Add files using upload-large-folder tool

.gitattributes

@@ -33,3 +33,18 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+205317/wandb/run-20241229_205323-2n31umej/files/media/videos/eval/1_eval_1_f35843717fd499551c4b.mp4 filter=lfs diff=lfs merge=lfs -text
+205317/wandb/run-20241229_205323-2n31umej/files/media/videos/eval/2_eval_2_c03a74d7b4bcab766fc2.mp4 filter=lfs diff=lfs merge=lfs -text
+205317/wandb/run-20241229_205323-2n31umej/files/media/videos/eval/0_eval_0_2dd247d1c4c04fa7cea7.mp4 filter=lfs diff=lfs merge=lfs -text
+205317/wandb/run-20241229_205323-2n31umej/files/media/videos/eval/3_eval_3_94417ff346f07f5fa23e.mp4 filter=lfs diff=lfs merge=lfs -text
+205317/wandb/run-20241229_205323-2n31umej/files/media/videos/eval/4_eval_4_20ea9dbf78cd7cd19362.mp4 filter=lfs diff=lfs merge=lfs -text
+205317/eval_video/3_eval.mp4 filter=lfs diff=lfs merge=lfs -text
+205317/eval_video/1_eval.mp4 filter=lfs diff=lfs merge=lfs -text
+205317/eval_video/4_eval.mp4 filter=lfs diff=lfs merge=lfs -text
+205317/eval_video/2_eval.mp4 filter=lfs diff=lfs merge=lfs -text
+205317/eval_video/0_eval.mp4 filter=lfs diff=lfs merge=lfs -text
+205317/episode_rosbags/episode_4_2024-12-29-20-57-31.bag filter=lfs diff=lfs merge=lfs -text
+205317/episode_rosbags/episode_1_2024-12-29-20-55-29.bag filter=lfs diff=lfs merge=lfs -text
+205317/episode_rosbags/episode_2_2024-12-29-20-56-10.bag filter=lfs diff=lfs merge=lfs -text
+205317/episode_rosbags/episode_3_2024-12-29-20-56-49.bag filter=lfs diff=lfs merge=lfs -text
+205317/episode_rosbags/episode_0_2024-12-29-20-54-40.bag filter=lfs diff=lfs merge=lfs -text

205131/.hydra/config.yaml

@@ -0,0 +1,350 @@
+root_dir: /home/${oc.env:USER}/fish_leon
+nstep: 3
+seed: 41
+dataset_shuffle_seed: ${seed}
+device: cuda
+save_video: true
+save_buffer: true
+use_tb: true
+baseline: false
+use_wandb: true
+eval: true
+process_contact_features: ${eval}
+obs_type: pixels
+use_color: true
+use_depth: true
+use_masks: false
+mask_list:
+- EE_obj_mask
+mask_representation: channels
+crop_hw:
+- 144
+- 144
+crop_down_offset: 48
+color_crop_type: null
+depth_crop_type: null
+segmask_crop_type: null
+add_crop_binary_mask: false
+add_coord_conv_map: false
+use_context_color: false
+use_context_depth: false
+use_context_segmask: false
+context_color_crop_type: null
+context_depth_crop_type: null
+context_segmask_crop_type: null
+context_add_crop_binary_mask: false
+context_add_coord_conv_map: false
+use_contact_map: false
+use_sdf_maps: false
+use_normals_maps: false
+which_objects: both
+max_contact_prob: 0.1
+max_depth: 2.0
+grasped_dtc_max_value: 0.105
+env_dtc_max_value: 0.425
+grasped_normals_mask_max_dtc_value: 0.105
+env_normals_mask_max_dtc_value: 0.425
+clamp_dtc: true
+dtc_adaptive_normalization: false
+mask_normals_within_sdf: true
+adaptive_normals_mask: true
+learnable_contact_preprocess_params: false
+contact_model_name: local_multitask_outhd64all_home_crop_h144w144d48_ctxt_seed_183386_epoch_9
+contact_estimation_model_ckpt_path: ~/fish_leon/contact_estimation/artifacts/175604_2/checkpoints/epoch=09-val_loss=0.00.ckpt
+num_eval: 5
+debug_timestamps: false
+open_loop: false
+action_trajectories: true
+stop_after_action: false
+interpolation_frequency: 25
+policy_frequency: 5
+wait_for_new_camera_frames: true
+random_start: false
+eval_starts: ${root_dir}/FISH/eval_starts/${suite.name}_${obs_type}/${task_name}
+train_demo_idxs_list_or_num: null
+num_valid_demos: null
+val_num_groups: 3
+name_of_expert_demo: 112_240x320_all_twodim_left_to_right_annotated_start_idx_5hz_zstd7_EE_pxl_coords_expert_demos_imp_act
+expert_dataset_dirpath: ${root_dir}/FISH/expert_demos/${suite.name}/${task_name}/${name_of_expert_demo}
+expert_dataset: ${expert_dataset_dirpath}/demos.zarr
+action_key: ${oc.if_else:${action_trajectories}, 'action_trajectory_${interpolation_frequency}hz',
+  'action'}
+semantic_demo_grouping_name: semantic_demo_grouping.yaml
+semantic_demo_grouping: ${expert_dataset_dirpath}/${semantic_demo_grouping_name}
+expert_dataset_config: ${expert_dataset_dirpath}/demo_config.yaml
+bc_regularize: false
+bc_weight_type: qfilter
+load_checkpoint: ${agent.load_checkpoint}
+wandb_run_id: '1045_0'
+true_action_history: false
+wandb_notes: null
+checkpoint_epoch: 12000
+load_residual_weight: false
+checkpoint_root_dir: /home/${oc.env:USER}/fish_leon/FISH
+checkpoint_weight_dir: ${checkpoint_root_dir}/exp_local/${suite.name}_${obs_type}/${task_name}/${wandb_run_id}
+residual_weight: ${root_dir}/FISH/weights/${suite.name}_${obs_type}/${task_name}/weight.pt
+experiment_dir: ./exp_local/${suite.name}_${obs_type}/${task_name}/${wandb_run_id}
+final_experiment_dir: ${experiment_dir}/${now:%H%M%S}
+agent:
+  _target_: agent.diffusion_policy.DiffusionPolicyAgent
+  name: diffusion_policy
+  load_checkpoint: ${eval}
+  device: ${device}
+  n_obs_steps: ${.config.policy_cfg.n_obs_steps}
+  suite_name: ${suite.name}
+  obs_type: ${obs_type}
+  enable_arm: ${eval}
+  enable_camera: ${eval}
+  use_tb: ${use_tb}
+  desired_image_shape:
+  - 13
+  - 180
+  - 240
+  orig_cam_shape:
+  - 3
+  - 240
+  - 320
+  config:
+    _target_: agent.diffusion_policy.DiffusionPolicyAgentConfig
+    compile: false
+    device: ${device}
+    cam_resize_shape: ${agent.desired_image_shape}
+    orig_cam_shape: ${agent.orig_cam_shape}
+    policy_frequency: ${policy_frequency}
+    interpolation_frequency: ${interpolation_frequency}
+    policy_cfg:
+      _target_: lerobot.common.policies.diffusion.configuration_diffusion.DiffusionConfig
+      n_obs_steps: 1
+      horizon: 36
+      n_action_steps: ${agent.config.policy_cfg.horizon}
+      input_shapes:
+        observation.image: ${agent.config.cam_resize_shape}
+        context_observation.image: ${agent.config.cam_resize_shape}
+        observation.state:
+        - 8
+        observation.action_history:
+        - 7
+      output_shapes:
+        action:
+        - 7
+      input_normalization_modes:
+        observation.image: mean_std
+        observation.state: min_max
+        observation.action_history: min_max
+      output_normalization_modes:
+        action: min_max
+      vision_backbone: resnet18
+      pretrained_backbone_weights: null
+      transforms:
+      - _target_: torchaug.transforms.RandomAffine
+        degrees:
+        - -5
+        - 5
+        translate:
+        - 0.05
+        - 0.05
+        batch_transform: true
+        num_chunks: -1
+        batch_inplace: true
+      - _target_: torchaug.transforms.RandomColorJitter
+        brightness: 0.3
+        contrast: 0.4
+        saturation: 0.5
+        hue: 0.08
+        batch_transform: true
+        num_chunks: -1
+        batch_inplace: true
+      use_group_norm: true
+      spatial_softmax_num_keypoints: 32
+      action_history_encoder_config:
+        _target_: lerobot.common.policies.diffusion.configuration_diffusion.Unet1dEncoderConfig
+        in_channels: 7
+        out_channels: 32
+        history_length: ${agent.config.policy_cfg.n_action_steps}
+        kernel_size: ${agent.config.policy_cfg.kernel_size}
+        downsample_kernel_size: 3
+        downsample_stride: 2
+        downsample_padding: 1
+      down_dims:
+      - 256
+      - 512
+      - 1024
+      kernel_size: 5
+      n_groups: 8
+      diffusion_step_embed_dim: 128
+      use_film_scale_modulation: true
+      noise_scheduler_type: DDIM
+      beta_schedule: squaredcos_cap_v2
+      beta_start: 0.0001
+      beta_end: 0.02
+      prediction_type: epsilon
+      clip_sample: true
+      clip_sample_range: 1.0
+      num_train_timesteps: 50
+      num_inference_steps: 10
+      do_mask_loss_for_padding: false
+    train_cfg:
+      _target_: utils.TrainConfig
+      lr: 0.0001
+      lr_scheduler: cosine
+      lr_warmup_steps: 500
+      adam_betas:
+      - 0.95
+      - 0.999
+      adam_eps: 1.0e-08
+      adam_weight_decay: 1.0e-06
+      grad_clip_norm: 10
+      offline_steps: ${num_train_frames_diffusion}
+      use_amp: true
+    observation_cfg:
+      _target_: agent.encoder.VisualFeatureSet
+      use_depth: ${use_depth}
+      use_color: ${use_color}
+      mask_input_dict:
+        _target_: agent.encoder.MaskInputDict
+        enable: ${use_masks}
+        representation: ${mask_representation}
+        mask_list: ${mask_list}
+      crop_input_config:
+        _target_: agent.encoder.CropInputConfig
+        color_crop_type: ${color_crop_type}
+        depth_crop_type: ${depth_crop_type}
+        segmask_crop_type: ${segmask_crop_type}
+        crop_hw: ${crop_hw}
+        crop_down_offset: ${crop_down_offset}
+        add_crop_binary_mask: ${add_crop_binary_mask}
+        add_coord_conv_map: ${add_coord_conv_map}
+      context_input_config:
+        _target_: agent.encoder.ContextInputConfig
+        use_color: ${use_context_color}
+        use_depth: ${use_context_depth}
+        mask_input_dict:
+          _target_: agent.encoder.MaskInputDict
+          enable: ${use_context_segmask}
+          representation: ${mask_representation}
+          mask_list: ${mask_list}
+        crop_input_config:
+          _target_: agent.encoder.CropInputConfig
+          color_crop_type: ${context_color_crop_type}
+          depth_crop_type: ${context_depth_crop_type}
+          segmask_crop_type: ${context_segmask_crop_type}
+          crop_hw: ${crop_hw}
+          crop_down_offset: ${crop_down_offset}
+          add_crop_binary_mask: ${context_add_crop_binary_mask}
+          add_coord_conv_map: ${context_add_coord_conv_map}
+      mask_soft_approx_scheduler_config:
+        _target_: agent.encoder.MaskSoftApproxSchedulerConfig
+        num_steps: 40000
+        initial_value: 10.0
+        final_value: 1000.0
+        interpolation_scheme: constant
+      use_contact_map: ${use_contact_map}
+      use_sdf_maps: ${use_sdf_maps}
+      use_normals_maps: ${use_normals_maps}
+      which_objects: ${which_objects}
+      grasped_dtc_max_value: ${grasped_dtc_max_value}
+      env_dtc_max_value: ${env_dtc_max_value}
+      grasped_normals_mask_max_dtc_value: ${grasped_normals_mask_max_dtc_value}
+      env_normals_mask_max_dtc_value: ${env_normals_mask_max_dtc_value}
+      clamp_dtc: ${clamp_dtc}
+      max_contact_prob: ${max_contact_prob}
+      mask_normals_within_sdf: ${mask_normals_within_sdf}
+      dtc_adaptive_normalization: ${dtc_adaptive_normalization}
+      adaptive_normals_mask: ${adaptive_normals_mask}
+      max_depth: ${max_depth}
+      image_shape: ${agent.desired_image_shape}
+      learnable_contact_preprocess_params: ${learnable_contact_preprocess_params}
+      learning_rate: ${agent.config.train_cfg.lr}
+      weight_decay: 0.0
+      contact_model_name: ${contact_model_name}
+      zero_centered: false
+suite:
+  suite: frankagym
+  name: frankagym
+  frame_stack: ${agent.n_obs_steps}
+  action_repeat: 1
+  discount: 0.99
+  hidden_dim: 1024
+  num_train_frames: 2010
+  num_seed_frames: 260
+  num_train_epochs: 5000
+  validate_every_epochs: 100
+  validate_diffusion_on_action_loss_every_epochs: 500
+  train_eval_diffusion_on_action_loss_every_epochs: 500
+  check_topk_every_epochs: 10
+  save_snapshot_every_epochs: 5000
+  eval_every_frames: 2000
+  num_eval_episodes: 5
+  save_snapshot: true
+  wait_for_user_to_start_episode: true
+  task_make_fn:
+    _target_: suite.frankagym.make
+    name: ${task_name}
+    height: 240
+    width: 320
+    frame_stack: ${suite.frame_stack}
+    action_repeat: ${suite.action_repeat}
+    seed: ${seed}
+    enable_arm: ${agent.enable_arm}
+    enable_gripper: ${enable_gripper}
+    start_with_gripper_open: ${start_with_gripper_open}
+    enable_camera: ${agent.enable_camera}
+    path_to_depth_extrinsics: ${path_to_depth_extrinsics}
+    contact_estimation_model_ckpt_path: ${contact_estimation_model_ckpt_path}
+    x_limit: ${x_limit}
+    y_limit: ${y_limit}
+    z_limit: ${z_limit}
+    device: ${device}
+    interpolation_frequency: ${interpolation_frequency}
+    policy_frequency: ${policy_frequency}
+    debug_timestamps: ${debug_timestamps}
+    stop_after_action: ${stop_after_action}
+    open_loop: ${open_loop}
+    wait_for_new_camera_frames: ${wait_for_new_camera_frames}
+    action_key: ${action_key}
+    action_trajectory_horizon: ${agent.config.policy_cfg.horizon}
+    action_trajectories: ${action_trajectories}
+    path_to_zarr_dataset: ${expert_dataset}
+    agent_policy_cfg: ???
+    true_action_history: ${true_action_history}
+num_train_frames_bc: 50000
+num_train_frames_drq: 1100000
+stddev_schedule_drq: linear(1.0,0.1,100000)
+task_name: FrankaInsertion-v1
+num_train_frames_vinn: 25000
+num_train_frames_diffusion: 1000000
+num_train_epochs_bc: 5000
+num_train_epochs_diffusion: 5000
+validate_every_epochs_bc: 5
+validate_every_epochs_diffusion: 25
+validate_diffusion_on_action_loss_every_epochs: 50
+train_eval_diffusion_on_action_loss_every_epochs: 500
+check_topk_every_epochs: 5
+check_topk_every_epochs_diffusion: ${validate_diffusion_on_action_loss_every_epochs}
+save_snapshot_every_epochs_diffusion: 5000
+x_limit:
+- 0.2
+- 0.7
+y_limit:
+- -0.4
+- 0.4
+z_limit:
+- -0.05
+- 0.55
+home_displacement:
+- 0.55
+- 0.0
+- 0.55
+- 180.0
+- 0.0
+- 0.0
+enable_gripper: true
+start_with_gripper_open: true
+offset_mask:
+- 1
+- 1
+- 1
+- 1
+- 1
+- 1
+path_to_depth_extrinsics: ~/fish_leon/FISH/cfgs/camera_poses/camera_poses_L515/20240904-122305/color_tf_world.npy

