Tags: ARISE-Initiative/robomimic
Tags
v0.5 release (#243) * added ResNet50Conv * dataset_states_to_obs: copy over ep_info (if applicable) * extract cam intrinsics in convert_r2d2.py * env update * added bc_rnn config * update name * fix critical bug with filter_key * Multiple backbones for ACT * deleting extraneous files * bc_rnn gen: use gmm * data_set_observations now multiprocessed * updating hyperparameters * process number tuning * Update diffusion_policy.py: fix bug in _get_action_trajectory function to use ema model everywhere * multiprocessing statse ot obs * batched evals; wip * multicore dataset states to obs * fix rollout camera settings, save all videos * set torch/torchvision versions in setup.py * dataset to obs * Update README.md * Update README.md * multiprocessed dataset to obs and update rollout horizon in bc_xfmr * use of lock in work retrievla * add language conditioning * remove extra reset call; treat lang as optional * reverting some file changes, and keeping old dataset_states_to_obs.py script * don't include lang_emb in observations if none * accompanying changes for datasets with multiple layouts/styles * save ep_meta in get_state func * added support for libero envs * added libero download instructions to readme * removed need to specify data_format libero * minor changes to dp config * update readme, fix dp config * added dp config for all libero90kitchen * minor update dp config * added dp config for libero-90 living * updated dp config for libero-90 living * added libero env to config_gen helper * added lang to dp config * update dp config * added bc config * minor update to bc config * added bc config * added bc config * add lang cond to 2-task bc3.json * added bc config * added bc config * added bc_transformer config * fix lang cond * removed algo.language_conditioned * update bc configs * update bc_transformer config * added mimicgen imports to env_robosuite * added bc configs for exp one_scene_two_task_lang_cond * added bc config * removed norm in config * added bc configs for exp * added bc_rnn configs for exp * Adding files for multitask eval * added bc config * added noLang config * added noLang-noActNorm config * added new lowdims to cofig * mapped gripper_states and joint_states obs keys from robosuite to libero * minor change to render cam * added rgb color-rand config * changed rollout rate in config * added bc config for exp * added bc config with no low dim * changed rollout rate in config * added support for libero mimicgen datasets * added config for libero_mg data * fixed obs keys for libero * added libero_mg configs * added libero_mg config * added libero_mg config for all 7 datasets * added configs * fixed libero-pretrained configs * added configs for testing rgbFeat1024 * update libero_mg configs with dataset names * added two-task config libero_mg * fixed two-task config libero_mg * fixed two-task config libero_mg * a few changes to the configs * added libero_mg config 7tasks-noEval * added multistack configs * minor fix * added multistack rgbFeat1024 config * adding changes for multitask eval * removing jsons * fixing multitask documentation * added libero_mg exp config * try adding film layers * added key to dataset config to be used for naming during rollouts * added configs for long-hor-similarity exps * update config * added long-hor-sim exp2 configs * update configs * added task-hor-sim exp3 configs * added single-task and open-loop configs * added exp1-batch2 openloop config * added three-piece-assembly-d2 config * added more three-piece-assembly configs * demo limit changes from matt's branch * added 100 demos configs * added 50 demo limit configs * minor fix * added liberoD1 configs * added exp2, exp3 configs * added demo_libero script * (minor) added examples to demo_libero.py * added support for demo_limit per dataset * added rmv2 exp1 (visual and spatial) train configs * added exp1-varyTask config * added singleTask configs for pretrain tasks in exp1 * increased horizon 400->1600 for singleTask configs * added exp1 all co-train config * changed horizon len in configs 400->1600 * added libero_finetune single task configs * minor fix to configs * added target100-cotrain1000 config * update tianshou requirement * enabled batched rollout in one exp * minor fixes for batched rollout * added libero_finetune exp2 configs * added libero_finetune exp2 target50 configs * added libero_finetune target10 configs * added libero_finetune exp2 weighted cotraining configs * added exp3 configs, and exp2 bigCamPose5Normal configs for newCamPose task * get normalize_weights_by_ds_size from config * minor fix in vid rollout * added exp3-newScene configs * added example usages in demo_libero * added scripts/teleop_to_env_meta.py * added exp3 target10 cotrain1000 5050 configs * updated configs * change clip download dir * added custom weighted