What's Changed

• chore: Bump to 0.4.0 by @imstevenpmwork in #1653
• Fix policy construction by @fracapuano in #1665
• fix(docs): Update links in il_robots.mdx and il_sim.mdx to use absolute URLs by @tc-huang in #1313
• fix(typo): fixing typo in LeRobot authors names by @CarolinePascal in #1673
• docs: update installation instructions by @imstevenpmwork in #1686
• feat(pipeline): universal processor for LeRobot by @AdilZouitine in #1431
• fix(deps): ceil torch pkg versions by @imstevenpmwork in #1689
• chore: Bump to 0.3.3 by @imstevenpmwork in #1690

Full Changelog: v0.3.2...v0.3.3

What's Changed

• Added poetry setup by @aliberts in #2
• Style & Formatting by @aliberts in #3
• Added pusht dataset auto-download by @aliberts in #4
• Add run on cpu-only compatibility by @aliberts in #5
• Make diffusion work by @Cadene in #6
• Add CI, Remove old install files by @aliberts in #8
• Add AbstractReplayBuffer by @Cadene in #7
• Add Aloha replay buffer + 4 sim datasets by @Cadene in #9
• Fix env tests by @Cadene in #10
• Add pusht test artifact by @aliberts in #12
• Add end-to-end pusht eval test by @aliberts in #15
• Integrate diffusion policy by @aliberts in #16
• Fix bugs with normalization by @Cadene in #17
• Configure wandb entity outside config by @aliberts in #19
• Add license by @aliberts in #20
• Add dynamic build versioning by @aliberts in #18
• Update README.md by @kashif in #23
• Add Aloha env and ACT policy by @Cadene in #11
• CI env fix by @aliberts in #25
• Download datasets from hugging face by @Cadene in #28
• Use Path type instead of str for data_dir by @Cadene in #30
• Fix wandb artifact name and add disable option by @alexander-soare in #31
• Switch between train and eval modes by @alexander-soare in #35
• Add replay_buffer directory and dataset versioning by @Cadene in #36
• Improve README/LICENSE by @Cadene in #37
• Incorporate SerialEnv and introduct multistep policy logic by @alexander-soare in #26
• [WIP] Fix remaining issues with training PushT Image for diffusion policy by @alexander-soare in #22
• Update stats for Pusht and Aloha datasets (v1.0 -> v1.1) by @Cadene in #39
• Enable mps backend for Apple silicon devices by @aliberts in #40
• Fixes issues with PushT diffusion by @alexander-soare in #41
• Run model from HuggingFace hub by @alexander-soare in #44
• Reproduce original diffusion policy pusht image eval by @alexander-soare in #45
• Fix bug with stats.pth + Move from cadene to lerobot + Update datasets to v1.2 by @Cadene in #46
• Fix normalization of last frame and data type in visualization by @alexander-soare in #48
• Port simxarm, upgrade gym to gymnasium by @aliberts in #47
• Improve README by @Cadene in #49
• Update diffusion model by @alexander-soare in #58
• Improve readme by @aliberts in #57
• Add code coverage, more end-to-end tests by @aliberts in #56
• Add examples 2 and 3 by @alexander-soare in #53
• fix stats computation by @alexander-soare in #66
• Remove envs by @aliberts in #69
• Refactor act by @alexander-soare in #68
• Remove torchrl by @Cadene in #64
• Use policy configs instead of passing arguments directly to policy classes by @alexander-soare in #72
• Partial refactor of Diffusion Policy by @alexander-soare in #71
• Refactor pyproject by @aliberts in #76
• Setup contributions by @aliberts in #74
• Use Hugging Face datasets.Dataset by @Cadene in #73
• Fix issue with saving freshly computed stats by @alexander-soare in #77
• Unify policy API by @alexander-soare in #80
• Improve dataset examples by @Cadene in #82
• Tests cleaning & simplification by @aliberts in #81
• Fix tolerance for delta_timestamps by @alexander-soare in #84
• Hotfix test_examples.py by @aliberts in #87
• Quality of life patches for eval.py by @alexander-soare in #86
• Loads episode_data_index and stats during dataset init by @Cadene in #85
• Fix online training by @Cadene in #94
• Move normalization to policy for act and diffusion by @Cadene in #90
• Refactor datasets into LeRobotDataset by @Cadene in #91
• CI nightlies cpu/gpu & cleanup by @aliberts in #75
• Make sure targets are normalized too by @alexander-soare in #106
• More CI cleanup, add style workflow by @aliberts in #107
• Add test-docker-build workflow by @aliberts in #109
• Update readme & remove example 1 by @aliberts in #108
• Add torchaudio by @aliberts in #110
• Add UMI-gripper dataset by @AdilZouitine in #83
• Refactor the download and publication of the datasets and convert it into CLI script by @AdilZouitine in #95
• Remove warnings by @aliberts in #111
• Fix policy defaults by @alexander-soare in #113
• Update UmiProcessor default fps to 10 by @AdilZouitine in #116
• Remove update method from the policy by @qgallouedec in #99
• Remove mask check key in XarmProcessor by @AdilZouitine in #117
• Speed up CI, add more checks by @aliberts in #115
• Refactor push_dataset_to_hub by @Cadene in #118
• Tidy up yaml configs by @alexander-soare in #121
• Nest ACT model in ACT Policy by @alexander-soare in #122
• Follow transformers single file naming conventions by @alexander-soare in #124
• Use PytorchModelHubMixin to save models as safetensors by @alexander-soare in #125
• Refactor TD-MPC by @alexander-soare in #103
• Fix missing policy.to(device) in policy factory by @alexander-soare in #126
• Add video decoding to LeRobotDataset by @Cadene in #92
• Publish environments by @aliberts in #120
• Refactor eval.py by @alexander-soare in #127
• Remove batch length restrictions in select_action by @aliberts in #123
• Add dataset visualization with rerun.io by @Cadene in #131
• Add regression tests by @aliberts in #119
• Fix loops in eval.py by @alexander-soare in #128
• Remove EMA model from Diffusion Policy by @alexander-soare in #134
• Rename Diffusion Policy model repository by @alexander-soare in #136
• Remove additional frame from eval render by @alexander-soare in #137
• Add 25 real world datasets (static + mobile aloha) by @Cadene in #133
• Release cleanup by @aliberts in #132
• Remove loss masking from diffusion policy by @alexander-soare in #135
• Relax version contraints & poetry lock by @aliberts in #139
• Add policy and dataset tags to WandB logging by @alexander-soare in #144
• Override pretrained model config by @alexander-soare in #147
• Fix SpatialSoftmax input shape by @alexander-soare in #150
• Support for DDIMScheduler in Diffusion Policy by @kashya...