CoolFace
Modelpublic

serialexperimentsleon/2170_0

sourceHugging Faceupdated 1y agoView on Hugging Face
0likes1downloads
config.yaml563 linesDownload Raw Back to root
1root_dir: /mnt/kostas-graid/datasets/extrinsic_contact_data2replay_buffer_size: 1500003replay_buffer_num_workers: 24nstep: 35batch_size: 1286seed: 07dataset_shuffle_seed: 28device: cuda9save_video: true10save_train_video: true11use_tb: true12use_wandb: true13wandb_run_id: '2170_0'14wandb_notes: 2170_0_restarted_1015eval: false16true_action_history: false17train_pad_after: 418process_contact_features: ${eval}19obs_type: pixels20use_color: true21use_depth: true22use_masks: true23mask_list:24- EE_obj_mask25mask_representation: channels26crop_hw:27- 14428- 14429crop_down_offset: 4830color_crop_type: null31depth_crop_type: null32segmask_crop_type: null33add_crop_binary_mask: false34add_coord_conv_map: false35use_context_color: false36use_context_depth: false37use_context_segmask: false38context_color_crop_type: null39context_depth_crop_type: null40context_segmask_crop_type: null41context_add_crop_binary_mask: false42context_add_coord_conv_map: false43use_contact_map: true44use_sdf_maps: true45use_normals_maps: true46which_objects: both47max_contact_prob: 0.148max_depth: 2.049grasped_dtc_max_value: 0.250env_dtc_max_value: 0.451grasped_normals_mask_max_dtc_value: 0.252env_normals_mask_max_dtc_value: 0.453clamp_dtc: true54dtc_adaptive_normalization: false55mask_normals_within_sdf: true56adaptive_normals_mask: true57learnable_contact_preprocess_params: true58contact_model_name: local_multitask_outhd64all_home_crop_h144w144d48_mask_ctxtmask_seed_220979_epoch_959contact_estimation_model_ckpt_path: ~/fish_leon/contact_estimation/artifacts/175604_2/checkpoints/epoch=09-val_loss=0.00.ckpt60encoder_type: small61debug_timestamps: false62open_loop: false63action_trajectories: true64stop_after_action: false65interpolation_frequency: 2566policy_frequency: 567wait_for_new_camera_frames: true68baseline: false69train_demo_idxs_list_or_num: -170log_train_every_steps: 2571name_of_expert_demo: 120_240x320_all_twodim_left_to_right_annotated_start_idx_5hz_zstd7_EE_pxl_coords_expert_demos_imp_act72expert_dataset_dirpath: ${root_dir}/FISH/expert_demos/${suite.name}/${task_name}/${name_of_expert_demo}73store_dataset_in_memory: false74expert_dataset: ${expert_dataset_dirpath}/demos.zarr75action_key: ${oc.if_else:${action_trajectories}, 'action_trajectory_${interpolation_frequency}hz',76  'action'}77semantic_demo_grouping_name: semantic_demo_grouping.yaml78semantic_demo_grouping: ${expert_dataset_dirpath}/${semantic_demo_grouping_name}79include_groups_list: all80expert_dataset_config: ${expert_dataset_dirpath}/demo_config.yaml81name_of_valid_demo: 120_240x320_all_twodim_left_to_right_annotated_start_idx_5hz_zstd7_EE_pxl_coords_expert_demos_imp_act82valid_dataset_dir: ${root_dir}/FISH/expert_demos/${suite.name}/${task_name}/${name_of_valid_demo}/demos.zarr83valid_demo_idxs_list_or_num: null84val_num_groups: 385load_bc: ${agent.load_checkpoint}86checkpoint_epoch_list:87- 9988- 19989- 29990- 39991- 49992- 59993- 69994- 79995- 89996- 99997- 124998- 149999- 1749100- 1999101- 2999102- 3999103- 4999104- 5999105- 6999106- 7999107- 8999108- 9999109snapshot_root_dir: /mnt/grasp_high_usage/leonmkim/contact_estimation/FISH110save_snapshot: true111save_last_snapshot: true112save_snapshot_when_done: true113top_k_checkpoints: 5114save_snapshot_link_to_weights_dir: deprecated115bc_regularize: false116bc_weight_type: qfilter117experiment_dir: ./exp_local/${suite.name}_${obs_type}/${task_name}/${wandb_run_id}118agent:119  _target_: agent.diffusion_policy.DiffusionPolicyAgent120  name: diffusion_policy121  load_checkpoint: ${eval}122  device: ${device}123  n_obs_steps: ${.config.policy_cfg.n_obs_steps}124  suite_name: ${suite.name}125  obs_type: ${obs_type}126  enable_arm: ${eval}127  enable_camera: ${eval}128  use_tb: ${use_tb}129  