data sampler * added second libero import to subside error * cleanup configs * added support for cosine lr schedule, changes to bc and dp configs * added resume functionality * removed bddl_file_name processing in file_utils, added mimiclabs to env_robosuite * cleanup libero * moved diffusion policy nets to a new file in models, minor styling fixes * moved postprocess_visual_obs from env to RolloutPolicy * deprecate EnvGibsonMOMART * fixes for chaining obs randomizers * added mimiclabs import * update default config for diffusion policy * ran generate_config_templates.py * prefer last ckpt over ckpt_path when resuming * update diffusion_policy default optim params * remove r2d2 datasets; still keeping conversion script, config generators, and in mse viz in algo.py * removed commented out sanity check in obs_encoder_kwargs_from_config() * moved deep_update to python_utils and removed script_utils * removed mse logging, updated base_config and algo config templates * added backwards compatibility for loading ckpts saved before adding resume support * removed batched rollouts support using tianshou * updated docs for diffusion policy * removed tianshou in setup.py * removed ACT * removed scripts demo_libero.py teleop_to_env_meta.py * removed ACT - (minor) remove act_config import * removed more non-useful scripts * removed ACT - removed config gen * fix dataset config, minor cleanup * removed data_format (ds_format) * moved action_utils into python_utils * minor fixes in train.py for storing list of env and shape metas * backwards compatibility for string dataset path in config * added default action_config to base_config * removed shuffled_obs_key_groups * added docs postprocessing, rename _static to static * added robosuite-1.5 support in robosuite_add_absolute_actions.py * minor naming fix for kwargs in RolloutPolicy * removed config_gen scripts * updated requirements, removed robosuite141 conversion script, minor comments cleaned * minor styling fix in diffusion_policy.py * removed changes in README.md * removed TODO comment in wrappers.py for action padding and action hist in obs * added new script dataset_states_to_obs_mp.py * added docstring for meta dataset classes * added acknowledgement for pytorch3d * update docs installation with make command for renaming _static * removed unused eval in robosuite_add_absolute_actions.py, added comments * minor change in .gitignore * minor changes to scripts * update acknowledgements * added docstring to abs action script, update team * remove folder * minor docs updates * changed docs/prepare_html.py * minor fix in setup.py * fix missing gifs in readme * update version string * bump version in setup * add Vaibhav as author * minor docs changes * fixed success check and exception handling for rollouts in train_utils.py * add pointers to old v04 docs * fixed support for additional_envs * removed config.algo.transformer.casual, updated templates * put adding abs actions behind arg in convert_robosuite.py * added docstrings in train_utils.py * added actions_abs to dataset_states_to_obs_mp.py * added multi_dataset_training.md * added action configs tutorial * fixed bug in gaussian action norm * minor changes and comment fixes * update docs for resume functionality, added lang cond tutorial * load optimizer state from ckpt only when resuming * minor fixes in docs * update docs for env_meta_update_dict * fix for config.experiment.validate * torch>=2.6 requires weights_only kwarg for torch.load * fixes to continue supporting hbc and iris * made getting and loading state dicts recursive over list/dict * fix to suppress error from torch.from_numpy() about negative stride * support None and single randomizer in ObservationEncoder.register_obs_key(), fix test and docs * suppress negative stride error from torch.from_numpy() for depth obs * fix tests * remove obs unprocess in run_trained_agent --------- Co-authored-by: snasiriany <[email protected]> Co-authored-by: Soroush Nasiriany <[email protected]> Co-authored-by: Aaron Lo <[email protected]> Co-authored-by: Soroush Nasiriany <[email protected]> Co-authored-by: Vaibhav Saxena <[email protected]> Co-authored-by: MBronars <[email protected]>
v0.4 release (#219) * add robocasa import and ep_meta support * update logic for robosuite v1.5 check * bump version * Hf v15 (#214) * some minor changes * release robomimic datasets compatible with robosuite v1.5 on huggingface * some fixes for visualization with robosuite v1.5 * potential fix for video render when using depth images during training * minor fix * new paths for test hdf5 based on robosuite version and HF * update docs --------- Co-authored-by: snasiriany <[email protected]> Co-authored-by: Soroush Nasiriany <[email protected]>
