix(policies): remove @DataClass from GR00TN15Config to fix import crash on transformers ≥ 5.5 by @ChuyaoShen in #3414
• fix(imports): realsense import name is platform dependent by @imstevenpmwork in #3451
• fix(dataset): no default overwrite in lerobot tool recompute stats by @imstevenpmwork in #3452
• fix(pi): avoid peak RAM in PiGemma construction by freeing replaced submodules by @imstevenpmwork in #3454
• chore(dependencies): update uv.lock by @imstevenpmwork in #3437
• feat(rollout): decouple policy deployment from data recording with new lerobot-rollout CLI by @imstevenpmwork in #3413
• chore(ci): bump docs workflows by @imstevenpmwork in #3476
• Reward models refactor by @s1lent4gnt in #3142
• refactor(datasets): replace untyped dict with typed DatasetInfo dataclass by @Maximellerbach in #3472
• fix(rollout): propagate policy-specific CLI config paramaters by @imstevenpmwork in #3483
• Migrate legacy RA-BC fields in train config loading by @s1lent4gnt in #3480
• Expose sarm package API and ship reward model card template by @s1lent4gnt in #3477
• fix(pi0&pi05): keep training sampling outside compiled forwards by @HaomingSong in #3487
• fix(peft): fix LoRA resume from Hub (PosixPath + double wrap) by @jashshah999 in #3485
• chore(ci): upgrade docker internal by @imstevenpmwork in #3505
• chore(ci): increase time stale by @imstevenpmwork in #3507
• chore(dep): bump transformers to 5.4.0 by @imstevenpmwork in #3374
• feat(dagger): adding smooth handover by @Maximellerbach in #3506
• fix(robots): openarm features compatibility with openarm_mini by @imstevenpmwork in #3524
• feat(policies): add EO-1 model by @HaomingSong in #3403
• fix(ci): bump robotwin benchmark image to CUDA 12.6 by @pkooij in #3525
• feat(policy): use pretrained vision encoder weights by default for diffusion and vqbet by @villekuosmanen in #3202
• fix(c...