205131/.hydra/hydra.yaml

@@ -0,0 +1,169 @@
+hydra:
+  run:
+    dir: ${final_experiment_dir}
+  sweep:
+    dir: ${final_experiment_dir}
+    subdir: ${hydra.job.num}
+  launcher:
+    submitit_folder: ${final_experiment_dir}/.slurm
+    timeout_min: 60
+    cpus_per_task: null
+    gpus_per_node: null
+    tasks_per_node: 1
+    mem_gb: null
+    nodes: 1
+    name: ${hydra.job.name}
+    stderr_to_stdout: false
+    _target_: hydra_plugins.hydra_submitit_launcher.submitit_launcher.LocalLauncher
+  sweeper:
+    _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper
+    max_batch_size: null
+    params: null
+  help:
+    app_name: ${hydra.job.name}
+    header: '${hydra.help.app_name} is powered by Hydra.
+
+      '
+    footer: 'Powered by Hydra (https://hydra.cc)
+
+      Use --hydra-help to view Hydra specific help
+
+      '
+    template: '${hydra.help.header}
+
+      == Configuration groups ==
+
+      Compose your configuration from those groups (group=option)
+
+
+      $APP_CONFIG_GROUPS
+
+
+      == Config ==
+
+      Override anything in the config (foo.bar=value)
+
+
+      $CONFIG
+
+
+      ${hydra.help.footer}
+
+      '
+  hydra_help:
+    template: 'Hydra (${hydra.runtime.version})
+
+      See https://hydra.cc for more info.
+
+
+      == Flags ==
+
+      $FLAGS_HELP
+
+
+      == Configuration groups ==
+
+      Compose your configuration from those groups (For example, append hydra/job_logging=disabled
+      to command line)
+
+
+      $HYDRA_CONFIG_GROUPS
+
+
+      Use ''--cfg hydra'' to Show the Hydra config.
+
+      '
+    hydra_help: ???
+  hydra_logging:
+    version: 1
+    formatters:
+      simple:
+        format: '[%(asctime)s][HYDRA] %(message)s'
+    handlers:
+      console:
+        class: logging.StreamHandler
+        formatter: simple
+        stream: ext://sys.stdout
+    root:
+      level: INFO
+      handlers:
+      - console
+    loggers:
+      logging_example:
+        level: DEBUG
+    disable_existing_loggers: false
+  job_logging:
+    version: 1
+    formatters:
+      simple:
+        format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s'
+    handlers:
+      console:
+        class: logging.StreamHandler
+        formatter: simple
+        stream: ext://sys.stdout
+      file:
+        class: logging.FileHandler
+        formatter: simple
+        filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log
+    root:
+      level: INFO
+      handlers:
+      - console
+      - file
+    disable_existing_loggers: false
+  env: {}
+  mode: RUN
+  searchpath: []
+  callbacks: {}
+  output_subdir: .hydra
+  overrides:
+    hydra:
+    - hydra.mode=RUN
+    task:
+    - agent=diffusion
+    - suite=frankagym
+    - suite/frankagym_task@_global_=insertion
+  job:
+    name: eval_policy
+    chdir: true
+    override_dirname: agent=diffusion,suite/frankagym_task@_global_=insertion,suite=frankagym
+    id: ???
+    num: ???
+    config_name: config_eval
+    env_set: {}
+    env_copy: []
+    config:
+      override_dirname:
+        kv_sep: '='
+        item_sep: ','
+        exclude_keys: []
+  runtime:
+    version: 1.3.2
+    version_base: '1.1'
+    cwd: /home/leonmkim/fish_leon/FISH
+    config_sources:
+    - path: hydra.conf
+      schema: pkg
+      provider: hydra
+    - path: /home/leonmkim/fish_leon/FISH/cfgs
+      schema: file
+      provider: main
+    - path: ''
+      schema: structured
+      provider: schema
+    output_dir: /home/leonmkim/fish_leon/FISH/exp_local/frankagym_pixels/FrankaInsertion-v1/1045_0/205131
+    choices:
+      suite: frankagym
+      suite/frankagym_task@_global_: insertion
+      agent: diffusion
+      hydra/env: default
+      hydra/callbacks: null
+      hydra/job_logging: default
+      hydra/hydra_logging: default
+      hydra/hydra_help: default
+      hydra/help: default
+      hydra/sweeper: basic
+      hydra/launcher: submitit_local
+      hydra/output: default
+  verbose: false

205131/.hydra/overrides.yaml

@@ -0,0 +1,3 @@
+- agent=diffusion
+- suite=frankagym
+- suite/frankagym_task@_global_=insertion

205131/eval_policy.log

@@ -0,0 +1,15 @@
+[2024-12-29 20:51:31,071][py.warnings][WARNING] - /home/leonmkim/fish_leon/FISH/eval_policy.py:428: UserWarning: 
+The version_base parameter is not specified.
+Please specify a compatability version level, or None.
+Will assume defaults for version 1.1
+  @hydra.main(config_path='cfgs', config_name='config_eval')
+
+[2024-12-29 20:51:31,075][py.warnings][WARNING] - /home/leonmkim/fish_leon/FISH/eval_policy.py:365: UserWarning: 
+The version_base parameter is not specified.
+Please specify a compatability version level, or None.
+Will assume defaults for version 1.1
+  hydra.initialize(
+
+[2024-12-29 20:51:34,320][py.warnings][WARNING] - /home/leonmkim/fish_leon/FISH/eval_policy.py:414: FutureWarning: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature.
+  payload = torch.load(f)
+