desired_image_shape:130  - 13131  - 180132  - 240133  orig_cam_shape:134  - 3135  - 240136  - 320137  config:138    _target_: agent.diffusion_policy.DiffusionPolicyAgentConfig139    compile: false140    device: ${device}141    cam_resize_shape: ${agent.desired_image_shape}142    orig_cam_shape: ${agent.orig_cam_shape}143    policy_frequency: ${policy_frequency}144    interpolation_frequency: ${interpolation_frequency}145    policy_cfg:146      _target_: lerobot.common.policies.diffusion.configuration_diffusion.DiffusionConfig147      n_obs_steps: 1148      horizon: 36149      n_action_steps: ${agent.config.policy_cfg.horizon}150      output_shapes:151        action:152        - 7153      input_normalization_modes:154        observation.image: mean_std155        observation.state: min_max156        observation.action_history: min_max157      output_normalization_modes:158        action: min_max159      vision_backbone: resnet18160      pretrained_backbone_weights: null161      transforms:162      - _target_: torchaug.transforms.RandomAffine163        degrees:164        - -5165        - 5166        translate:167        - 0.05168        - 0.05169        batch_transform: true170        num_chunks: -1171        batch_inplace: true172      - _target_: torchaug.transforms.RandomColorJitter173        brightness: 0.3174        contrast: 0.4175        saturation: 0.5176        hue: 0.08177        batch_transform: true178        num_chunks: -1179        batch_inplace: true180      use_group_norm: true181      spatial_softmax_num_keypoints: 32182      action_history_encoder_config:183        _target_: lerobot.common.policies.diffusion.configuration_diffusion.Unet1dEncoderConfig184        in_channels: 7185        out_channels: 32186        history_length: 6187        kernel_size: ${agent.config.policy_cfg.kernel_size}188        downsample_kernel_size: 3189        downsample_stride: 2190        downsample_padding: 1191      down_dims:192      - 256193      - 512194      - 1024195      kernel_size: 5196      n_groups: 8197      diffusion_step_embed_dim: 128198      use_film_scale_modulation: true199      noise_scheduler_type: DDIM200      beta_schedule: squaredcos_cap_v2201      beta_start: 0.0001202      beta_end: 0.02203      prediction_type: epsilon204      clip_sample: true205      clip_sample_range: 1.0206      num_train_timesteps: 50207      num_inference_steps: 10208      do_mask_loss_for_padding: false209      input_shapes:210        observation.image:211        - 13212        - 180213        - 240214        context_observation.image:215        - 13216        - 180217        - 240218        observation.state:219        - 8220        observation.action_history:221        - 7222    train_cfg:223      _target_: utils.TrainConfig224      lr: 0.0001225      lr_scheduler: cosine226      lr_warmup_steps: 500227      adam_betas:228      - 0.95229      - 0.999230      adam_eps: 1.0e-08231      adam_weight_decay: 1.0e-06232      grad_clip_norm: 10233      offline_steps: ${num_train_frames_diffusion}234      use_amp: true235    observation_cfg:236      _target_: agent.encoder.VisualFeatureSet237      use_depth: ${use_depth}238      use_color: ${use_color}239      mask_input_dict:240        _target_: agent.encoder.MaskInputDict241        enable: ${use_masks}242        representation: ${mask_representation}243        mask_list: ${mask_list}244      crop_input_config:245        _target_: agent.encoder.CropInputConfig246        color_crop_type: ${color_crop_type}247        depth_crop_type: ${depth_crop_type}248        segmask_crop_type: ${segmask_crop_type}249        crop_hw: ${crop_hw}250        crop_down_offset: ${crop_down_offset}251        add_crop_binary_mask: ${add_crop_binary_mask}252        add_coord_conv_map: ${add_coord_conv_map}253      context_input_config:254        _target_: agent.encoder.ContextInputConfig255        use_color: ${use_context_color}256        use_depth: ${use_context_depth}257        mask_input_dict:258          _target_: agent.encoder.MaskInputDict259          enable: ${use_context_segmask}260          representation: ${mask_representation}261          mask_list: ${mask_list}262        crop_input_config:263          _target_: agent.encoder.CropInputConfig264          color_crop_type: ${context_color_crop_type}265          depth_crop_type: ${context_depth_crop_type}266          segmask_crop_type: ${context_segmask_crop_type}267          crop_hw: ${crop_hw}268          crop_down_offset: ${crop_down_offset}269          add_crop_binary_mask: ${context_add_crop_binary_mask}270          add_coord_conv_map: ${context_add_coord_conv_map}271      mask_soft_approx_scheduler_config:272        _target_: agent.encoder.MaskSoftApproxSchedulerConfig273        num_steps: 40000274        initial_value: 10.0275        final_value: 1000.0276        interpolation_scheme: cosine277      use_contact_map: ${use_contact_map}278      use_sdf_maps: ${use_sdf_maps}279      use_normals_maps: ${use_normals_maps}280      which_objects: ${which_objects}281      grasped_dtc_max_value: ${grasped_dtc_max_value}282      env_dtc_max_value: ${env_dtc_max_value}283      grasped_normals_mask_max_dtc_value: ${grasped_normals_mask_max_dtc_value}284      env_normals_mask_max_dtc_value: ${env_normals_mask_max_dtc_value}285      clamp_dtc: ${clamp_dtc}286      max_contact_prob: ${max_contact_prob}287      mask_normals_within_sdf: ${mask_normals_within_sdf}288      dtc_adaptive_normalization: ${dtc_adaptive_normalization}289      adaptive_normals_mask: ${adaptive_normals_mask}290      max_depth: ${max_depth}291      image_shape: ${agent.desired_image_shape}292      learnable_contact_preprocess_params: ${learnable_contact_preprocess_params}293      learning_rate: 0.0001294      weight_decay: 0.0295      contact_model_name: ${contact_model_name}296      zero_centered: false297suite:298  suite: frankagym299  name: frankagym300  frame_stack: ${agent.n_obs_steps}301  action_repeat: 1302  discount: 0.99303  hidden_dim: 1024304  num_train_frames: 1000000305  num_seed_frames: 0306  num_train_epochs: 15000307  validate_every_epochs: 250308  validate_diffusion_on_action_loss_every_epochs: 250309  train_eval_diffusion_on_action_loss_every_epochs: 250310  check_topk_every_epochs: 250311  save_snapshot_every_epochs: 1500312  eval_every_frames: 2000313  num_eval_episodes: 5314  save_snapshot: true315  wait_for_user_to_start_episode: true316  task_make_fn:317    _target_: suite.frankagym.make318    name: ${task_name}319    height: 240320    width: 320321    frame_stack: ${suite.frame_stack}322    action_repeat: ${suite.action_repeat}323    seed: ${seed}324    enable_arm: ${agent.enable_arm}325    enable_gripper: ${enable_gripper}326    start_with_gripper_open: ${start_with_gripper_open}327    enable_camera: ${agent.enable_camera}328    path_to_depth_extrinsics: ${path_to_depth_extrinsics}329    contact_estimation_model_ckpt_path: ${contact_estimation_model_ckpt_path}330    x_limit: ${x_limit}331    y_limit: ${y_limit}332    z_limit: ${z_limit}333    device: ${device}334    interpolation_frequency: ${interpolation_frequency}335    policy_frequency: ${policy_frequency}336    debug_timestamps: ${debug_timestamps}337    stop_after_action: ${stop_after_action}338    open_loop: ${open_loop}339    wait_for_new_camera_frames: ${wait_for_new_camera_frames}340    action_key: ${action_key}341    action_trajectory_horizon: ${agent.config.policy_cfg.horizon}342    action_trajectories: ${action_trajectories}343    path_to_zarr_dataset: ${expert_dataset}344    agent_policy_cfg:345      _target_: agent.diffusion_policy.DiffusionPolicyAgentConfig346      compile: false347      device: ${device}348      cam_resize_shape: ${agent.desired_image_shape}349      orig_cam_shape: ${agent.orig_cam_shape}350      policy_frequency: ${policy_frequency}351      interpolation_frequency: ${interpolation_frequency}352      policy_cfg:353        _target_: lerobot.common.policies.diffusion.configuration_diffusion.DiffusionConfig354        n_obs_steps: 1355        horizon: 36356        n_action_steps: ${agent.config.policy_cfg.horizon}357        output_shapes:358          action:359          - 7360        input_normalization_modes:361          observation.image: mean_std362          observation.state: min_max363          observation.action_history: min_max364        output_normalization_modes:365          action: min_max366        vision_backbone: resnet18367        pretrained_backbone_weights: null368        transforms:369        - _target_: torchaug.transforms.RandomAffine370          degrees:371          - -5372          - 5373          translate:374          - 0.05375          - 0.05376          batch_transform: true377          num_chunks: -1378          batch_inplace: true379        - _target_: torchaug.transforms.RandomColorJitter380          brightness: 0.3381          contrast: 0.4382          saturation: 0.5383          hue: 0.08384          batch_transform: true385          num_chunks: -1386          batch_inplace: true387        use_group_norm: true388        spatial_softmax_num_keypoints: 32389        action_history_encoder_config:390          _target_: lerobot.common.policies.diffusion.configuration_diffusion.Unet1dEncoderConfig391          in_channels: 7392          out_channels: 32393          history_length: 6394          kernel_size: ${agent.config.policy_cfg.kernel_size}395          downsample_kernel_size: 3396          downsample_stride: 2397          downsample_padding: 1398        down_dims:399        - 256400        - 512401        - 1024402        kernel_size: 5403        n_groups: 8404        diffusion_step_embed_dim: 128405        use_film_scale_modulation: true406        noise_scheduler_type: DDIM407        beta_schedule: squaredcos_cap_v2408        beta_start: 0.0001409        beta_end: 0.02410        prediction_type: epsilon411        clip_sample: true412        clip_sample_range: 1.0413        num_train_timesteps: 50414        num_inference_steps: 10415        do_mask_loss_for_padding: false416        input_shapes:417          observation.image:418          - 13419          - 180420          - 240421          context_observation.image:422          - 13423          - 180424          - 240425          observation.state:426          - 8427          observation.action_history:428          - 7429      train_cfg:430        _target_: utils.TrainConfig431        lr: 0.0001432        lr_scheduler: cosine433        lr_warmup_steps: 500434        adam_betas:435        - 0.95436        - 0.999437        adam_eps: 1.0e-08438        adam_weight_decay: 1.0e-06439        grad_clip_norm: 10440        offline_steps: ${num_train_frames_diffusion}441        use_amp: true442      observation_cfg:443        _target_: agent.encoder.VisualFeatureSet444        use_depth: ${use_depth}445        use_color: ${use_color}446        mask_input_dict:447          _target_: agent.encoder.MaskInputDict448          enable: ${use_masks}449          representation: ${mask_representation}450          mask_list: ${mask_list}451        crop_input_config:452          _target_: agent.encoder.CropInputConfig453          color_crop_type: ${color_crop_type}454          depth_crop_type: ${depth_crop_type}455          segmask_crop_type: ${segmask_crop_type}456          crop_hw: ${crop_hw}457          crop_down_offset: ${crop_down_offset}458          add_crop_binary_mask: ${add_crop_binary_mask}459          add_coord_conv_map: ${add_coord_conv_map}460        context_input_config:461          _target_: agent.encoder.ContextInputConfig462          use_color: ${use_context_color}463          use_depth: ${use_context_depth}464          mask_input_dict:465            _target_: agent.encoder.MaskInputDict466            enable: ${use_context_segmask}467            