What's Changed

• chore(dependencies): Bump lerobot to 0.5.1 by @imstevenpmwork in #3118
• fix(unitree_g1): correct SDK detection and update installation docs by @nepyope in #3115
• docs: fix HF_USER to ignore bash ANSI codes by @j3soon in #3119
• chore(dependencies): Increase opencv-python-headless upper bound by @traversaro in #3120
• Fix action padding key at SmolVLA by @ymd-h in #1717
• Update Dockerfile.user by @heuzef in #1633
• fix(policies): crop losses based on the action dof by @imstevenpmwork in #3133
• Add option to disable tags on WandB by @brunomachado37 in #1339
• fix(links): replacing relative links with absolute links in the contribution guide by @CarolinePascal in #3141
• fix(script help): fixing lerobot-train --help by removing % in the docstrings by @CarolinePascal in #3161
• chore(dependecies): untangle dependecies across internal modules by @imstevenpmwork in #3149
• refactor(datasets): module cleanup by @imstevenpmwork in #3169
• chore(dataset): basic house-keeping by @imstevenpmwork in #3170
• refactor(dataset): modular files by @imstevenpmwork in #3171
• Fix/earth rover dataset features by @s1lent4gnt in #3088
• fix(vqbet): use in-place fill_ to avoid overwriting DDP GPU buffers with CPU tensors by @Altman-conquer in #3128
• fix(links): update GitHub issues link in async docs by @0xPraedico in #3186
• chore(docs): add rename map and empty cam guide by @jadechoghari in #3065
• refactor(dataset): split LeRobotDataset into DatasetReader & DatasetWriter (+ API cleanup) by @imstevenpmwork in #3180
• fix(logging): suppress noisy httpx INFO logs by @reeceomahoney in #3173
• fix(deps): breaking change from transformers 5.4.0 by @Maximellerbach in #3231
• chore(docs): add more guidance to bring your own policies tutorial by @Maximellerbach in #3230
• fix(dataset): use revision-safe Hub cache for downloaded datasets by @imstevenpmwork in #3233
• Feature/add multitask diffusion transformer policy implementation by @brysonjones in #2545
• fix(datasets): remove unreachable timestamp branch in add_frame by @SevenFo in #3163
• feat(scripts): add transformers version by @imstevenpmwork in #3248
• Revert "fix(deps): breaking change from transformers 5.4.0" by @imstevenpmwork in #3249
• Fix SO-101 assembly instruction order to match videos by @jaikr-dev in #3242
• feat(policies): add relative action support for pi0, pi0.5, and pi0_fast by @pkooij in #2970
• Add create reward visualization by @pkooij in #3155
• docs(toctree): add Benchmarks section for LIBERO and Meta-World by @pkooij in #3268
• feat(dagger): Add HIL/Dagger/HG-Dagger/RaC style data collection by @pkooij in #2833
• docs(benchmarks): add benchmark integration guide and standardize benchmark docs by @pkooij in #3270
• docs(policy): fix gr00t license docs by @imstevenpmwork in #3284
• chore(dependencies): bump minimum torch from 2.2.1 to 2.7 by @CarolinePascal in #3156
• chore(docs): no policy readme in src code by @imstevenpmwork in #3286
• chore(policy): multi dit docs by @imstevenpmwork in #3285
• feat(ci): add uv.lock by @imstevenpmwork in #3292
• docs(ci): add readme for dockerfile by @imstevenpmwork in #3295
• fix(ci): latest deps tests permissions by @imstevenpmwork in #3296
• chore(dependencies): update uv.lock by @imstevenpmwork in #3302
• chore(docs): new badge for readme by @imstevenpmwork in #3303
• fix(scripts): Update action values when running robot without interpolation by @anthonyhchan in #3281
• feat(dataset): registering torchvision transforms by @fracapuano in #3153
• 🔒 Pin GitHub Actions to commit SHAs by @paulinebm in #3265

New Contributors

• @j3soon made their first contribution in #3119
• @traversaro made their first contribution in #3120
• @ymd-h made their first contribution in #1717
• @heuzef made their first contribution in #1633
• @brunomachado37 made their first contribution in #1339
• @Altman-conquer made their first contribution in #3128
• @0xPraedico made their first contribution in #3186
• @SevenFo made their first contribution in #3163
• @anthonyhchan made their first contribution in #3281

Full Changelog: v0.5.0...v0.5.1

Release Blog

https://huggingface.co/blog/lerobot-release-v050

What's Changed

• chore(dependencies): Bump lerobot to 0.4.5 by @imstevenpmwork in #3051
• chore: add AI policy by @imstevenpmwork in #3055
• Improve policy_device documentation for async.mdx by @btelles in #3060
• fix(frame_index): making rerun's "frame_index" timeline compatible with behaviour1k datasets by @CarolinePascal in #3068
• fix(dataset edit tools): clarifying root argument usage + adding related features by @CarolinePascal in #3049
• chore(root): fixing root semantics in convert_dataset script by @CarolinePascal in #3073
• chore(docs): updating deprecated huggingface-cli to hf by @Maximellerbach in #3071
• Fixing metadata indexing when writing new Parquet file by @skiingpacman in #2941
• chore(docstrings): updating v2.1-v3.0 conversion script docstrings to match the new task label by @CarolinePascal in #3077
• chore(dependencies): bump transformers v5 by @imstevenpmwork in #2964
• fix(ci): log in HF for gated repo in nightly workflows by @imstevenpmwork in #3089
• fix(num_frames): fixing redundant frames count in conversion script by @CarolinePascal in #3091
• Feat/slurm compute rabc script by @pkooij in #3041
• fix(ci): test only multi-gpu tests in multi-gpu runner by @imstevenpmwork in #3092
• feat(dependencies): require Python 3.12+ as minimum version by @imstevenpmwork in #3023
• chore(dependencies): sync intelrealsense + added notes by @imstevenpmwork in #3094
• chore(docs): remove pi installation note by @imstevenpmwork in #3095
• fix(ci): skip HF log in (and tests) in forks and community PRs by @imstevenpmwork in #3097
• fix(async_inference): restore robot module imports in robot_client.py by @aShunSasaki in #3081
• feat(robots): Unitree G1 WBC implementation by @nepyope in #2876
• feat(train): add cudnn_deterministic option for reproducible training by @imgeorgiev in #3102
• chore(docs): add uv installation instructions by @imstevenpmwork in #3105
• chore(readme): update citation with ICLR26 paper by @imstevenpmwork in #3107
• test(cameras): skip flaky async_read test by @imstevenpmwork in #3106
• chore(deps): update requirements file by @imstevenpmwork in #3114
• chore(dependencies): Bump lerobot to 0.5.0 by @imstevenpmwork in #3117