205132/.hydra/config.yaml

@@ -0,0 +1,350 @@
+root_dir: /home/${oc.env:USER}/fish_leon
+nstep: 3
+seed: 41
+dataset_shuffle_seed: ${seed}
+device: cuda
+save_video: true
+save_buffer: true
+use_tb: true
+baseline: false
+use_wandb: true
+eval: true
+process_contact_features: ${eval}
+obs_type: pixels
+use_color: true
+use_depth: true
+use_masks: false
+mask_list:
+- EE_obj_mask
+mask_representation: channels
+crop_hw:
+- 144
+- 144
+crop_down_offset: 48
+color_crop_type: null
+depth_crop_type: null
+segmask_crop_type: null
+add_crop_binary_mask: false
+add_coord_conv_map: false
+use_context_color: false
+use_context_depth: false
+use_context_segmask: false
+context_color_crop_type: null
+context_depth_crop_type: null
+context_segmask_crop_type: null
+context_add_crop_binary_mask: false
+context_add_coord_conv_map: false
+use_contact_map: false
+use_sdf_maps: false
+use_normals_maps: false
+which_objects: both
+max_contact_prob: 0.1
+max_depth: 2.0
+grasped_dtc_max_value: 0.105
+env_dtc_max_value: 0.425
+grasped_normals_mask_max_dtc_value: 0.105
+env_normals_mask_max_dtc_value: 0.425
+clamp_dtc: true
+dtc_adaptive_normalization: false
+mask_normals_within_sdf: true
+adaptive_normals_mask: true
+learnable_contact_preprocess_params: false
+contact_model_name: local_multitask_outhd64all_home_crop_h144w144d48_ctxt_seed_183386_epoch_9
+contact_estimation_model_ckpt_path: ~/fish_leon/contact_estimation/artifacts/175604_2/checkpoints/epoch=09-val_loss=0.00.ckpt
+num_eval: 5
+debug_timestamps: false
+open_loop: false
+action_trajectories: true
+stop_after_action: false
+interpolation_frequency: 25
+policy_frequency: 5
+wait_for_new_camera_frames: true
+random_start: false
+eval_starts: ${root_dir}/FISH/eval_starts/${suite.name}_${obs_type}/${task_name}
+train_demo_idxs_list_or_num: null
+num_valid_demos: null
+val_num_groups: 3
+name_of_expert_demo: 112_240x320_all_twodim_left_to_right_annotated_start_idx_5hz_zstd7_EE_pxl_coords_expert_demos_imp_act
+expert_dataset_dirpath: ${root_dir}/FISH/expert_demos/${suite.name}/${task_name}/${name_of_expert_demo}
+expert_dataset: ${expert_dataset_dirpath}/demos.zarr
+action_key: ${oc.if_else:${action_trajectories}, 'action_trajectory_${interpolation_frequency}hz',
+  'action'}
+semantic_demo_grouping_name: semantic_demo_grouping.yaml
+semantic_demo_grouping: ${expert_dataset_dirpath}/${semantic_demo_grouping_name}
+expert_dataset_config: ${expert_dataset_dirpath}/demo_config.yaml
+bc_regularize: false
+bc_weight_type: qfilter
+load_checkpoint: ${agent.load_checkpoint}
+wandb_run_id: '1045_0'
+true_action_history: false
+wandb_notes: null
+checkpoint_epoch: 12000
+load_residual_weight: false
+checkpoint_root_dir: /home/${oc.env:USER}/fish_leon/FISH
+checkpoint_weight_dir: ${checkpoint_root_dir}/exp_local/${suite.name}_${obs_type}/${task_name}/${wandb_run_id}
+residual_weight: ${root_dir}/FISH/weights/${suite.name}_${obs_type}/${task_name}/weight.pt
+experiment_dir: ./exp_local/${suite.name}_${obs_type}/${task_name}/${wandb_run_id}
+final_experiment_dir: ${experiment_dir}/${now:%H%M%S}
+agent:
+  _target_: agent.diffusion_policy.DiffusionPolicyAgent
+  name: diffusion_policy
+  load_checkpoint: ${eval}
+  device: ${device}
+  n_obs_steps: ${.config.policy_cfg.n_obs_steps}
+  suite_name: ${suite.name}
+  obs_type: ${obs_type}
+  enable_arm: ${eval}
+  enable_camera: ${eval}
+  use_tb: ${use_tb}
+  desired_image_shape:
+  - 13
+  - 180
+  - 240
+  orig_cam_shape:
+  - 3
+  - 240
+  - 320
+  config:
+    _target_: agent.diffusion_policy.DiffusionPolicyAgentConfig
+    compile: false
+    device: ${device}
+    cam_resize_shape: ${agent.desired_image_shape}
+    orig_cam_shape: ${agent.orig_cam_shape}
+    policy_frequency: ${policy_frequency}
+    interpolation_frequency: ${interpolation_frequency}
+    policy_cfg:
+      _target_: lerobot.common.policies.diffusion.configuration_diffusion.DiffusionConfig
+      n_obs_steps: 1
+      horizon: 36
+      n_action_steps: ${agent.config.policy_cfg.horizon}
+      input_shapes:
+        observation.image: ${agent.config.cam_resize_shape}
+        context_observation.image: ${agent.config.cam_resize_shape}
+        observation.state:
+        - 8
+        observation.action_history:
+        - 7
+      output_shapes:
+        action:
+        - 7
+      input_normalization_modes:
+        observation.image: mean_std
+        observation.state: min_max
+        observation.action_history: min_max
+      output_normalization_modes:
+        action: min_max
+      vision_backbone: resnet18
+      pretrained_backbone_weights: null
+      transforms:
+      - _target_: torchaug.transforms.RandomAffine
+        degrees:
+        - -5
+        - 5
+        translate:
+        - 0.05
+        - 0.05
+        batch_transform: true
+        num_chunks: -1
+        batch_inplace: true
+      - _target_: torchaug.transforms.RandomColorJitter
+        brightness: 0.3
+        contrast: 0.4
+        saturation: 0.5
+        hue: 0.08
+        batch_transform: true
+        num_chunks: -1
+        batch_inplace: true
+      use_group_norm: true
+      spatial_softmax_num_keypoints: 32
+      action_history_encoder_config:
+        _target_: lerobot.common.policies.diffusion.configuration_diffusion.Unet1dEncoderConfig
+        in_channels: 7
+        out_channels: 32
+        history_length: ${agent.config.policy_cfg.n_action_steps}
+        kernel_size: ${agent.config.policy_cfg.kernel_size}
+        downsample_kernel_size: 3
+        downsample_stride: 2
+        downsample_padding: 1
+      down_dims:
+      - 256
+      - 512
+      - 1024
+      kernel_size: 5
+      n_groups: 8
+      diffusion_step_embed_dim: 128
+      use_film_scale_modulation: true
+      noise_scheduler_type: DDIM
+      beta_schedule: squaredcos_cap_v2
+      beta_start: 0.0001
+      beta_end: 0.02
+      prediction_type: epsilon
+      clip_sample: true
+      clip_sample_range: 1.0
+      num_train_timesteps: 50
+      num_inference_steps: 10
+      do_mask_loss_for_padding: false
+    train_cfg:
+      _target_: utils.TrainConfig
+      lr: 0.0001
+      lr_scheduler: cosine
+      lr_warmup_steps: 500
+      adam_betas:
+      - 0.95
+      - 0.999
+      adam_eps: 1.0e-08
+      adam_weight_decay: 1.0e-06
+      grad_clip_norm: 10
+      offline_steps: ${num_train_frames_diffusion}
+      use_amp: true
+    observation_cfg:
+      _target_: agent.encoder.VisualFeatureSet
+      use_depth: ${use_depth}
+      use_color: ${use_color}
+      mask_input_dict:
+        _target_: agent.encoder.MaskInputDict
+        enable: ${use_masks}
+        representation: ${mask_representation}
+        mask_list: ${mask_list}
+      crop_input_config:
+        _target_: agent.encoder.CropInputConfig
+        color_crop_type: ${color_crop_type}
+        depth_crop_type: ${depth_crop_type}
+        segmask_crop_type: ${segmask_crop_type}
+        crop_hw: ${crop_hw}
+        crop_down_offset: ${crop_down_offset}
+        add_crop_binary_mask: ${add_crop_binary_mask}
+        add_coord_conv_map: ${add_coord_conv_map}
+      context_input_config:
+        _target_: agent.encoder.ContextInputConfig
+        use_color: ${use_context_color}
+        use_depth: ${use_context_depth}
+        mask_input_dict:
+          _target_: agent.encoder.MaskInputDict
+          enable: ${use_context_segmask}
+          representation: ${mask_representation}
+          mask_list: ${mask_list}
+        crop_input_config:
+          _target_: agent.encoder.CropInputConfig
+          color_crop_type: ${context_color_crop_type}
+          depth_crop_type: ${context_depth_crop_type}
+          segmask_crop_type: ${context_segmask_crop_type}
+          crop_hw: ${crop_hw}
+          crop_down_offset: ${crop_down_offset}
+          add_crop_binary_mask: ${context_add_crop_binary_mask}
+          add_coord_conv_map: ${context_add_coord_conv_map}
+      mask_soft_approx_scheduler_config:
+        _target_: agent.encoder.MaskSoftApproxSchedulerConfig
+        num_steps: 40000
+        initial_value: 10.0
+        final_value: 1000.0
+        interpolation_scheme: constant
+      use_contact_map: ${use_contact_map}
+      use_sdf_maps: ${use_sdf_maps}
+      use_normals_maps: ${use_normals_maps}
+      which_objects: ${which_objects}
+      grasped_dtc_max_value: ${grasped_dtc_max_value}
+      env_dtc_max_value: ${env_dtc_max_value}
+      grasped_normals_mask_max_dtc_value: ${grasped_normals_mask_max_dtc_value}
+      env_normals_mask_max_dtc_value: ${env_normals_mask_max_dtc_value}
+      clamp_dtc: ${clamp_dtc}
+      max_contact_prob: ${max_contact_prob}
+      mask_normals_within_sdf: ${mask_normals_within_sdf}
+      dtc_adaptive_normalization: ${dtc_adaptive_normalization}
+      adaptive_normals_mask: ${adaptive_normals_mask}
+      max_depth: ${max_depth}
+      image_shape: ${agent.desired_image_shape}
+      learnable_contact_preprocess_params: ${learnable_contact_preprocess_params}
+      learning_rate: ${agent.config.train_cfg.lr}
+      weight_decay: 0.0
+      contact_model_name: ${contact_model_name}
+      zero_centered: false
+suite:
+  suite: frankagym
+  name: frankagym
+  frame_stack: ${agent.n_obs_steps}
+  action_repeat: 1
+  discount: 0.99
+  hidden_dim: 1024
+  num_train_frames: 2010
+  num_seed_frames: 260
+  num_train_epochs: 5000
+  validate_every_epochs: 100
+  validate_diffusion_on_action_loss_every_epochs: 500
+  train_eval_diffusion_on_action_loss_every_epochs: 500
+  check_topk_every_epochs: 10
+  save_snapshot_every_epochs: 5000
+  eval_every_frames: 2000
+  num_eval_episodes: 5
+  save_snapshot: true
+  wait_for_user_to_start_episode: true
+  task_make_fn:
+    _target_: suite.frankagym.make
+    name: ${task_name}
+    height: 240
+    width: 320
+    frame_stack: ${suite.frame_stack}
+    action_repeat: ${suite.action_repeat}
+    seed: ${seed}
+    enable_arm: ${agent.enable_arm}
+    enable_gripper: ${enable_gripper}
+    start_with_gripper_open: ${start_with_gripper_open}
+    enable_camera: ${agent.enable_camera}
+    path_to_depth_extrinsics: ${path_to_depth_extrinsics}
+    contact_estimation_model_ckpt_path: ${contact_estimation_model_ckpt_path}
+    x_limit: ${x_limit}
+    y_limit: ${y_limit}
+    z_limit: ${z_limit}
+    device: ${device}
+    interpolation_frequency: ${interpolation_frequency}
+    policy_frequency: ${policy_frequency}
+    debug_timestamps: ${debug_timestamps}
+    stop_after_action: ${stop_after_action}
+    open_loop: ${open_loop}
+    wait_for_new_camera_frames: ${wait_for_new_camera_frames}
+    action_key: ${action_key}
+    action_trajectory_horizon: ${agent.config.policy_cfg.horizon}
+    action_trajectories: ${action_trajectories}
+    path_to_zarr_dataset: ${expert_dataset}
+    agent_policy_cfg: ???
+    true_action_history: ${true_action_history}
+num_train_frames_bc: 50000
+num_train_frames_drq: 1100000
+stddev_schedule_drq: linear(1.0,0.1,100000)
+task_name: FrankaInsertion-v1
+num_train_frames_vinn: 25000
+num_train_frames_diffusion: 1000000
+num_train_epochs_bc: 5000
+num_train_epochs_diffusion: 5000
+validate_every_epochs_bc: 5
+validate_every_epochs_diffusion: 25
+validate_diffusion_on_action_loss_every_epochs: 50
+train_eval_diffusion_on_action_loss_every_epochs: 500
+check_topk_every_epochs: 5
+check_topk_every_epochs_diffusion: ${validate_diffusion_on_action_loss_every_epochs}
+save_snapshot_every_epochs_diffusion: 5000
+x_limit:
+- 0.2
+- 0.7
+y_limit:
+- -0.4
+- 0.4
+z_limit:
+- -0.05
+- 0.55
+home_displacement:
+- 0.55
+- 0.0
+- 0.55
+- 180.0
+- 0.0
+- 0.0
+enable_gripper: true
+start_with_gripper_open: true
+offset_mask:
+- 1
+- 1
+- 1
+- 1
+- 1
+- 1
+path_to_depth_extrinsics: ~/fish_leon/FISH/cfgs/camera_poses/camera_poses_L515/20240904-122305/color_tf_world.npy

205132/.hydra/hydra.yaml

@@ -0,0 +1,169 @@
+hydra:
+  run:
+    dir: ${final_experiment_dir}
+  sweep:
+    dir: ${final_experiment_dir}
+    subdir: ${hydra.job.num}
+  launcher:
+    submitit_folder: ${final_experiment_dir}/.slurm
+    timeout_min: 60
+    cpus_per_task: null
+    gpus_per_node: null
+    tasks_per_node: 1
+    mem_gb: null
+    nodes: 1
+    name: ${hydra.job.name}
+    stderr_to_stdout: false
+    _target_: hydra_plugins.hydra_submitit_launcher.submitit_launcher.LocalLauncher
+  sweeper:
+    _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper
+    max_batch_size: null
+    params: null
+  help:
+    app_name: ${hydra.job.name}
+    header: '${hydra.help.app_name} is powered by Hydra.
+
+      '
+    footer: 'Powered by Hydra (https://hydra.cc)
+
+      Use --hydra-help to view Hydra specific help
+
+      '
+    template: '${hydra.help.header}
+
+      == Configuration groups ==
+
+      Compose your configuration from those groups (group=option)
+
+
+      $APP_CONFIG_GROUPS
+
+
+      == Config ==
+
+      Override anything in the config (foo.bar=value)
+
+
+      $CONFIG
+
+
+      ${hydra.help.footer}
+
+      '
+  hydra_help:
+    template: 'Hydra (${hydra.runtime.version})
+
+      See https://hydra.cc for more info.
+
+
+      == Flags ==
+
+      $FLAGS_HELP
+
+
+      == Configuration groups ==
+
+      Compose your configuration from those groups (For example, append hydra/job_logging=disabled
+      to command line)
+
+
+      $HYDRA_CONFIG_GROUPS
+
+
+      Use ''--cfg hydra'' to Show the Hydra config.
+
+      '
+    hydra_help: ???
+  hydra_logging:
+    version: 1
+    formatters:
+      simple:
+        format: '[%(asctime)s][HYDRA] %(message)s'
+    handlers:
+      console:
+        class: logging.StreamHandler
+        formatter: simple
+        stream: ext://sys.stdout
+    root:
+      level: INFO
+      handlers:
+      - console
+    loggers:
+      logging_example:
+        level: DEBUG
+    disable_existing_loggers: false
+  job_logging:
+    version: 1
+    formatters:
+      simple:
+        format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s'
+    handlers:
+      console:
+        class: logging.StreamHandler
+        formatter: simple
+        stream: ext://sys.stdout
+      file:
+        class: logging.FileHandler
+        formatter: simple
+        filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log
+    root:
+      level: INFO
+      handlers:
+      - console
+      - file
+    disable_existing_loggers: false
+  env: {}
+  mode: RUN
+  searchpath: []
+  callbacks: {}
+  output_subdir: .hydra
+  overrides:
+    hydra:
+    - hydra.mode=RUN
+    task:
+    - agent=diffusion
+    - suite=frankagym
+    - suite/frankagym_task@_global_=insertion
+  job:
+    name: eval_robot
+    chdir: true
+    override_dirname: agent=diffusion,suite/frankagym_task@_global_=insertion,suite=frankagym
+    id: ???
+    num: ???
+    config_name: config_eval
+    env_set: {}
+    env_copy: []
+    config:
+      override_dirname:
+        kv_sep: '='
+        item_sep: ','
+        exclude_keys: []
+  runtime:
+    version: 1.3.2
+    version_base: '1.1'
+    cwd: /home/leonmkim/fish_leon/FISH
+    config_sources:
+    - path: hydra.conf
+      schema: pkg
+      provider: hydra
+    - path: /home/leonmkim/fish_leon/FISH/cfgs
+      schema: file
+      provider: main
+    - path: ''
+      schema: structured
+      provider: schema
+    output_dir: /home/leonmkim/fish_leon/FISH/exp_local/frankagym_pixels/FrankaInsertion-v1/1045_0/205132
+    choices:
+      suite: frankagym
+      suite/frankagym_task@_global_: insertion
+      agent: diffusion
+      hydra/env: default
+      hydra/callbacks: null
+      hydra/job_logging: default
+      hydra/hydra_logging: default
+      hydra/hydra_help: default
+      hydra/help: default
+      hydra/sweeper: basic
+      hydra/launcher: submitit_local
+      hydra/output: default
+  verbose: false

205132/.hydra/overrides.yaml

@@ -0,0 +1,3 @@
+- agent=diffusion
+- suite=frankagym
+- suite/frankagym_task@_global_=insertion

205132/eval_robot.log

@@ -0,0 +1,13 @@
+[2024-12-29 20:51:32,787][py.warnings][WARNING] - /home/leonmkim/fish_leon/FISH/eval_robot.py:503: UserWarning: 
+The version_base parameter is not specified.
+Please specify a compatability version level, or None.
+Will assume defaults for version 1.1
+  @hydra.main(config_path='cfgs', config_name='config_eval')
+
+[2024-12-29 20:51:32,790][py.warnings][WARNING] - /home/leonmkim/fish_leon/FISH/eval_robot.py:439: UserWarning: 
+The version_base parameter is not specified.
+Please specify a compatability version level, or None.
+Will assume defaults for version 1.1
+  hydra.initialize(
+
+[2024-12-29 20:51:33,311][rospy.core][INFO] - signal_shutdown [atexit]