Release v0.3 (#63) * colab links in readme and docs * move colab link to getting_started in docs, add colab banner in readme * basic support for logging warnings * wandb logging; revamped warning logging; macros * addressed PR comments; moved wandb proj name to experiments.logging section; hyperparam generator populates tags * change tags to meta. add wandb api key to macros * check that Macros.WANDB_ENTITY is not None * fix depth unprocessing, fix config modality handling * limit torch threads * space saving options in dataset extraction script * changes to make image uint8 to float conversion happen on GPU * disable load next obs if BC * small bug fix * add storage flags to docs * add some 2022 refs * add viola * add link to old docs on tab bar and logo into tab bar * make sure kwargs are empty in default configs for easy override * move some models into obs core file * minor change * work with older robosuite versions * add some 2023 refs * making validation hdf5 key explicit * IQL implementation; todo: comments and testing * revamped comments for IQL * add minimal iql test and fix template * added comments and tests; removed filter_type feature * updated docs and configs for d4rl v2 envs; added iql results * update iql template json config to reflect v0.3 changes * generate iql configs for d4rl * fixed small bug with d4rl output dirs * update version numbers for d4rl docs * adding r3m and mvp pretraining * clean * add doc for pretrain * fix * revert change to env robosuite * minor rename * update pretrained reps tutorial * fix minor issues * small update to iql config * small docs fix * add orbit reference * more 2023 papers * working version of color jitter net * add gaussian noise randomizer * fix bug in GaussianNoiseRandomizer * randomizer visualization * center cropping for CropRandomizer eval mode * viz randomizer * move randomizers to obs_core.py * remove dups * remove dups * missing import * doc instruction for visualizing randomizer * Fix tab issue * update reqs with matplotlib * modified docs for new datasets, new algos, wandb support * add note about using offline_study datasets * postprocess xml for v1.3 * update postprocess function * added verification exps * limit num torch threads * updated dataset gen script * mj211 verification runs * started conversion script for robosuite v1.3.1. WIP * initial conversion script ready * add verion to env_robosuite; fix bug from observation modality PR * added v_1_3_1 verification jsons * converted lift dataset, wip on can * playback with can appears working but state divergence still high * all tasks minimally in place now without mujoco exceptions * minimal script for robosuite v140 conversion; created configs * updated robosuite v140 convertion script * added version checks for robosuite * update dataset registry to point to v141 raw datasets * generating and downloading datasets for v141 * updated obs extraction script to use v141 robosuite datasets * remove unsued scripts * remove unnec script * delete old verification scripts * remove unnec changes * add low_dim dataset links back in, more informative message for unstored datasets * Transformer support (#38) * initial BC-mingpt implementation * added adamw and weight scheduler * updated mingpt configs for sweep * rename mingpt -> transformer * cleaned up bc transformer in bc.py * cleaned up code comments * remove internal config gen scripts * addressing Danfei's comments in PR * update template configs and make some cosmetic changes * add ack for mingpt --------- Co-authored-by: Ajay Mandlekar <[email protected]> Co-authored-by: amandlek <[email protected]> * bump version and minor docs update * remove file * some minor docs fixes for updated test_v141 hdf5 * small change to gitignore for test hdf5 * some minor docs fixes * update 2023 refs * Update obs_core.py with import * Add wandb description in "viewing results" tutorial page * added more wandb info to docs * fix observation postprocessing for nested obs dicts * added bc transformer tests * V0.3 transformer tutorial (#59) * added transformer tutorial * Update bc_transformer.json * Update bc_transformer.json * Update training_transformers.md * adjusting tutorial docs based on feedback * changes to pretrained representations and transformer tutorial * added data loader params for pretrained reps tutorial --------- Co-authored-by: Bronars <[email protected]> * add readme note * V0.3 pretrained rep fix (#64) * fixed error in base_nets for R3MConv * added error messages for Sequential * added doc string for Sequential init --------- Co-authored-by: Danfei Xu <[email protected]> Co-authored-by: snasiriany <[email protected]> Co-authored-by: josiah_wong <[email protected]> Co-authored-by: Soroush Nasiriany <[email protected]> Co-authored-by: Soroush Nasiriany <[email protected]> Co-authored-by: j96w <[email protected]> Co-authored-by: danfei <[email protected]> Co-authored-by: MBronars <[email protected]> Co-authored-by: Bronars <[email protected]>