New Contributors

• @btelles made their first contribution in #3060
• @Maximellerbach made their first contribution in #3071
• @skiingpacman made their first contribution in #2941
• @aShunSasaki made their first contribution in #3081
• @imgeorgiev made their first contribution in #3102

Full Changelog: v0.4.4...v0.5.0

What's Changed

• chore(dependencies): Bump lerobot to 0.4.4 by @imstevenpmwork in #2840
• feat(ci): trigger manually documentation release version by @imstevenpmwork in #2841
• fix(robots): update gripper configuration and calibration settings for OMX by @Woojin-Crive in #2815
• feat(envs): Add task ids to libero env cfg by @reeceomahoney in #2842
• feat(motors): add damiao motors & can bus by @imstevenpmwork in #2788
• Add robot images to docs by @pkooij in #2862
• docs: update libero by @reeceomahoney in #2857
• Fix(aggregate.py) Aggregation of datasets when sub-datasets are already a result of a previous merge by @michel-aractingi in #2861
• feat(robots): add OpenArm robot & teleoperator by @imstevenpmwork in #2795
• add g1 teleoperation by @nepyope in #2791
• feat(robots): add bi manual openarm follower and leader by @imstevenpmwork in #2835
• refactor(cameras): cameras API extension by @imstevenpmwork in #2808
• fix(sac): make temperature a property to fix checkpoint resume bug by @michel-aractingi in #2877
• Feat(dataset_tools.py) Add modify tasks tool by @michel-aractingi in #2875
• docs(cameras): revising and improving docs on cameras by @CarolinePascal in #2878
• fix(find zmq): adding a clearer not implemented warning for the ZMQ find_cameras method by @CarolinePascal in #2879
• feat(dataset): add subtask support by @jadechoghari in #2860
• docs: update document in response to Simplify configs PR by @HiroIshida in #1596
• Add missing RL config options: add_ee_pose_to_observation and gripper_penalty_in_reward by @michel-aractingi in #2873
• Limit W&B tag length to comply with 64‑character limit by @iory in #995
• Small docstring fix in diffusion configuration by @jwang078 in #2847
• feat(datasets): improve image transform support by @reeceomahoney in #2885
• fix(policies): wandb warning during smolvla training by @reeceomahoney in #2903
• fix(scripts): correct can import check by @imstevenpmwork in #2937
• fix(pipeline): use FeatureType for STATE features in Libero processor by @stepanfeduniak in #2888
• Change LIBERO init_state_id when reset. by @AoqunJin in #2899
• docs: clarify installation steps are sequential, not optional by @jaikr-dev in #2925
• [Bug Fix] fix(ci): prevent runner group error on fork pushes by @whats2000 in #2911
• chore(linter): ensure motors module passes MyPy type checks by @imstevenpmwork in #2939
• chore: use is_connected decorators by @imstevenpmwork in #2948
• fix(motors): use decorator to fix precommit by @imstevenpmwork in #2951
• feat(cameras): make backend configurable to the CLI by @imstevenpmwork in #2945
• Fixing wrong argparse examples in the comments (using 'True' not 'true') by @taken-yjyoon in #1040
• fix(dataset tools draccus): fixing draccus parsing for dataset edit operation type specification by @CarolinePascal in #2949
• Always use degrees by @pkooij in #2968
• feat(datasets): Add info operation to lerobot-edit-dataset command by @masato-ka in #2917
• (Chore) Bump upper bound for torch version by @sovrasov in #2897
• fix(scripts): serve grpc for a web viewer by @sotanakamura in #2881