205317/.hydra/config.yaml

@@ -0,0 +1,350 @@
+root_dir: /home/${oc.env:USER}/fish_leon
+nstep: 3
+seed: 41
+dataset_shuffle_seed: ${seed}
+device: cuda
+save_video: true
+save_buffer: true
+use_tb: true
+baseline: false
+use_wandb: true
+eval: true
+process_contact_features: ${eval}
+obs_type: pixels
+use_color: true
+use_depth: true
+use_masks: false
+mask_list:
+- EE_obj_mask
+mask_representation: channels
+crop_hw:
+- 144
+- 144
+crop_down_offset: 48
+color_crop_type: null
+depth_crop_type: null
+segmask_crop_type: null
+add_crop_binary_mask: false
+add_coord_conv_map: false
+use_context_color: false
+use_context_depth: false
+use_context_segmask: false
+context_color_crop_type: null
+context_depth_crop_type: null
+context_segmask_crop_type: null
+context_add_crop_binary_mask: false
+context_add_coord_conv_map: false
+use_contact_map: false
+use_sdf_maps: false
+use_normals_maps: false
+which_objects: both
+max_contact_prob: 0.1
+max_depth: 2.0
+grasped_dtc_max_value: 0.105
+env_dtc_max_value: 0.425
+grasped_normals_mask_max_dtc_value: 0.105
+env_normals_mask_max_dtc_value: 0.425
+clamp_dtc: true
+dtc_adaptive_normalization: false
+mask_normals_within_sdf: true
+adaptive_normals_mask: true
+learnable_contact_preprocess_params: false
+contact_model_name: local_multitask_outhd64all_home_crop_h144w144d48_ctxt_seed_183386_epoch_9
+contact_estimation_model_ckpt_path: ~/fish_leon/contact_estimation/artifacts/175604_2/checkpoints/epoch=09-val_loss=0.00.ckpt
+num_eval: 5
+debug_timestamps: false
+open_loop: false
+action_trajectories: true
+stop_after_action: false
+interpolation_frequency: 25
+policy_frequency: 5
+wait_for_new_camera_frames: true
+random_start: false
+eval_starts: ${root_dir}/FISH/eval_starts/${suite.name}_${obs_type}/${task_name}
+train_demo_idxs_list_or_num: null
+num_valid_demos: null
+val_num_groups: 3
+name_of_expert_demo: 112_240x320_all_twodim_left_to_right_annotated_start_idx_5hz_zstd7_EE_pxl_coords_expert_demos_imp_act
+expert_dataset_dirpath: ${root_dir}/FISH/expert_demos/${suite.name}/${task_name}/${name_of_expert_demo}
+expert_dataset: ${expert_dataset_dirpath}/demos.zarr
+action_key: ${oc.if_else:${action_trajectories}, 'action_trajectory_${interpolation_frequency}hz',
+  'action'}
+semantic_demo_grouping_name: semantic_demo_grouping.yaml
+semantic_demo_grouping: ${expert_dataset_dirpath}/${semantic_demo_grouping_name}
+expert_dataset_config: ${expert_dataset_dirpath}/demo_config.yaml
+bc_regularize: false
+bc_weight_type: qfilter
+load_checkpoint: ${agent.load_checkpoint}
+wandb_run_id: '1045_0'
+true_action_history: false
+wandb_notes: null
+checkpoint_epoch: 12000
+load_residual_weight: false
+checkpoint_root_dir: /home/${oc.env:USER}/fish_leon/FISH
+checkpoint_weight_dir: ${checkpoint_root_dir}/exp_local/${suite.name}_${obs_type}/${task_name}/${wandb_run_id}
+residual_weight: ${root_dir}/FISH/weights/${suite.name}_${obs_type}/${task_name}/weight.pt
+experiment_dir: ./exp_local/${suite.name}_${obs_type}/${task_name}/${wandb_run_id}
+final_experiment_dir: ${experiment_dir}/${now:%H%M%S}
+agent:
+  _target_: agent.diffusion_policy.DiffusionPolicyAgent
+  name: diffusion_policy
+  load_checkpoint: ${eval}
+  device: ${device}
+  n_obs_steps: ${.config.policy_cfg.n_obs_steps}
+  suite_name: ${suite.name}
+  obs_type: ${obs_type}
+  enable_arm: ${eval}
+  enable_camera: ${eval}
+  use_tb: ${use_tb}
+  desired_image_shape:
+  - 13
+  - 180
+  - 240
+  orig_cam_shape:
+  - 3
+  - 240
+  - 320
+  config:
+    _target_: agent.diffusion_policy.DiffusionPolicyAgentConfig
+    compile: false
+    device: ${device}
+    cam_resize_shape: ${agent.desired_image_shape}
+    orig_cam_shape: ${agent.orig_cam_shape}
+    policy_frequency: ${policy_frequency}
+    interpolation_frequency: ${interpolation_frequency}
+    policy_cfg:
+      _target_: lerobot.common.policies.diffusion.configuration_diffusion.DiffusionConfig
+      n_obs_steps: 1
+      horizon: 36
+      n_action_steps: ${agent.config.policy_cfg.horizon}
+      input_shapes:
+        observation.image: ${agent.config.cam_resize_shape}
+        context_observation.image: ${agent.config.cam_resize_shape}
+        observation.state:
+        - 8
+        observation.action_history:
+        - 7
+      output_shapes:
+        action:
+        - 7
+      input_normalization_modes:
+        observation.image: mean_std
+        observation.state: min_max
+        observation.action_history: min_max
+      output_normalization_modes:
+        action: min_max
+      vision_backbone: resnet18
+      pretrained_backbone_weights: null
+      transforms:
+      - _target_: torchaug.transforms.RandomAffine
+        degrees:
+        - -5
+        - 5
+        translate:
+        - 0.05
+        - 0.05
+        batch_transform: true
+        num_chunks: -1
+        batch_inplace: true
+      - _target_: torchaug.transforms.RandomColorJitter
+        brightness: 0.3
+        contrast: 0.4
+        saturation: 0.5
+        hue: 0.08
+        batch_transform: true
+        num_chunks: -1
+        batch_inplace: true
+      use_group_norm: true
+      spatial_softmax_num_keypoints: 32
+      action_history_encoder_config:
+        _target_: lerobot.common.policies.diffusion.configuration_diffusion.Unet1dEncoderConfig
+        in_channels: 7
+        out_channels: 32
+        history_length: ${agent.config.policy_cfg.n_action_steps}
+        kernel_size: ${agent.config.policy_cfg.kernel_size}
+        downsample_kernel_size: 3
+        downsample_stride: 2
+        downsample_padding: 1
+      down_dims:
+      - 256
+      - 512
+      - 1024
+      kernel_size: 5
+      n_groups: 8
+      diffusion_step_embed_dim: 128
+      use_film_scale_modulation: true
+      noise_scheduler_type: DDIM
+      beta_schedule: squaredcos_cap_v2
+      beta_start: 0.0001
+      beta_end: 0.02
+      prediction_type: epsilon
+      clip_sample: true
+      clip_sample_range: 1.0
+      num_train_timesteps: 50
+      num_inference_steps: 10
+      do_mask_loss_for_padding: false
+    train_cfg:
+      _target_: utils.TrainConfig
+      lr: 0.0001
+      lr_scheduler: cosine
+      lr_warmup_steps: 500
+      adam_betas:
+      - 0.95
+      - 0.999
+      adam_eps: 1.0e-08
+      adam_weight_decay: 1.0e-06
+      grad_clip_norm: 10
+      offline_steps: ${num_train_frames_diffusion}
+      use_amp: true
+    observation_cfg:
+      _target_: agent.encoder.VisualFeatureSet
+      use_depth: ${use_depth}
+      use_color: ${use_color}
+      mask_input_dict:
+        _target_: agent.encoder.MaskInputDict
+        enable: ${use_masks}
+        representation: ${mask_representation}
+        mask_list: ${mask_list}
+      crop_input_config:
+        _target_: agent.encoder.CropInputConfig
+        color_crop_type: ${color_crop_type}
+        depth_crop_type: ${depth_crop_type}
+        segmask_crop_type: ${segmask_crop_type}
+        crop_hw: ${crop_hw}
+        crop_down_offset: ${crop_down_offset}
+        add_crop_binary_mask: ${add_crop_binary_mask}
+        add_coord_conv_map: ${add_coord_conv_map}
+      context_input_config:
+        _target_: agent.encoder.ContextInputConfig
+        use_color: ${use_context_color}
+        use_depth: ${use_context_depth}
+        mask_input_dict:
+          _target_: agent.encoder.MaskInputDict
+          enable: ${use_context_segmask}
+          representation: ${mask_representation}
+          mask_list: ${mask_list}
+        crop_input_config:
+          _target_: agent.encoder.CropInputConfig
+          color_crop_type: ${context_color_crop_type}
+          depth_crop_type: ${context_depth_crop_type}
+          segmask_crop_type: ${context_segmask_crop_type}
+          crop_hw: ${crop_hw}
+          crop_down_offset: ${crop_down_offset}
+          add_crop_binary_mask: ${context_add_crop_binary_mask}
+          add_coord_conv_map: ${context_add_coord_conv_map}
+      mask_soft_approx_scheduler_config:
+        _target_: agent.encoder.MaskSoftApproxSchedulerConfig
+        num_steps: 40000
+        initial_value: 10.0
+        final_value: 1000.0
+        interpolation_scheme: constant
+      use_contact_map: ${use_contact_map}
+      use_sdf_maps: ${use_sdf_maps}
+      use_normals_maps: ${use_normals_maps}
+      which_objects: ${which_objects}
+      grasped_dtc_max_value: ${grasped_dtc_max_value}
+      env_dtc_max_value: ${env_dtc_max_value}
+      grasped_normals_mask_max_dtc_value: ${grasped_normals_mask_max_dtc_value}
+      env_normals_mask_max_dtc_value: ${env_normals_mask_max_dtc_value}
+      clamp_dtc: ${clamp_dtc}
+      max_contact_prob: ${max_contact_prob}
+      mask_normals_within_sdf: ${mask_normals_within_sdf}
+      dtc_adaptive_normalization: ${dtc_adaptive_normalization}
+      adaptive_normals_mask: ${adaptive_normals_mask}
+      max_depth: ${max_depth}
+      image_shape: ${agent.desired_image_shape}
+      learnable_contact_preprocess_params: ${learnable_contact_preprocess_params}
+      learning_rate: ${agent.config.train_cfg.lr}
+      weight_decay: 0.0
+      contact_model_name: ${contact_model_name}
+      zero_centered: false
+suite:
+  suite: frankagym
+  name: frankagym
+  frame_stack: ${agent.n_obs_steps}
+  action_repeat: 1
+  discount: 0.99
+  hidden_dim: 1024
+  num_train_frames: 2010
+  num_seed_frames: 260
+  num_train_epochs: 5000
+  validate_every_epochs: 100
+  validate_diffusion_on_action_loss_every_epochs: 500
+  train_eval_diffusion_on_action_loss_every_epochs: 500
+  check_topk_every_epochs: 10
+  save_snapshot_every_epochs: 5000
+  eval_every_frames: 2000
+  num_eval_episodes: 5
+  save_snapshot: true
+  wait_for_user_to_start_episode: true
+  task_make_fn:
+    _target_: suite.frankagym.make
+    name: ${task_name}
+    height: 240
+    width: 320
+    frame_stack: ${suite.frame_stack}
+    action_repeat: ${suite.action_repeat}
+    seed: ${seed}
+    enable_arm: ${agent.enable_arm}
+    enable_gripper: ${enable_gripper}
+    start_with_gripper_open: ${start_with_gripper_open}
+    enable_camera: ${agent.enable_camera}
+    path_to_depth_extrinsics: ${path_to_depth_extrinsics}
+    contact_estimation_model_ckpt_path: ${contact_estimation_model_ckpt_path}
+    x_limit: ${x_limit}
+    y_limit: ${y_limit}
+    z_limit: ${z_limit}
+    device: ${device}
+    interpolation_frequency: ${interpolation_frequency}
+    policy_frequency: ${policy_frequency}
+    debug_timestamps: ${debug_timestamps}
+    stop_after_action: ${stop_after_action}
+    open_loop: ${open_loop}
+    wait_for_new_camera_frames: ${wait_for_new_camera_frames}
+    action_key: ${action_key}
+    action_trajectory_horizon: ${agent.config.policy_cfg.horizon}
+    action_trajectories: ${action_trajectories}
+    path_to_zarr_dataset: ${expert_dataset}
+    agent_policy_cfg: ???
+    true_action_history: ${true_action_history}
+num_train_frames_bc: 50000
+num_train_frames_drq: 1100000
+stddev_schedule_drq: linear(1.0,0.1,100000)
+task_name: FrankaInsertion-v1
+num_train_frames_vinn: 25000
+num_train_frames_diffusion: 1000000
+num_train_epochs_bc: 5000
+num_train_epochs_diffusion: 5000
+validate_every_epochs_bc: 5
+validate_every_epochs_diffusion: 25
+validate_diffusion_on_action_loss_every_epochs: 50
+train_eval_diffusion_on_action_loss_every_epochs: 500
+check_topk_every_epochs: 5
+check_topk_every_epochs_diffusion: ${validate_diffusion_on_action_loss_every_epochs}
+save_snapshot_every_epochs_diffusion: 5000
+x_limit:
+- 0.2
+- 0.7
+y_limit:
+- -0.4
+- 0.4
+z_limit:
+- -0.05
+- 0.55
+home_displacement:
+- 0.55
+- 0.0
+- 0.55
+- 180.0
+- 0.0
+- 0.0
+enable_gripper: true
+start_with_gripper_open: true
+offset_mask:
+- 1
+- 1
+- 1
+- 1
+- 1
+- 1
+path_to_depth_extrinsics: ~/fish_leon/FISH/cfgs/camera_poses/camera_poses_L515/20240904-122305/color_tf_world.npy