representation: ${mask_representation}468            mask_list: ${mask_list}469          crop_input_config:470            _target_: agent.encoder.CropInputConfig471            color_crop_type: ${context_color_crop_type}472            depth_crop_type: ${context_depth_crop_type}473            segmask_crop_type: ${context_segmask_crop_type}474            crop_hw: ${crop_hw}475            crop_down_offset: ${crop_down_offset}476            add_crop_binary_mask: ${context_add_crop_binary_mask}477            add_coord_conv_map: ${context_add_coord_conv_map}478        mask_soft_approx_scheduler_config:479          _target_: agent.encoder.MaskSoftApproxSchedulerConfig480          num_steps: 40000481          initial_value: 10.0482          final_value: 1000.0483          interpolation_scheme: cosine484        use_contact_map: ${use_contact_map}485        use_sdf_maps: ${use_sdf_maps}486        use_normals_maps: ${use_normals_maps}487        which_objects: ${which_objects}488        grasped_dtc_max_value: ${grasped_dtc_max_value}489        env_dtc_max_value: ${env_dtc_max_value}490        grasped_normals_mask_max_dtc_value: ${grasped_normals_mask_max_dtc_value}491        env_normals_mask_max_dtc_value: ${env_normals_mask_max_dtc_value}492        clamp_dtc: ${clamp_dtc}493        max_contact_prob: ${max_contact_prob}494        mask_normals_within_sdf: ${mask_normals_within_sdf}495        dtc_adaptive_normalization: ${dtc_adaptive_normalization}496        adaptive_normals_mask: ${adaptive_normals_mask}497        max_depth: ${max_depth}498        image_shape: ${agent.desired_image_shape}499        learnable_contact_preprocess_params: ${learnable_contact_preprocess_params}500        learning_rate: 0.0001501        weight_decay: 0.0502        contact_model_name: ${contact_model_name}503        zero_centered: false504    true_action_history: ${true_action_history}505num_train_frames_bc: 50000506num_train_frames_drq: 1100000507stddev_schedule_drq: linear(1.0,0.1,100000)508task_name: FrankaInsertion-v1509num_train_frames_vinn: 25000510num_train_frames_diffusion: 1000000511num_train_epochs_bc: 5000512num_train_epochs_diffusion: 15000513validate_every_epochs_bc: 5514validate_every_epochs_diffusion: 250515validate_diffusion_on_action_loss_every_epochs: 250516train_eval_diffusion_on_action_loss_every_epochs: 250517check_topk_every_epochs: 5518check_topk_every_epochs_diffusion: 250519save_snapshot_every_epochs_diffusion: 1500520x_limit:521- 0.2522- 0.7523y_limit:524- -0.4525- 0.4526z_limit:527- -0.05528- 0.55529home_displacement:530- 0.55531- 0.0532- 0.55533- 180.0534- 0.0535- 0.0536enable_gripper: true537start_with_gripper_open: true538offset_mask:539- 1540- 1541- 1542- 1543- 1544- 1545path_to_depth_extrinsics: ~/fish_leon/FISH/cfgs/camera_poses/camera_poses_L515/20240904-122305/color_tf_world.npy546test:547  username: leonmkim548  hostname: grasp-login1549  name_of_expert_demo: 112_240x320_all_twodim_left_to_right_annotated_start_idx_5hz_zstd7_EE_pxl_coords_expert_demos_imp_act550  root_dir: ~/fish_leon551  expert_dataset_dirpath: ${test.root_dir}/FISH/expert_demos/${suite.name}/${task_name}/${test.name_of_expert_demo}552  expert_dataset: ${test.expert_dataset_dirpath}/demos.zarr553  semantic_demo_grouping_name: semantic_demo_grouping.yaml554  semantic_demo_grouping: ${test.expert_dataset_dirpath}/${test.semantic_demo_grouping_name}555  include_groups_list: all556  batch_size: 128557  store_dataset_in_memory: false558  use_tb: true559  local_snapshot_root_dir: /mnt/bighdd/fish_contact_backup560  local_snapshot_dir: ${test.local_snapshot_root_dir}/exp_local/${suite.name}_${obs_type}/${task_name}561  resume_wandb_run: false562feature_type: 180x240_1_RGB_D_2.0_msk_channels_EE_obj_mask_cntct_0.1_DTC_clmpd_lrnbl_nrmls_DTCmask_adpt_lrnbl_both_lr_0.0001_wd_0.0_local_multitask_outhd64all_home_crop_h144w144d48_mask_ctxtmask_seed_220979_epoch_9_acthst_hst6_out32_dwnkrnl3_dwnstrd2_dwnpd1563