• fix(docs): update env variable LEROBOT_HOME to HF_LEROBOT_HOME by @tc-huang in #2973
• fix(utils): more conservative sleep_margin default value in precise_sleep by @imstevenpmwork in #2985
• chore(scripts): warn if we don't respect the target FPS by @imstevenpmwork in #2986
• fix(filtering): fixing episodes filtering in load_nested_dataset to always use .from_parquet() by @CarolinePascal in #2982
• feat(robots): use read_latest() camera by @imstevenpmwork in #2987
• chore: remove usernames + use entrypoints in docs, comments & sample commands by @imstevenpmwork in #2988
• Fix gym-hil integration with the new LeRobot pipeline. by @s1lent4gnt in #2482
• chore(deps): bump ceil datasets by @imstevenpmwork in #2946
• feat(dataset): add streaming video encoding + HW encoder support by @imstevenpmwork in #2974
• Add metadata_buffer_size to dataset creation by @yuecideng in #2998
• feat(motors): add RobStride CAN implementation by @imstevenpmwork in #2821
• chore(docs): update the document for Phone teleop to clarify how to use the examples by @ynaka81 in #2991
• fix/docs: small comment change by @gui-miotto in #2990
• docs: fix HF_USER export command to correctly parse username by @WilbertYuan in #2932
• docs: add WSL evdev installation note by @DominikPalo in #2855
• feat: Enable torch.compile for DiffusionPolicy inference by @jashshah999 in #2486
• fix(ci): permissions triton by @imstevenpmwork in #3011
• feat(scripts): Integrate tqdm for training progress visualization by @imstevenpmwork in #3010
• fix(dataset): Fix reindexing bug for videos on splits by @mkiefel in #2548
• Fix metaworld_config.json missing from pip installs and AttributeError crash by @jashshah999 in #3017
• fix(video): replace assertions with proper exceptions in video frame decoding by @jashshah999 in #3016
• Fix SARM dense_only mode: always load episodes_df for target computation by @s1lent4gnt in #3021
• feat: add visualization badge to card template and update dataset card creation with repo_id by @mishig25 in #3005
• fix(policies): support dp train when n_obs_steps=1 by @Hukongtao in #2430
• Feat(teleoperators): add OpenArm Mini teleoperator by @s1lent4gnt in #3022
• Fix: remove device_map from SmolVLA model loading by @s1lent4gnt in #3029
• feat(async_inference) Enable plugins with async inference by @IamPhytan in #2425
• fix(datasets): only warn in MultiLeRobotDataset when features differ by @thatmich in #3019
• chore(scripts): no spam log when no action by @imstevenpmwork in #3042
• fix(scripts): register OpenArm Mini in all lerobot CLI scripts by @s1lent4gnt in #3028
• feat(policies): add Smolvla torch compile support by @jadechoghari in #3043
• Fix(diffusion): enforce no-crop behavior when crop_ratio=1.0 by @s1lent4gnt in #3046
• fix(logging): avoid double-counting samples across processes by @pkooij in #3045
• chore(docstrings): fixing deprecated root argument description in LeRobotDataset class by @CarolinePascal in #3035

New Contributors

• @reeceomahoney made their first contribution in #2842
• @iory made their first contribution in #995
• @jwang078 made their first contribution in #2847
• @stepanfeduniak made their first contribution in #2888
• @AoqunJin made their first contribution in #2899
• @jaikr-dev made their first contribution in #2925
• @whats2000 made their first contribution in #2911
• @taken-yjyoon made their first contribution in #1040
• @yuecideng made their first contribution in https:...