205317/.hydra/hydra.yaml

@@ -0,0 +1,169 @@
+hydra:
+  run:
+    dir: ${final_experiment_dir}
+  sweep:
+    dir: ${final_experiment_dir}
+    subdir: ${hydra.job.num}
+  launcher:
+    submitit_folder: ${final_experiment_dir}/.slurm
+    timeout_min: 60
+    cpus_per_task: null
+    gpus_per_node: null
+    tasks_per_node: 1
+    mem_gb: null
+    nodes: 1
+    name: ${hydra.job.name}
+    stderr_to_stdout: false
+    _target_: hydra_plugins.hydra_submitit_launcher.submitit_launcher.LocalLauncher
+  sweeper:
+    _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper
+    max_batch_size: null
+    params: null
+  help:
+    app_name: ${hydra.job.name}
+    header: '${hydra.help.app_name} is powered by Hydra.
+
+      '
+    footer: 'Powered by Hydra (https://hydra.cc)
+
+      Use --hydra-help to view Hydra specific help
+
+      '
+    template: '${hydra.help.header}
+
+      == Configuration groups ==
+
+      Compose your configuration from those groups (group=option)
+
+
+      $APP_CONFIG_GROUPS
+
+
+      == Config ==
+
+      Override anything in the config (foo.bar=value)
+
+
+      $CONFIG
+
+
+      ${hydra.help.footer}
+
+      '
+  hydra_help:
+    template: 'Hydra (${hydra.runtime.version})
+
+      See https://hydra.cc for more info.
+
+
+      == Flags ==
+
+      $FLAGS_HELP
+
+
+      == Configuration groups ==
+
+      Compose your configuration from those groups (For example, append hydra/job_logging=disabled
+      to command line)
+
+
+      $HYDRA_CONFIG_GROUPS
+
+
+      Use ''--cfg hydra'' to Show the Hydra config.
+
+      '
+    hydra_help: ???
+  hydra_logging:
+    version: 1
+    formatters:
+      simple:
+        format: '[%(asctime)s][HYDRA] %(message)s'
+    handlers:
+      console:
+        class: logging.StreamHandler
+        formatter: simple
+        stream: ext://sys.stdout
+    root:
+      level: INFO
+      handlers:
+      - console
+    loggers:
+      logging_example:
+        level: DEBUG
+    disable_existing_loggers: false
+  job_logging:
+    version: 1
+    formatters:
+      simple:
+        format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s'
+    handlers:
+      console:
+        class: logging.StreamHandler
+        formatter: simple
+        stream: ext://sys.stdout
+      file:
+        class: logging.FileHandler
+        formatter: simple
+        filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log
+    root:
+      level: INFO
+      handlers:
+      - console
+      - file
+    disable_existing_loggers: false
+  env: {}
+  mode: RUN
+  searchpath: []
+  callbacks: {}
+  output_subdir: .hydra
+  overrides:
+    hydra:
+    - hydra.mode=RUN
+    task:
+    - agent=diffusion
+    - suite=frankagym
+    - suite/frankagym_task@_global_=insertion
+  job:
+    name: eval_robot
+    chdir: true
+    override_dirname: agent=diffusion,suite/frankagym_task@_global_=insertion,suite=frankagym
+    id: ???
+    num: ???
+    config_name: config_eval
+    env_set: {}
+    env_copy: []
+    config:
+      override_dirname:
+        kv_sep: '='
+        item_sep: ','
+        exclude_keys: []
+  runtime:
+    version: 1.3.2
+    version_base: '1.1'
+    cwd: /home/leonmkim/fish_leon/FISH
+    config_sources:
+    - path: hydra.conf
+      schema: pkg
+      provider: hydra
+    - path: /home/leonmkim/fish_leon/FISH/cfgs
+      schema: file
+      provider: main
+    - path: ''
+      schema: structured
+      provider: schema
+    output_dir: /home/leonmkim/fish_leon/FISH/exp_local/frankagym_pixels/FrankaInsertion-v1/1045_0/205317
+    choices:
+      suite: frankagym
+      suite/frankagym_task@_global_: insertion
+      agent: diffusion
+      hydra/env: default
+      hydra/callbacks: null
+      hydra/job_logging: default
+      hydra/hydra_logging: default
+      hydra/hydra_help: default
+      hydra/help: default
+      hydra/sweeper: basic
+      hydra/launcher: submitit_local
+      hydra/output: default
+  verbose: false

205317/.hydra/overrides.yaml

@@ -0,0 +1,3 @@
+- agent=diffusion
+- suite=frankagym
+- suite/frankagym_task@_global_=insertion

205317/episode_rosbags/aligned_depth_to_color_K.npy

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4a962c703d20da282f5e009d432dff51df4ebd22f3386699b6754ea9cfc06a55
+size 200

205317/episode_rosbags/cam_tf_world.npy

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:29313ba240dd65ebbc79056d5582a97c9586cf2a1d4a1e0db13b87b49152cc9e
+size 256

205317/episode_rosbags/color_K.npy

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4a962c703d20da282f5e009d432dff51df4ebd22f3386699b6754ea9cfc06a55
+size 200

205317/episode_rosbags/depth_K.npy

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cc601d2fecd31c5513c76a88b5d4d8059adc1f92dad682e2d755d89a66d8fdf7
+size 200

205317/episode_rosbags/episode_0_2024-12-29-20-54-40.bag

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:263d6ba64875ad3e5c9380992b1ae9ac5466595aba489e10a184139ca4012da9
+size 1329494909

205317/episode_rosbags/episode_1_2024-12-29-20-55-29.bag

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6a5a4219be5162013cee691ff311f9a727cad82b88d356453a7fc8c13f6ad650
+size 1329599493

205317/episode_rosbags/episode_2_2024-12-29-20-56-10.bag

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:78194c48348d8759d08fca04bd5f3115792307c2b9618c258e792bad42c075f3
+size 1333012907

205317/episode_rosbags/episode_3_2024-12-29-20-56-49.bag

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:078503c64c23429f55328a6c2c550475f7561dca7fcf620a4e33a7346816978e
+size 1327707902

205317/episode_rosbags/episode_4_2024-12-29-20-57-31.bag

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:975a676b531d6e8139daea375e7609373d248fda8540766565f7604c286a86e3
+size 1333042300

205317/eval_robot.log

@@ -0,0 +1,12 @@
+[2024-12-29 20:53:17,250][py.warnings][WARNING] - /home/leonmkim/fish_leon/FISH/eval_robot.py:503: UserWarning: 
+The version_base parameter is not specified.
+Please specify a compatability version level, or None.
+Will assume defaults for version 1.1
+  @hydra.main(config_path='cfgs', config_name='config_eval')
+
+[2024-12-29 20:53:17,254][py.warnings][WARNING] - /home/leonmkim/fish_leon/FISH/eval_robot.py:439: UserWarning: 
+The version_base parameter is not specified.
+Please specify a compatability version level, or None.
+Will assume defaults for version 1.1
+  hydra.initialize(
+

205317/eval_video/0_eval.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2dd247d1c4c04fa7cea7a03c76f9bcf0fac708360da8e1551f468b1b47fc66b9
+size 1171585

205317/eval_video/1_eval.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f35843717fd499551c4b0926cca736afc0945a08b7388201dffb14db7ec3fefa
+size 1166652

205317/eval_video/2_eval.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c03a74d7b4bcab766fc29ee131b9e9ac693b435c897bde76d32747b19fb58257
+size 1184865

205317/eval_video/3_eval.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:94417ff346f07f5fa23e9c53f6b0137c96c29dcb3d88023f1ed9c85959c133e3
+size 1210929

205317/eval_video/4_eval.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:20ea9dbf78cd7cd19362e85b72b43f16e349294c086d44f5e3adc1fd73292d66
+size 1225749

205317/tb/events.out.tfevents.1735523604.leonmkim-ROG-Strix-G15CS-G15CS.1355277.0

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f35fcfee4a918882d02da32f4394ab050801d8e7f028ae8272dd24cece614507
+size 1103