What's Changed

• chore(dependencies): Bump lerobot to 0.4.3 by @imstevenpmwork in #2540
• fix(scripts): remove the data sampler because data access to the subset of a dataset has been fixed by @sotanakamura in #2521
• Feat/add unitree g1 robot by @nepyope in #2530
• fix(ci): move hub artifacts to /mnt to avoid runners' No space left on device by @imstevenpmwork in #2564
• fix pi05 forward compile by @michel-aractingi in #2551
• Make transport module Mypy Compliant [issue#1731] by @sezan92 in #2433
• feat(policies): Allow users to register 3rd party policies - pip install lerobot_policy_mypolicy by @danielsanjosepro in #2308
• doc: fix wrong package name in installation doc by @yanghua in #2513
• Earth Rover Mini Plus integration by @s1lent4gnt in #2544
• Revert "Earth Rover Mini Plus integration" by @imstevenpmwork in #2574
• feat(policies): Add X-VLA by @jadechoghari in #2405
• feat(robots): Earth Rover Mini Plus integration by @imstevenpmwork in #2575
• fix(docs): remove duplicated package in install instructions by @imstevenpmwork in #2573
• Fix link to lerobot-train script in documentation by @cagataycali in #2466
• chore(docs): reorganize assembly instructions setup before assembly by @ozten in #2333
• Fix argument typo in async inference documentation by @kevinjosethomas in #2361
• fix(scripts): missing so101 import by @imstevenpmwork in #2577
• Add a documentation page with a brief intro to hw backends by @sovrasov in #2385
• Fix video indexing bugs in dataset aggregation with multi video datatasets by @michel-aractingi in #2550
• chore(ci): skip workflows if not lerobot repository by @imstevenpmwork in #2601
• Fix torch.compile in diffusion sampling loops at inference time by @michel-aractingi in #2600
• feat(dataset): add tool to convert images to video datasets by @jadechoghari in #2560
• Docs: Update X-VLA training strategies/commands by @jadechoghari in #2611
• enable variable image sizes to pi0/pi0.5 by @michel-aractingi in #2609
• docs: add auto in docs for xvla by @jadechoghari in #2645
• feat(robot): Add support for OMX robot by @Woojin-Crive in #2614
• Feat/add mj env by @nepyope in #2613
• docs: update IL robots API example and add OpenCV workaround by @cagataycali in #2648
• fix(scripts): finalize the dataset after recording by @sotanakamura in #2496
• feat(dataset): expose tolerance_s argument to training config by @imstevenpmwork in #2653
• fix(policies): add device back to smolvlm expert by @imstevenpmwork in #2662
• feat(scripts): add more info to lerobot-info by @imstevenpmwork in #2663
• docs(dataset): add visualization section by @imstevenpmwork in #2664
• chore(ci): update PR template by @imstevenpmwork in #2665
• feat(ci): simple automatic labelling by @imstevenpmwork in #2667
• chore(ci): update issue template by @imstevenpmwork in #2666
• fix(ci): better heuristic + issue type template fix by @imstevenpmwork in #2672
• fix(ci): close bracket pattern by @imstevenpmwork in #2674
• chore(ci): minor improvements auto labeling by @imstevenpmwork in #2675
• chore(ci): minor improvement bug-report template & pr auto label by @imstevenpmwork in #2676
• Add sarm by @pkooij in #2639
• feat(docs): modernize readme by @imstevenpmwork in #2660
• Fixes ReadMe Policies Classification by @fracapuano in #2682
• fix(optim): enable and resolve mypy type errors by @cverrier in #2683
• Add readme by @pkooij in #2698
• feat: support wallx model by @wut19 in #2593
• modify the README file for wallx by @wut19 in #2705
• chore: modernize contributing.md by @imstevenpmwork in #2677
• use syslink for wall-x readme by @pkooij in #2708
• pi fixes for dependencies by @pkooij in #2706
• feat(ci): adds release versioning to docs by @imstevenpmwork in #2709
• chore(ci): make keyword matching more conservative by @imstevenpmwork in #2711
• Upgrade GitHub Actions for Node 24 compatibility by @salmanmkc in #2691
• fix: use importlib.metadata for plugin discovery to support PEP 660 by @aalvan in #2687
• resolve path for RABC correctly by @pkooij in #2710
• chore: adds dynamic README handling and setup script by @imstevenpmwork in #2724
• feat(pi0): add train_expert_only and freeze_vision_encoder flags to pi0 and pi0.5 by @pxe3 in #2727
• feat(envs): Add NVIDIA IsaacLab-Arena Lerobot by @nv-sachdevkartik in #2699
• IsaacLab Arena Integration documentation update by @liorbenhorin in #2749
• chore(readme): update discord invitation link by @imstevenpmwork in #2750
• Add basic PEFT support to train script + record module by @githubnemo in #1411
• Fix invalid syntax by @pkooij in #2752
• fix a bug for kwargs in wallx by @wut19 in #2714
• docs: LightWheel to Lightwheel by @jadechoghari in #2758
• fix: precise_sleep is never called with negative value by @imstevenpmwork in #2757
• refactor(ci): Docker Hub image env by @paulinebm in #2755
• chore(dependencies): bound new dependecies by @imstevenpmwork in #2759
• add holosoma locomotion by @nepyope in #2669
• feat(visualization): allow remote viewer + compress rerun images by @imstevenpmwork in #2756
• Revert "feat(visualization): allow remote viewer + compress rerun images" by @imstevenpmwork in #2766
• feat(visualization): allow remote viewer + compress rerun images by @imstevenpmwork in #2767
• feat(robots): consolidate SO arms implementation by @imstevenpmwork in #2763
• chore(docs): update code block syntax to specify python for clarity by @imstevenpmwork in #2770
• feat(datasets): expose video codec option for dataset recording by @leot13 in #2771
• skip peft cmd test in cli by @pkooij in #2776
• feat(policies): add autoregressive VLAs with tokenization PiFast by @jadechoghari in #2734
• Fix: Respect policy.device=cpu config in training by @michel-aractingi in #2778
• docs: improve assets by @jadechoghari in #2777
• feat(robots): consolidates bi SO setups by @imstevenpmwork in #2780
• Feat/g1 improvements record sim by @nepyope in #2765
• fix(robots): add reachy2 fixes by @imstevenpmwork in #2783
• Move cfg.validate() earlier to fix NoneType error with --policy.path by @samet-rob in #2782
• fix: train tokenizer CLI entry point by @jadechoghari in #2784
• chore: use alias & constants by @imstevenpmwork in #2785
• fix(teleop): add is_connected check to get_action by @imstevenpmwork in #2801
• chore(ci): merge annoying section in PR template by @imstevenpmwork in https://git...