205317/wandb/debug.log

@@ -0,0 +1,31 @@
+2024-12-29 20:53:23,743 INFO    MainThread:1355277 [wandb_setup.py:_flush():76] Current SDK version is 0.17.5
+2024-12-29 20:53:23,743 INFO    MainThread:1355277 [wandb_setup.py:_flush():76] Configure stats pid to 1355277
+2024-12-29 20:53:23,743 INFO    MainThread:1355277 [wandb_setup.py:_flush():76] Loading settings from /home/leonmkim/.config/wandb/settings
+2024-12-29 20:53:23,743 INFO    MainThread:1355277 [wandb_setup.py:_flush():76] Loading settings from /home/leonmkim/fish_leon/FISH/exp_local/frankagym_pixels/FrankaInsertion-v1/1045_0/205317/wandb/settings
+2024-12-29 20:53:23,743 INFO    MainThread:1355277 [wandb_setup.py:_flush():76] Loading settings from environment variables: {}
+2024-12-29 20:53:23,743 INFO    MainThread:1355277 [wandb_setup.py:_flush():76] Applying setup settings: {'_disable_service': False}
+2024-12-29 20:53:23,743 INFO    MainThread:1355277 [wandb_setup.py:_flush():76] Inferring run settings from compute environment: {'program_relpath': 'FISH/eval_robot.py', 'program_abspath': '/home/leonmkim/fish_leon/FISH/eval_robot.py', 'program': '/home/leonmkim/fish_leon/FISH/eval_robot.py'}
+2024-12-29 20:53:23,743 INFO    MainThread:1355277 [wandb_setup.py:_flush():76] Applying login settings: {}
+2024-12-29 20:53:23,743 INFO    MainThread:1355277 [wandb_init.py:_log_setup():529] Logging user logs to /home/leonmkim/fish_leon/FISH/exp_local/frankagym_pixels/FrankaInsertion-v1/1045_0/205317/wandb/run-20241229_205323-2n31umej/logs/debug.log
+2024-12-29 20:53:23,743 INFO    MainThread:1355277 [wandb_init.py:_log_setup():530] Logging internal logs to /home/leonmkim/fish_leon/FISH/exp_local/frankagym_pixels/FrankaInsertion-v1/1045_0/205317/wandb/run-20241229_205323-2n31umej/logs/debug-internal.log
+2024-12-29 20:53:23,744 INFO    MainThread:1355277 [wandb_init.py:init():569] calling init triggers
+2024-12-29 20:53:23,744 INFO    MainThread:1355277 [wandb_init.py:init():576] wandb.init called with sweep_config: {}
+config: {'root_dir': '/home/leonmkim/fish_leon', 'replay_buffer_size': 150000, 'replay_buffer_num_workers': 2, 'nstep': 3, 'batch_size': 128, 'seed': 0, 'dataset_shuffle_seed': 5, 'device': 'cuda', 'save_video': True, 'save_train_video': True, 'use_tb': True, 'use_wandb': True, 'wandb_run_id': '1045_0', 'wandb_notes': '1045_0_', 'eval': True, 'true_action_history': False, 'train_pad_after': 4, 'process_contact_features': True, 'obs_type': 'pixels', 'use_color': True, 'use_depth': True, 'use_masks': False, 'mask_list': ['EE_obj_mask'], 'mask_representation': 'channels', 'crop_hw': [144, 144], 'crop_down_offset': 48, 'color_crop_type': None, 'depth_crop_type': None, 'segmask_crop_type': None, 'add_crop_binary_mask': False, 'add_coord_conv_map': False, 'use_context_color': False, 'use_context_depth': False, 'use_context_segmask': False, 'context_color_crop_type': None, 'context_depth_crop_type': None, 'context_segmask_crop_type': None, 'context_add_crop_binary_mask': False, 'context_add_coord_conv_map': False, 'use_contact_map': False, 'use_sdf_maps': False, 'use_normals_maps': False, 'which_objects': 'both', 'max_contact_prob': 0.1, 'max_depth': 2.0, 'grasped_dtc_max_value': 0.2, 'env_dtc_max_value': 0.4, 'grasped_normals_mask_max_dtc_value': 0.2, 'env_normals_mask_max_dtc_value': 0.4, 'clamp_dtc': True, 'dtc_adaptive_normalization': False, 'mask_normals_within_sdf': True, 'adaptive_normals_mask': True, 'learnable_contact_preprocess_params': True, 'contact_model_name': 'local_multitask_outhd64all_home_crop_h144w144d48_mask_ctxtmask_seed_220979_epoch_9', 'contact_estimation_model_ckpt_path': '~/fish_leon/contact_estimation/artifacts/175604_2/checkpoints/epoch=09-val_loss=0.00.ckpt', 'encoder_type': 'small', 'debug_timestamps': False, 'open_loop': False, 'action_trajectories': True, 'stop_after_action': False, 'interpolation_frequency': 25, 'policy_frequency': 5, 'wait_for_new_camera_frames': True, 'baseline': False, 'train_demo_idxs_list_or_num': -1, 'log_train_every_steps': 25, 'name_of_expert_demo': '120_240x320_all_twodim_left_to_right_annotated_start_idx_5hz_zstd7_EE_pxl_coords_expert_demos_imp_act', 'expert_dataset_dirpath': '/home/leonmkim/fish_leon/FISH/expert_demos/frankagym/FrankaInsertion-v1/120_240x320_all_twodim_left_to_right_annotated_start_idx_5hz_zstd7_EE_pxl_coords_expert_demos_imp_act', 'store_dataset_in_memory': False, 'expert_dataset': '/home/leonmkim/fish_leon/FISH/expert_demos/frankagym/FrankaInsertion-v1/120_240x320_all_twodim_left_to_right_annotated_start_idx_5hz_zstd7_EE_pxl_coords_expert_demos_imp_act/demos.zarr', 'action_key': 'action_trajectory_25hz', 'semantic_demo_grouping_name': 'semantic_demo_grouping.yaml', 'semantic_demo_grouping': '/home/leonmkim/fish_leon/FISH/expert_demos/frankagym/FrankaInsertion-v1/120_240x320_all_twodim_left_to_right_annotated_start_idx_5hz_zstd7_EE_pxl_coords_expert_demos_imp_act/semantic_demo_grouping.yaml', 'include_groups_list': ['greece_twodim_nominal', 'greece_twodim_recovery'], 'expert_dataset_config': '/home/leonmkim/fish_leon/FISH/expert_demos/frankagym/FrankaInsertion-v1/120_240x320_all_twodim_left_to_right_annotated_start_idx_5hz_zstd7_EE_pxl_coords_expert_demos_imp_act/demo_config.yaml', 'name_of_valid_demo': '120_240x320_all_twodim_left_to_right_annotated_start_idx_5hz_zstd7_EE_pxl_coords_expert_demos_imp_act', 'valid_dataset_dir': '/home/leonmkim/fish_leon/FISH/expert_demos/frankagym/FrankaInsertion-v1/120_240x320_all_twodim_left_to_right_annotated_start_idx_5hz_zstd7_EE_pxl_coords_expert_demos_imp_act/demos.zarr', 'valid_demo_idxs_list_or_num': None, 'val_num_groups': 0, 'load_bc': True, 'checkpoint_epoch_list': [99, 199, 299, 399, 499, 599, 699, 799, 899, 999, 1249, 1499, 1749, 1999, 2999, 3999, 4999, 5999, 6999, 7999, 8999, 9999], 'snapshot_root_dir': '/mnt/grasp_high_usage/leonmkim/contact_estimation/FISH', 'save_snapshot': True, 'save_last_snapshot': True, 'save_snapshot_when_done': True, 'top_k_checkpoints': 5, 'save_snapshot_link_to_weights_dir': 'deprecated', 'bc_regularize': False, 'bc_weight_type': 'qfilter', 'experiment_dir': './exp_local/frankagym_pixels/FrankaInsertion-v1/1045_0', 'agent': {'_target_': 'agent.diffusion_policy.DiffusionPolicyAgent', 'name': 'diffusion_policy', 'load_checkpoint': True, 'device': 'cuda', 'n_obs_steps': 1, 'suite_name': 'frankagym', 'obs_type': 'pixels', 'enable_arm': True, 'enable_camera': True, 'use_tb': True, 'desired_image_shape': [13, 180, 240], 'orig_cam_shape': [3, 240, 320], 'config': {'_target_': 'agent.diffusion_policy.DiffusionPolicyAgentConfig', 'compile': False, 'device': 'cuda', 'cam_resize_shape': [13, 180, 240], 'orig_cam_shape': [3, 240, 320], 'policy_frequency': 5, 'interpolation_frequency': 25, 'policy_cfg': {'_target_': 'lerobot.common.policies.diffusion.configuration_diffusion.DiffusionConfig', 'n_obs_steps': 1, 'horizon': 36, 'n_action_steps': 36, 'output_shapes': {'action': [7]}, 'input_normalization_modes': {'observation.image': 'mean_std', 'observation.state': 'min_max', 'observation.action_history': 'min_max'}, 'output_normalization_modes': {'action': 'min_max'}, 'vision_backbone': 'resnet18', 'pretrained_backbone_weights': None, 'transforms': [{'_target_': 'torchaug.transforms.RandomAffine', 'degrees': [-5, 5], 'translate': [0.05, 0.05], 'batch_transform': True, 'num_chunks': -1, 'batch_inplace': True}, {'_target_': 'torchaug.transforms.RandomColorJitter', 'brightness': 0.3, 'contrast': 0.4, 'saturation': 0.5, 'hue': 0.08, 'batch_transform': True, 'num_chunks': -1, 'batch_inplace': True}], 'use_group_norm': True, 'spatial_softmax_num_keypoints': 32, 'action_history_encoder_config': {'_target_': 'lerobot.common.policies.diffusion.configuration_diffusion.Unet1dEncoderConfig', 'in_channels': 7, 'out_channels': 32, 'history_length': 6, 'kernel_size': 5, 'downsample_kernel_size': 3, 'downsample_stride': 2, 'downsample_padding': 1}, 'down_dims': [256, 512, 1024], 'kernel_size': 5, 'n_groups': 8, 'diffusion_step_embed_dim': 128, 'use_film_scale_modulation': True, 'noise_scheduler_type': 'DDIM', 'beta_schedule': 'squaredcos_cap_v2', 'beta_start': 0.0001, 'beta_end': 0.02, 'prediction_type': 'epsilon', 'clip_sample': True, 'clip_sample_range': 1.0, 'num_train_timesteps': 50, 'num_inference_steps': 10, 'do_mask_loss_for_padding': False, 'input_shapes': {'observation.image': [13, 180, 240], 'context_observation.image': [13, 180, 240], 'observation.state': [8], 'observation.action_history': [7]}}, 'train_cfg': {'_target_': 'utils.TrainConfig', 'lr': 0.0001, 'lr_scheduler': 'cosine', 'lr_warmup_steps': 500, 'adam_betas': [0.95, 0.999], 'adam_eps': 1e-08, 'adam_weight_decay': 1e-06, 'grad_clip_norm': 10, 'offline_steps': 1000000, 'use_amp': True}, 'observation_cfg': {'_target_': 'agent.encoder.VisualFeatureSet', 'use_depth': True, 'use_color': True, 'mask_input_dict': {'_target_': 'agent.encoder.MaskInputDict', 'enable': False, 'representation': 'channels', 'mask_list': ['EE_obj_mask']}, 'crop_input_config': {'_target_': 'agent.encoder.CropInputConfig', 'color_crop_type': None, 'depth_crop_type': None, 'segmask_crop_type': None, 'crop_hw': [144, 144], 'crop_down_offset': 48, 'add_crop_binary_mask': False, 'add_coord_conv_map': False}, 'context_input_config': {'_target_': 'agent.encoder.ContextInputConfig', 'use_color': False, 'use_depth': False, 'mask_input_dict': {'_target_': 'agent.encoder.MaskInputDict', 'enable': False, 'representation': 'channels', 'mask_list': ['EE_obj_mask']}, 'crop_input_config': {'_target_': 'agent.encoder.CropInputConfig', 'color_crop_type': None, 'depth_crop_type': None, 'segmask_crop_type': None, 'crop_hw': [144, 144], 'crop_down_offset': 48, 'add_crop_binary_mask': False, 'add_coord_conv_map': False}}, 'mask_soft_approx_scheduler_config': {'_target_': 'agent.encoder.MaskSoftApproxSchedulerConfig', 'num_steps': 40000, 'initial_value': 10.0, 'final_value': 1000.0, 'interpolation_scheme': 'cosine'}, 'use_contact_map': False, 'use_sdf_maps': False, 'use_normals_maps': False, 'which_objects': 'both', 'grasped_dtc_max_value': 0.2, 'env_dtc_max_value': 