What's Changed

• chore(dependencies): Bump lerobot to 0.4.2 by @imstevenpmwork in #2423
• Fixes in port droid scripts by @michel-aractingi in #2455
• Fix episode filtering bug when requesting a subset of the episodes in a dataset by @michel-aractingi in #2456
• [RTC] Real Time Chunking for Pi0, Smolvla, Pi0.5 by @helper2424 in #1698
• feat(envs): add envs pre-post processor by @jadechoghari in #2474
• Add absolute_to_reative_idx for remapping indices to fix delta_timestamp bug by @michel-aractingi in #2490
• docs: Add LeIsaac x LeRobot Envhub tutorial by @jadechoghari in #2498
• feat(dataset): speed-up encoding time by @imstevenpmwork in #2514
• feat(dataset): dynamic compress_level depending on the type of dataset (video or image) by @imstevenpmwork in #2517
• fix(examples): wrap all of them into a main function by @imstevenpmwork in #2524
• feat(parallel encoding): making parallel encoding the default choice over all platforms by @CarolinePascal in #2525
• fix(benchmark) : fixing video benchmark by @CarolinePascal in #2094
• feat(utils): precise_sleep() less CPU hungry without sacrificing accuracy by @imstevenpmwork in #2526
• fix(scripts): better prints teleop by @imstevenpmwork in #2538

Full Changelog: v0.4.1...v0.4.2

What's Changed

• chore(dependecies): Bump lerobot to 0.4.1 by @imstevenpmwork in #2299
• docs(readme): update installation instructions for 0.4.0 by @imstevenpmwork in #2310
• docs(install): improve groot and libero installation instructions by @imstevenpmwork in #2314
• fix(video_key typo): fixing video_key typo in update_video_info by @CarolinePascal in #2323
• fix(make_policy): rename mapping edge cases in training by @jadechoghari in #2332
• dep(upgrade): add libero as a pypi package by @jadechoghari in #2365
• feat(sim): EnvHub - allow loading envs from the hub by @jadechoghari in #2121
• fix(dataset_tools) Critical bug in modify features by @michel-aractingi in #2342
• fix(dataset): fix data access bottleneck for faster training by @imstevenpmwork in #2408
• chore(installation): remove libero installation patch by @imstevenpmwork in #2416

Full Changelog: v0.4.0...v0.4.1

Release Blog

https://huggingface.co/blog/lerobot-release-v040

What's Changed

• chore: Bump to 0.3.4 by @imstevenpmwork in #1691
• chore(docs): prioritize use of entry points in docs + fix nightly badge by @imstevenpmwork in #1692
• fix(busy_wait): fix busy_wait implementation for Windows platforms by @CarolinePascal in #1695
• fix(ci): inverted names by @imstevenpmwork in #1705
• fix(ci): rename libegl1-mesa in deb13 trixie by @imstevenpmwork in #1735
• fix: Add missing dynamixel encoding table entries by @jackvial in #1534
• fix(dataset): Correctly use episode_data parameter in save_episode by @lxk-221 in #1740
• Removed dupicate lines of code by @mgiac-hexagon in #1709
• Add feetech firmware update docs by @pkooij in #1793
• Lower limits by 50% for current and torque for gripper motor by @pkooij in #1809
• Allow max_relative_target to be a float by @jstnhuang in #1837
• 2 add reachy 2 to updated lerobot by @glannuzel in #1767
• fix (koch): Disable motor torque before applying calibration logic by @Gongsta in #1889
• fix(scripts): revert deletion of rs cam config import introduced by #1767 by @imstevenpmwork in #1876
• Dataset v3 by @michel-aractingi in #1412
• Add Streaming Dataset by @fracapuano in #1613
• Update dataset card by default by @michel-aractingi in #1936
• docs(dataset): add dataset v3 documentation by @jadechoghari in #1956
• feat(processors): use pipelines across the codebase by @AdilZouitine in #1452
• small tiny nit by @fracapuano in #1975
• fix (docs): image link for phone by @pkooij in #1977
• bump gym-hil version to be pipeline compatible by @michel-aractingi in #1983
• Add docs for LeRobot Image transforms by @pkooij in #1972
• fix formatting readme by @pkooij in #1987
• bump datasets to 4.0.0 by @michel-aractingi in #1990
• fix(trainer): overrides device to the target device, for the device processor on the preprocessor by @AdilZouitine in #1993
• feat(sim): Add Libero Env by @jadechoghari in #1984
• feat(ci): add stale GH action bot for stalled issues & PRs by @imstevenpmwork in #1996
• chore(ci): update time for stale issue/pr by @imstevenpmwork in #1997
• Updated Readme removed broken link by @complete-dope in #1989
• Fix: Resolve PermissionError and UnicodeDecodeError in Python scripts by @JivinDotL in #1980
• fix(dataset): cast fps to int instead of float by @imstevenpmwork in #2001
• chore(rl): move rl related code to its directory at top level by @imstevenpmwork in #2002
• docs(rl): fix path by @imstevenpmwork in #2004
• refactor(scripts): update system info script by @imstevenpmwork in #2005
• feat(script): add entry point for dataset viz by @imstevenpmwork in #2006
• feat(script): add entry point for image transform viz by @imstevenpmwork in #2007
• chore: move constants to utils by @imstevenpmwork in #2016