0.4, 'grasped_normals_mask_max_dtc_value': 0.2, 'env_normals_mask_max_dtc_value': 0.4, 'clamp_dtc': True, 'max_contact_prob': 0.1, 'mask_normals_within_sdf': True, 'dtc_adaptive_normalization': False, 'adaptive_normals_mask': True, 'max_depth': 2.0, 'image_shape': [13, 180, 240], 'learnable_contact_preprocess_params': True, 'learning_rate': 0.0001, 'weight_decay': 0.0, 'contact_model_name': 'local_multitask_outhd64all_home_crop_h144w144d48_mask_ctxtmask_seed_220979_epoch_9', 'zero_centered': False}}}, 'suite': {'suite': 'frankagym', 'name': 'frankagym', 'frame_stack': 1, 'action_repeat': 1, 'discount': 0.99, 'hidden_dim': 1024, 'num_train_frames': 2010, 'num_seed_frames': 260, 'num_train_epochs': 5000, 'validate_every_epochs': 100, 'validate_diffusion_on_action_loss_every_epochs': 500, 'train_eval_diffusion_on_action_loss_every_epochs': 500, 'check_topk_every_epochs': 10, 'save_snapshot_every_epochs': 5000, 'eval_every_frames': 2000, 'num_eval_episodes': 5, 'save_snapshot': True, 'wait_for_user_to_start_episode': True, 'task_make_fn': {'_target_': 'suite.frankagym.make', 'name': 'FrankaInsertion-v1', 'height': 240, 'width': 320, 'frame_stack': 1, 'action_repeat': 1, 'seed': 0, 'enable_arm': True, 'enable_gripper': True, 'start_with_gripper_open': True, 'enable_camera': True, 'path_to_depth_extrinsics': '~/fish_leon/FISH/cfgs/camera_poses/camera_poses_L515/20240904-122305/color_tf_world.npy', 'contact_estimation_model_ckpt_path': '~/fish_leon/contact_estimation/artifacts/175604_2/checkpoints/epoch=09-val_loss=0.00.ckpt', 'x_limit': [0.2, 0.7], 'y_limit': [-0.4, 0.4], 'z_limit': [-0.05, 0.55], 'device': 'cuda', 'interpolation_frequency': 25, 'policy_frequency': 5, 'debug_timestamps': False, 'stop_after_action': False, 'open_loop': False, 'wait_for_new_camera_frames': True, 'action_key': 'action_trajectory_25hz', 'action_trajectory_horizon': 36, 'action_trajectories': True, 'path_to_zarr_dataset': '/home/leonmkim/fish_leon/FISH/expert_demos/frankagym/FrankaInsertion-v1/120_240x320_all_twodim_left_to_right_annotated_start_idx_5hz_zstd7_EE_pxl_coords_expert_demos_imp_act/demos.zarr', 'agent_policy_cfg': {'_target_': 'agent.diffusion_policy.DiffusionPolicyAgentConfig', 'compile': False, 'device': 'cuda', 'cam_resize_shape': [13, 180, 240], 'orig_cam_shape': [3, 240, 320], 'policy_frequency': 5, 'interpolation_frequency': 25, 'policy_cfg': {'_target_': 'lerobot.common.policies.diffusion.configuration_diffusion.DiffusionConfig', 'n_obs_steps': 1, 'horizon': 36, 'n_action_steps': 36, 'output_shapes': {'action': [7]}, 'input_normalization_modes': {'observation.image': 'mean_std', 'observation.state': 'min_max', 'observation.action_history': 'min_max'}, 'output_normalization_modes': {'action': 'min_max'}, 'vision_backbone': 'resnet18', 'pretrained_backbone_weights': None, 'transforms': [{'_target_': 'torchaug.transforms.RandomAffine', 'degrees': [-5, 5], 'translate': [0.05, 0.05], 'batch_transform': True, 'num_chunks': -1, 'batch_inplace': True}, {'_target_': 'torchaug.transforms.RandomColorJitter', 'brightness': 0.3, 'contrast': 0.4, 'saturation': 0.5, 'hue': 0.08, 'batch_transform': True, 'num_chunks': -1, 'batch_inplace': True}], 'use_group_norm': True, 'spatial_softmax_num_keypoints': 32, 'action_history_encoder_config': {'_target_': 'lerobot.common.policies.diffusion.configuration_diffusion.Unet1dEncoderConfig', 'in_channels': 7, 'out_channels': 32, 'history_length': 6, 'kernel_size': 5, 'downsample_kernel_size': 3, 'downsample_stride': 2, 'downsample_padding': 1}, 'down_dims': [256, 512, 1024], 'kernel_size': 5, 'n_groups': 8, 'diffusion_step_embed_dim': 128, 'use_film_scale_modulation': True, 'noise_scheduler_type': 'DDIM', 'beta_schedule': 'squaredcos_cap_v2', 'beta_start': 0.0001, 'beta_end': 0.02, 'prediction_type': 'epsilon', 'clip_sample': True, 'clip_sample_range': 1.0, 'num_train_timesteps': 50, 'num_inference_steps': 10, 'do_mask_loss_for_padding': False, 'input_shapes': {'observation.image': [13, 180, 240], 'context_observation.image': [13, 180, 240], 'observation.state': [8], 'observation.action_history': [7]}}, 'train_cfg': {'_target_': 'utils.TrainConfig', 'lr': 0.0001, 'lr_scheduler': 'cosine', 'lr_warmup_steps': 500, 'adam_betas': [0.95, 0.999], 'adam_eps': 1e-08, 'adam_weight_decay': 1e-06, 'grad_clip_norm': 10, 'offline_steps': 1000000, 'use_amp': True}, 'observation_cfg': {'_target_': 'agent.encoder.VisualFeatureSet', 'use_depth': True, 'use_color': True, 'mask_input_dict': {'_target_': 'agent.encoder.MaskInputDict', 'enable': False, 'representation': 'channels', 'mask_list': ['EE_obj_mask']}, 'crop_input_config': {'_target_': 'agent.encoder.CropInputConfig', 'color_crop_type': None, 'depth_crop_type': None, 'segmask_crop_type': None, 'crop_hw': [144, 144], 'crop_down_offset': 48, 'add_crop_binary_mask': False, 'add_coord_conv_map': False}, 'context_input_config': {'_target_': 'agent.encoder.ContextInputConfig', 'use_color': False, 'use_depth': False, 'mask_input_dict': {'_target_': 'agent.encoder.MaskInputDict', 'enable': False, 'representation': 'channels', 'mask_list': ['EE_obj_mask']}, 'crop_input_config': {'_target_': 'agent.encoder.CropInputConfig', 'color_crop_type': None, 'depth_crop_type': None, 'segmask_crop_type': None, 'crop_hw': [144, 144], 'crop_down_offset': 48, 'add_crop_binary_mask': False, 'add_coord_conv_map': False}}, 'mask_soft_approx_scheduler_config': {'_target_': 'agent.encoder.MaskSoftApproxSchedulerConfig', 'num_steps': 40000, 'initial_value': 10.0, 'final_value': 1000.0, 'interpolation_scheme': 'cosine'}, 'use_contact_map': False, 'use_sdf_maps': False, 'use_normals_maps': False, 'which_objects': 'both', 'grasped_dtc_max_value': 0.2, 'env_dtc_max_value': 0.4, 'grasped_normals_mask_max_dtc_value': 0.2, 'env_normals_mask_max_dtc_value': 0.4, 'clamp_dtc': True, 'max_contact_prob': 0.1, 'mask_normals_within_sdf': True, 'dtc_adaptive_normalization': False, 'adaptive_normals_mask': True, 'max_depth': 2.0, 'image_shape': [13, 180, 240], 'learnable_contact_preprocess_params': True, 'learning_rate': 0.0001, 'weight_decay': 0.0, 'contact_model_name': 'local_multitask_outhd64all_home_crop_h144w144d48_mask_ctxtmask_seed_220979_epoch_9', 'zero_centered': False}}, 'true_action_history': False}}, 'num_train_frames_bc': 50000, 'num_train_frames_drq': 1100000, 'stddev_schedule_drq': 'linear(1.0,0.1,100000)', 'task_name': 'FrankaInsertion-v1', 'num_train_frames_vinn': 25000, 'num_train_frames_diffusion': 1000000, 'num_train_epochs_bc': 5000, 'num_train_epochs_diffusion': 15000, 'validate_every_epochs_bc': 5, 'validate_every_epochs_diffusion': 250, 'validate_diffusion_on_action_loss_every_epochs': 250, 'train_eval_diffusion_on_action_loss_every_epochs': 250, 'check_topk_every_epochs': 5, 'check_topk_every_epochs_diffusion': 250, 'save_snapshot_every_epochs_diffusion': 1500, 'x_limit': [0.2, 0.7], 'y_limit': [-0.4, 0.4], 'z_limit': [-0.05, 0.55], 'home_displacement': [0.55, 0.0, 0.55, 180.0, 0.0, 0.0], 'enable_gripper': True, 'start_with_gripper_open': True, 'offset_mask': [1, 1, 1, 1, 1, 1], 'path_to_depth_extrinsics': '~/fish_leon/FISH/cfgs/camera_poses/camera_poses_L515/20240904-122305/color_tf_world.npy', 'feature_type': '180x240_1_RGB_D_2.0_acthst_hst6_out32_dwnkrnl3_dwnstrd2_dwnpd1', 'save_buffer': True, 'num_eval': 5, 'random_start': False, 'eval_starts': '/home/leonmkim/fish_leon/FISH/eval_starts/frankagym_pixels/FrankaInsertion-v1', 'num_valid_demos': None, 'load_checkpoint': True, 'checkpoint_epoch': 12000, 'load_residual_weight': False, 'checkpoint_root_dir': '/home/leonmkim/fish_leon/FISH', 'checkpoint_weight_dir': '/home/leonmkim/fish_leon/FISH/exp_local/frankagym_pixels/FrankaInsertion-v1/1045_0', 'residual_weight': '/home/leonmkim/fish_leon/FISH/weights/frankagym_pixels/FrankaInsertion-v1/weight.pt', 'final_experiment_dir': './exp_local/frankagym_pixels/FrankaInsertion-v1/1045_0/205317'}
+2024-12-29 20:53:23,744 INFO    MainThread:1355277 [wandb_init.py:init():619] starting backend
+2024-12-29 20:53:23,744 INFO    MainThread:1355277 [wandb_init.py:init():623] setting up manager
+2024-12-29 20:53:23,748 INFO    MainThread:1355277 [backend.py:_multiprocessing_setup():105] multiprocessing start_methods=fork,spawn,forkserver, using: spawn
+2024-12-29 20:53:23,750 INFO    MainThread:1355277 [wandb_init.py:init():631] backend started and connected
+2024-12-29 20:53:23,761 INFO    MainThread:1355277 [wandb_init.py:init():720] updated telemetry
+2024-12-29 20:53:23,767 INFO    MainThread:1355277 [wandb_init.py:init():753] communicating run to backend with 90.0 second timeout
+2024-12-29 20:53:24,005 INFO    MainThread:1355277 [wandb_run.py:_on_init():2435] communicating current version
+2024-12-29 20:53:24,117 INFO    MainThread:1355277 [wandb_run.py:_on_init():2444] got version response upgrade_message: "wandb version 0.19.1 is available!  To upgrade, please run:\n $ pip install wandb --upgrade"
+
+2024-12-29 20:53:24,117 INFO    MainThread:1355277 [wandb_init.py:init():804] starting run threads in backend
+2024-12-29 20:53:24,460 INFO    MainThread:1355277 [wandb_run.py:_console_start():2413] atexit reg
+2024-12-29 20:53:24,460 INFO    MainThread:1355277 [wandb_run.py:_redirect():2255] redirect: wrap_raw
+2024-12-29 20:53:24,460 INFO    MainThread:1355277 [wandb_run.py:_redirect():2320] Wrapping output streams.
+2024-12-29 20:53:24,460 INFO    MainThread:1355277 [wandb_run.py:_redirect():2345] Redirects installed.
+2024-12-29 20:53:24,461 INFO    MainThread:1355277 [wandb_init.py:init():847] run started, returning control to user process
+2024-12-29 20:53:24,462 INFO    MainThread:1355277 [wandb_run.py:_tensorboard_callback():1544] tensorboard callback: /home/leonmkim/fish_leon/FISH/exp_local/frankagym_pixels/FrankaInsertion-v1/1045_0/205317/tb, True
+2024-12-29 20:53:30,487 INFO    MainThread:1355277 [wandb_run.py:_config_callback():1382] config_cb None None {'grasped_obj_name': 'greece', 'left_book_slot': 'twodim'}
+2024-12-29 20:58:20,684 WARNING MsgRouterThr:1355277 [router.py:message_loop():77] message_loop has been closed