• chore: move errors to utils by @imstevenpmwork in #2017
• chore(scripts): move find_cameras to scripts by @imstevenpmwork in #2018
• feat(script): add entry point for find joints limits by @imstevenpmwork in #2010
• chore(scripts): move find_port to scripts by @imstevenpmwork in #2019
• chore(scripts): move teleop to scripts by @imstevenpmwork in #2023
• chore(scripts): move record to scripts by @imstevenpmwork in #2022
• chore(scripts): move calibrate to scripts by @imstevenpmwork in #2024
• chore(scripts): move setup_motors to scripts by @imstevenpmwork in #2020
• chore(scripts): move replay to scripts by @imstevenpmwork in #2021
• chore(async): move async related code to its directory at top level by @imstevenpmwork in #2003
• chore(ci): update stale message by @imstevenpmwork in #2027
• fix(docs): update outdated links by @imstevenpmwork in #2026
• chore(utils): move benchmark and buffer to their respective modules by @imstevenpmwork in #2028
• chore(utils): move encoding utils and process to their respective modules by @imstevenpmwork in #2029
• chore(utils): remove unused utils legacy functions + rename init_rerun by @imstevenpmwork in #2031
• chore(utils): move queue utils and wandb_utils to their respective modules by @imstevenpmwork in #2030
• (docs): small change in dataset name by @jadechoghari in #2032
• chore(scripts): rename eval and train scripts by @imstevenpmwork in #2033
• chore: replace hard-coded obs values with constants throughout all the source code by @imstevenpmwork in #2037
• chore(mypy): add mypy configuration and module overrides for gradual type checking by @AdilZouitine in #2052
• chore: replace hard-coded action values with constants throughout all the source code by @imstevenpmwork in #2055
• Bug in conversion from v2.1 script by @michel-aractingi in #2057
• chore: replace hard-coded next values with constants throughout all the source code by @imstevenpmwork in #2056
• chore(utils): remove unused code by @imstevenpmwork in #2059
• Fix(train): add processor overrides with normalizer and unnormalizer stats by @AdilZouitine in #2038
• docs(fix): libero example command by @jadechoghari in #2060
• Save Cropped Dataset to Hub by @fracapuano in #2071
• Fix video concat path by @Tavish9 in #2076
• chore: remove unused code by @imstevenpmwork in #2062
• Improve V3 aggregate implementation by @Tavish9 in #2077
• chore: enable pyugrade ruff lint by @imstevenpmwork in #2084
• chore: enable simplify in ruff lint by @imstevenpmwork in #2085
• feat(normalization): add validation for empty features in NormalizerProcessorStep and UnnormalizerProcessorStep by @AdilZouitine in #2087
• Revert feat(normalization): add validation for empty features in NormalizerProcessorStep and UnnormalizerProcessorStep (#2087) by @AdilZouitine in #2089
• feat(policies): add noise parameter to action prediction methods by @AdilZouitine in #2063
• feat(train): include input and output features in processor overrides for normalization by @AdilZouitine in #2090
• fix(video_path): setting video_path to None during conversion for images datasets by @CarolinePascal in #2095
• feat(mypy): enable type checking for envs module and configure mypy settings in pyproject.toml by @AdilZouitine in #2099
• feat(mypy): configure mypy settings and add module overrides for gradual typing by @AdilZouitine in #2101
• Making Envs module pass MyPy checks by @iakhil in #2048
• Add OpenPi, Pi0 and Pi0.5 by @pkooij in #1910
• (chore) improve v3 message, allow converting local datasets to V3 by @michel-aractingi in #1948
• fix bug in augment_dataset_quantile_stats.py that was not detecting… by @michel-aractingi in https://github.com/hu...