205317/wandb/run-20241229_205323-2n31umej/files/code/FISH/eval_robot.py

@@ -0,0 +1,512 @@
+#%%
+import warnings
+import os
+
+os.environ['MKL_SERVICE_FORCE_INTEL'] = '1'
+os.environ['MUJOCO_GL'] = 'egl'
+from pathlib import Path
+#%%
+import hydra
+import numpy as np
+import torch
+
+import utils
+from utils import get_feature_dirname_from_configs
+
+from video import VideoRecorder
+import pickle
+import time
+import threading
+import shutil
+from logger import Logger
+
+import wandb 
+from omegaconf import OmegaConf, open_dict
+
+from replay_buffer_robot import RosbagEvalReplayBufferStorage
+from lerobot.common.utils.utils import _relative_path_between
+
+torch.backends.cudnn.benchmark = True
+warnings.filterwarnings('ignore', category=DeprecationWarning)
+
+# import specs for replay buffer
+from dm_env import specs
+
+import sys, signal
+import yaml
+
+# get path of current file
+current_path = os.path.dirname(os.path.realpath(__file__))
+sys.path.append(os.path.join(current_path, os.pardir))
+# from contact_estimation.src.utils.viz_utils import normalized_surface_normal_to_rgb, depth_map_to_im, grasped_env_dtc_map_to_im, contact_prob_map_to_im, desaturate_color_image, masked_overlay_im_list
+
+def make_agent(obs_spec, action_spec, cfg):
+	cfg.obs_shape = obs_spec['pixels'].shape
+	dataset_statistics = None # this will be loaded from the checkpoint
+	try:
+		cfg.action_shape = action_spec.shape
+	except:
+		pass
+	return hydra.utils.instantiate(cfg, dataset_statistics)
+
+class Workspace:
+	def __init__(self, cfg):
+		self.work_dir = Path.cwd()
+		print(f'workspace: {self.work_dir}')
+		
+		signal.signal(signal.SIGINT, self.signal_handler)
+		
+		self.cfg = cfg
+		self.loading_uncompiled_checkpoint_with_compile = False
+		self.loading_compiled_checkpoint_with_no_compile = False
+
+		snapshot_path = Path(self.cfg.checkpoint_weight_dir) / f'snapshot_{self.cfg.checkpoint_epoch}.pt' 
+		self.load_checkpoint_conf(snapshot_path=snapshot_path)
+
+		# load config for action trajectories
+		utils.set_seed_everywhere(self.cfg.seed)
+		self.device = torch.device(self.cfg.device)
+		self.setup()
+
+		# self.agent = make_agent(self.eval_env.observation_spec(),
+		# 						self.eval_env.action_spec(), self.cfg.agent)
+		self.timer = utils.Timer()
+		# self._global_step = 0
+		self._global_episode = 0
+		self._global_epoch = 0
+		self.num_episode_successes = 0
+
+		# Need to convert hydra config to primitive container for wandb https://docs.wandb.ai/guides/integrations/hydra
+		with open_dict(self.cfg):
+			self.cfg.feature_type = get_feature_dirname_from_configs(
+				hydra.utils.instantiate(self.cfg.agent.config.observation_cfg), 
+				self.cfg.agent.config.policy_cfg.input_shapes,
+				hydra.utils.instantiate(self.cfg.agent.config.policy_cfg.action_history_encoder_config) if 'observation.action_history' in self.cfg.agent.config.policy_cfg.input_shapes else None,
+			)
+
+		wandb_config = OmegaConf.to_container(
+			self.cfg, resolve=True, throw_on_missing=True
+		)
+		# must be called before any tf summary writer is created
+		if self.cfg.use_wandb:
+			wandb.init(project='extrinsic_contact_downstream', entity='serialexperimentsleon', job_type='eval', sync_tensorboard=self.cfg.use_tb, config=wandb_config)
+
+		self.logger = Logger(self.work_dir, use_tb=self.cfg.use_tb, use_wandb=self.cfg.use_wandb)
+
+		# if not self.loading_uncompiled_checkpoint_with_compile and self.cfg.agent.config.compile:
+		# 	self.agent.compile_modules()
+			
+		# self.load_checkpoint(snapshot_path=snapshot_path)
+
+		# if self.loading_uncompiled_checkpoint_with_compile: # need to call compile after loading the checkpoint
+		# 	self.agent.compile_modules()
+
+		print(f"loaded agent with feature_type: {self.cfg.feature_type}")
+
+	def check_for_key_press(self):
+		while self.continue_keypress_thread:
+			inp = input("Press 'r' to restart current episode, 'n' to stop current episode and skip to next, 'q' to break entire eval\n")
+			if inp == 'n':
+				self.preempt_episode = True
+				print("preempting episode")
+			elif inp in ['', '0', '1']: # enter key
+				if inp in ['0', '1']:
+					self.num_episode_successes += int(inp)
+				self.proceed_after_env_reset_event.set()
+				print("proceeding to start episode!")
+			elif inp == 'q':
+				self.proceed_after_env_reset_event.set()
+				self.preempt_episode = True
+				self.exit_eval = True
+				self.continue_keypress_thread = False # will stop the keypress thread
+				print("quitting eval")
+				break
+			elif inp == 'r':
+				print('restarting episode')
+				self.preempt_episode = True
+				self.restart_episode = True
+			else:
+				print("Invalid key press, try again")
+		
+		# self.keypress_input_thread.join() # wait for the keypress thread to finish
+
+	def signal_handler(self, signal, frame):
+		print("\nprogram exiting gracefully")
+		self.proceed_after_env_reset_event.set()
+		self.preempt_episode = True
+		self.exit_eval = True
+		self.continue_keypress_thread = False # will stop the keypress thread
+		self.keypress_input_thread.join() # wait for the keypress thread to finish
+		video_filepath = self.video_recorder.save()
+		# get the video file and convert to video tensor to log 
+		self.logger.log_video('eval/video', video_filepath, self.global_step)
+		sys.exit(0)
+		
+	def setup(self):
+		# create envs
+		self.eval_env = hydra.utils.call(self.cfg.suite.task_make_fn)
+		# expert_demo_config_path = os.path.join(os.path.dirname(self.cfg.expert_dataset), 'demo_config.yaml')
+		# self.expert_demo_config = yaml.load(open(expert_demo_config_path, 'r'), Loader=yaml.FullLoader)
+		# self.eval_env._env.action_trans_norm = expert_demo_config['max_translation_action_norm']
+		# self.eval_env._env.action_rot_norm = expert_demo_config['max_rotation_action_norm']
+		# self.eval_env._env.action_period = expert_demo_config['sample_period']
+		# print(f"setting max_translation_action_norm to {expert_demo_config['max_translation_action_norm']} and sample_period to {expert_demo_config['sample_period']}")
+		# print(f"setting max_rotation_action_norm to {expert_demo_config['max_rotation_action_norm']}")
+
+		# self.eval_env.set_demo_params(self.cfg.expert_dataset)
+
+		# Turn off random start
+		self.eval_env.random_start = False
+
+		# create replay buffer
+		# data_specs = [
+		# 	{
+		# 		'observation': self.eval_env.observation_spec(),
+		# 	},
+		# 	# self.eval_env.observation_spec()['features'],
+		# 	self.eval_env.action_spec(),
+		# 	specs.Array(self.eval_env.action_spec().shape, self.eval_env.action_spec().dtype, 'vinn_action'), 
+		# 	specs.Array((1, ), np.float32, 'reward'),
+		# 	specs.Array((1, ), np.float32, 'discount'),
+		# ]
+
+		# self.eval_replay_storage = ZarrEvalReplayBufferStorage(data_specs, self.work_dir / 'eval_buffer', debug_timestamps=self.cfg.debug_timestamps, save_buffer=self.cfg.save_buffer, debug_info_data_specs=self.eval_env.debug_info_data_specs, camera_info_dict=self.eval_env.get_camera_info_dict())
+		self.eval_replay_storage = RosbagEvalReplayBufferStorage(self.work_dir)
+
+		self.video_recorder = VideoRecorder(
+			self.work_dir if self.cfg.save_video else None,
+			ros_enabled=True,
+			fps=self.cfg.agent.config.policy_frequency,
+			)
+		
+		print('workspace setup complete')
+		
+	@property
+	def global_step(self):
+		# return self._global_step
+		return self.eval_env.get_global_step()
+
+	@property
+	def global_episode(self):
+		return self._global_episode
+
+	@property
+	def global_frame(self):
+		return self.global_step * self.cfg.action_repeat
+
+	@property
+	def global_epoch(self):
+		return self._global_epoch
+	
+	def reset(self, eval_idx):
+		if not self.eval_env.enable_arm:
+			return np.array([0,0,0], dtype=np.float32)
+		self.eval_env.arm_refresh(reset=False)
+		# Set start position
+		try:
+			self.eval_env.set_position(self.start_pos[eval_idx])
+		except:
+			self.eval_env.arm.set_position(self.start_pos[eval_idx])
+			if self.eval_env.arm.keep_gripper_closed:
+				self.eval_env.arm.close_gripper_fully()
+			else:
+				self.eval_env.arm.open_gripper_fully()
+		time.sleep(0.1)		
+		time_step = self.eval_env.step(np.zeros(self.eval_env.action_spec().shape[0], dtype=np.float32),
+									   np.zeros(self.eval_env.action_spec().shape[0], dtype=np.float32))	
+		return time_step
+
+	def eval(self):
+		# before evals start, prompt user for name of grasped object and the left book of the slot location
+		grasped_obj_name = input("Enter the name of the grasped object: ")
+		left_book_slot = input("Enter the left book slot location: ")
+		# update wandb config
+		if self.cfg.use_wandb:
+			wandb.config.update({'grasped_obj_name': grasped_obj_name, 'left_book_slot': left_book_slot})
+
+		self.preempt_episode = False
+		self.exit_eval = False
+		self.restart_episode = False
+
+		self.continue_keypress_thread = True
+		self.proceed_after_env_reset_event = threading.Event()
+		self.keypress_input_thread = threading.Thread(target=self.check_for_key_press)
+		self.keypress_input_thread.start()
+
+		# # Set model to eval mode
+		# self.agent.train(False)
+		
+		eval_until_episode = utils.Until(self.cfg.num_eval)
+
+		self.use_action_history = False
+		# if "dp" in repr(self.agent) and "observation.action_history" in self.cfg.agent.config.policy_cfg.input_shapes:
+		if "observation.action_history" in self.cfg.agent.config.policy_cfg.input_shapes:
+			self.use_action_history = True
+
+		# self.eval_replay_storage._new_eval_step(0)
+
+		# if 'vinn' in repr(self.agent) or 'openloop' in repr(self.agent):
+		# 	with open(self.cfg.expert_dataset, 'rb') as f:
+		# 		if self.cfg.obs_type == 'pixels':
+		# 			self.expert_demo, _, self.expert_action, self.expert_reward = pickle.load(f)
+		# 		elif self.cfg.obs_type == 'features':
+		# 			_, self.expert_demo, self.expert_action, self.expert_reward = pickle.load(f)
+				
+		# 		if self.cfg.action_trajectories:
+		# 			with open(self.cfg.expert_action_trajectories, 'rb') as f:
+		# 				self.expert_action = pickle.load(f)
+
+		# 		if isinstance(self.cfg.train_demo_idxs_list_or_num, int):
+		# 			if self.cfg.train_demo_idxs_list_or_num == -1:
+		# 				self.cfg.train_demo_idxs_list_or_num = len(self.expert_demo)
+		# 			train_demo_idxs_list_or_num = list(range(self.cfg.train_demo_idxs_list_or_num))
+				
+		# 		self.expert_demo = self.expert_demo[train_demo_idxs_list_or_num]
+		# 		self.expert_action = self.expert_action[train_demo_idxs_list_or_num]
+		# 		self.expert_reward = self.expert_reward[train_demo_idxs_list_or_num]
+		# 		# if self.cfg.action_plans:
+		# 		# 	self.expert_action_plans = self.expert_action_plans[self.cfg.train_demo_idxs_list_or_num]
+		# 		# self.expert_demo = self.expert_demo[:self.cfg.num_demos]
+		# 		# self.expert_action = self.expert_action[:self.cfg.num_demos]
+		# 		# self.expert_reward = self.expert_reward[:self.cfg.num_demos]
+
+		# 		self.expert_demo = np.concatenate(self.expert_demo, axis=0)
+		# 		self.expert_rgb_obs = np.ascontiguousarray(np.transpose(self.expert_demo, (0,2,3,1))[:, :,:,:3].astype(np.uint8))
+		# 		self.expert_action = np.concatenate(self.expert_action, axis=0)
+				
+		# 		self.agent.save_representations(self.expert_demo, self.expert_action, 128, config=self.expert_demo_config)
+		
+		# Get start points
+		if self.cfg.random_start:
+			eval_starts = Path(self.cfg.eval_starts) / 'starts.pkl'
+			if eval_starts.exists():
+				with eval_starts.open('rb') as f:
+					self.start_pos = pickle.load(f)
+			else:
+				eval_starts = Path(self.cfg.eval_starts)
+				eval_starts.mkdir(parents=True, exist_ok=True)
+				
+				# Generate start points
+				self.start_pos = []
+				try:
+					for _ in range(self.cfg.num_eval):
+						self.start_pos.append(self.eval_env.get_random_pos())
+				except:
+					for _ in range(self.cfg.num_eval):
+						self.start_pos.append(self.eval_env.arm.get_random_pos())
+
+				# Save start points for the task
+				eval_starts = eval_starts / 'starts.pkl'
+				with eval_starts.open('wb') as f:
+					pickle.dump(self.start_pos, f)
+
+		time_step = self.eval_env.reset()
+		# replay_thread = None
+		while eval_until_episode(self.global_episode) and not self.exit_eval:
+			# self.video_recorder.init(self.eval_env, video_filename=f'{self.global_episode}_eval.mp4')
+			print(f"Starting episode {self.global_episode}")
+			time_step = self.eval_env.reset() #Leon: need to call reset twice in case objects are trapped
+			self.video_recorder.init(self.eval_env, video_filename=f'{self.global_episode}_eval.mp4')
+			# x = input("Press Enter to continue... after reseting env")
+			print("Press Enter to continue... after reseting env. To rate prev episode, press 0 for failure and 1 for success")
+			self.proceed_after_env_reset_event.clear() # clear the event flag
+			self.proceed_after_env_reset_event.wait() # blocking wait for the event flag to be set
+			if self.global_episode > 0:
+				self.logger.log_metrics({'num_success': self.num_episode_successes}, self.global_step, 'eval', episode=self.global_episode)
+				self.logger.log_metrics({'success_rate': self.num_episode_successes/self.global_episode}, self.global_step, 'eval', episode=self.global_episode)
+			time_step = self.eval_env.reset()
+			# debug_info_dict = self.eval_env.debug_info_dict
+			# if replay_thread is not None:
+			# 	# wait for the last replay thread to finish
+			# 	replay_thread.join()
+
+			# self.eval_replay_storage.add(time_step._replace(observation=time_step.observation[self.cfg.obs_type]), debug_info_dict)
+			# replay_thread = threading.Thread(target=self.eval_replay_storage.add, args=(time_step._replace(observation=time_step.observation[self.cfg.obs_type]), debug_info_dict))
+			# replay_thread = threading.Thread(target=self.eval_replay_storage.add, args=(time_step, debug_info_dict))
+
+			# replay_thread.start()
+			if self.cfg.random_start:
+				time_step = self.reset(self.global_episode)
+				time.sleep(2) #5)
+			# if 'vinn' in repr(self.agent):
+			# 	self.agent.reset()
+			# 	# self.agent.buffer.reset()
+			# 	# if self.cfg.open_loop:
+			# 	# 	self.agent.current_step = 0
+			# if 'openloop' in repr(self.agent):
+			# 	self.agent.curr_step = 0
+			# at start of each episode, provide zero action for policies that use action history
+			# shape should be (T_o, T_a, action_dim)
+
+			# while not time_step.last() and not self.preempt_episode:
+			self.video_recorder.ros_start_recording()
+			self.eval_replay_storage.start_episode()
+			self.eval_env.start_policy_timer()
+			while not self.eval_env.episode_done() and not self.preempt_episode:
+				# with torch.no_grad(), utils.eval_mode(self.agent):
+				# 	# if self.cfg.agent.provide_topk:
+				# 	# 	action, vinn_action, topk = self.agent.act(
+				# 	# 		time_step.observation['pixels'],
+				# 	# 		self.global_step,
+				# 	# 		eval_mode=True)
+				# 	# elif self.cfg.agent.provide_obs:
+				# 	# 	action, vinn_action, obs = self.agent.act(
+				# 	# 		time_step.observation['pixels'],
+				# 	# 		self.global_step,
+				# 	# 		eval_mode=True)
+				# 	# else:
+				# 	action, vinn_action = self.agent.act(
+				# 		time_step.observation,
+				# 		self.global_step,
+				# 		eval_mode=True,
+				# 		obs_timestamp=time_step.observation['timestamp'],
+				# 		obs_seq=time_step.observation['seq'],
+				# 		action_history=action_history,
+				# 		action_history_start_timestamp=action_history_start_timestamp,
+				# 		)
+					# DONT WAIT FOR POLICY TO GET AN ACTION 
+					# we dont want to slow down grabbing obs and passing to sam/contact features
+
+				self.eval_env.run_policy_threads() # this just does a rospy sleep
+					
+				# if self.use_action_history:
+				# 	action_history_start_timestamp = time_step.observation['timestamp']
+				# 	# action_history = action[:self.cfg.agent.config.policy_cfg.action_history_encoder_config.history_length, ...]		
+				# 	# add n_obs_steps dimension to action_history, for now we assume n_obs_steps = 1
+				# 	# TODO: handle n_obs_steps > 1
+				# 	action_history = action[np.newaxis, ...]
+
+				# time_step = self.eval_env.step(action, vinn_action) # obs, reward after action has been taken
+				# debug_info_dict = self.eval_env.debug_info_dict
+				
+				# time_step = self.eval_env.ros_step()
+				
+				# replay_thread.join()
+
+				# time how long it takes to execute the step
+				# time_before_add = time.perf_counter()
+				# self.eval_replay_storage.add(time_step._replace(observation=time_step.observation[self.cfg.obs_type]), debug_info_dict)
+				# use thread to call the add function in a separate thread
+				# replay_thread = threading.Thread(target=self.eval_replay_storage.add, args=(time_step._replace(observation=time_step.observation[self.cfg.obs_type]), debug_info_dict))
+
+				# replay_thread = threading.Thread(target=self.eval_replay_storage.add, args=(time_step, debug_info_dict))
+				# replay_thread.start()
+				
+				# print(f"Time to add to replay buffer: {time.perf_counter() - time_before_add}")
+
+				# self.video_recorder.record(self.eval_env)
+				# self._global_step += 1
+			
+			self.eval_env.stop_policy_timer()
+			
+			if self.restart_episode:
+				# means we should delete the current episode and start again
+				self.restart_episode = False
+				self.eval_replay_storage.reset_current_episode()
+				self.video_recorder.reset_current_episode()
+
+			else:
+				self.eval_replay_storage.store_current_episode()
+				video_filepath = self.video_recorder.save()
+				self.logger.log_video(f"eval/{video_filepath.name.rstrip('.mp4')}", video_filepath, self.global_step)
+				self._global_episode += 1
+			
+			self.preempt_episode = False # reset preempt_episode flag
+		
+		# self.video_recorder.save(f'{episode}_eval.mp4')
+		# get the video file and convert to video tensor to log 
+
+		self.eval_env.reset()
+		
+		print("Evaluation finished. To wrap up, rate prev episode, press 0 for failure and 1 for success")
+		self.proceed_after_env_reset_event.clear() # clear the event flag
+		self.proceed_after_env_reset_event.wait() # blocking wait for the event flag to be set
+		if self.global_episode > 0:
+			# self.logger.log_metrics({'num_success': self.num_episode_successes}, self.global_step, 'eval', episode=self.global_episode)
+			self.logger.log_metrics({'num_success': self.num_episode_successes}, self.global_step, 'eval', episode=self.global_episode)
+			self.logger.log_metrics({'success_rate': self.num_episode_successes/self.global_episode}, self.global_step, 'eval', episode=self.global_episode)
+			
+		self.continue_keypress_thread = False # will stop the keypress thread
+		self.keypress_input_thread.join() # wait for the keypress thread to finish
+
+	def load_checkpoint_conf(self, snapshot_path):
+		config_path = snapshot_path.parent / 'config.yaml'
+		if not config_path.exists():
+			raise FileNotFoundError(f'No snapshot conf found at {config_path}')
+		else:
+			# load the omegaconf config
+			hydra.core.global_hydra.GlobalHydra.instance().clear()
+			hydra.initialize(
+				str(_relative_path_between(Path(config_path).absolute().parent, Path(__file__).absolute().parent)),
+			)
+			cfg = hydra.compose(Path(config_path).stem)
+			from deepdiff import DeepDiff
+			from omegaconf import open_dict
+			diff = DeepDiff(OmegaConf.to_container(cfg), OmegaConf.to_container(self.cfg)) # old, new
+			# import re
+			overwriteable_keys = [f"root{overwritable_key}" for overwritable_key in ["['use_wandb']", "['path_to_depth_extrinsics']", "['eval']", "['root_dir']", "['wandb_notes']", "['agent']['config']['train_cfg']['use_amp']", "['agent']['config']['compile']", "['agent']['config']['policy_cfg']['num_inference_steps']"]]
+			if "values_changed" in diff:
+				# top_k_checkpoints, wandb_notes, agent.config.train_cfg.use_amp, save_snapshot_every_epochs_diffusion, check_topk_every_epochs_diffusion, validate_diffusion_on_action_loss_every_epochs, train_eval_diffusion_on_action_loss_every_epochs, validate_every_epochs_diffusion
+				# for keys above, overwrite the old config with the new config
+				for k, v in diff['values_changed'].items():
+					# replace any keys that are under "root['suite']"
+					if k in overwriteable_keys or k.startswith("root['suite']"):
+						print(f"Found changed key {k} with value {v}. Overwriting old checkpoint config")
+						if k == "root['agent']['config']['compile']":
+							if diff['values_changed'][k]['new_value']:
+								self.loading_uncompiled_checkpoint_with_compile = True
+							elif not diff['values_changed'][k]['new_value']:
+								# raise ValueError("Cannot load a compiled checkpoint without compile")
+								self.loading_compiled_checkpoint_with_no_compile = True
+						exec(f"{k.replace('root[', 'cfg[')} = {k.replace('root[', 'self.cfg[')}")
+			# for any new values, update the old checkpoint config
+			if "dictionary_item_added" in diff:
+				for new_key in diff['dictionary_item_added']: # this is a list
+					# if new_key == "root['suite']['task_make_fn']['observation_cfg']":
+					if new_key == "root['suite']['task_make_fn']['agent_policy_cfg']":
+						# pass the agents observation_cfg to the suite task_make_fn
+						with open_dict(cfg): # to allow addition of non-existing keys
+							# cfg.suite.task_make_fn.observation_cfg = cfg.agent.config.observation_cfg
+							cfg.suite.task_make_fn.agent_policy_cfg = cfg.agent.config
+						continue
+					elif "['agent']['config']['policy_cfg']['input_shapes']" in new_key:
+						# skip adding the new key if it is the input_shapes of the policy_cfg
+						continue
+					else:
+						print(f"Found new key {new_key} with value {eval(new_key.replace('root[', 'self.cfg['))}. Adding to checkpoint config")
+					# eval(new_key.replace('root', 'cfg')) = eval(new_key.replace('root', 'self.cfg'))
+					if new_key == "root['agent']['config']['compile']":
+						if self.cfg.agent.config.compile:
+							self.loading_uncompiled_checkpoint_with_compile = True
+					
+					with open_dict(cfg):
+						exec(f"{new_key.replace('root[', 'cfg[')}={new_key.replace('root[', 'self.cfg[')}")
+			self.cfg = cfg
+
+	def load_checkpoint(self, snapshot_path, bc=False):
+		print(f'resuming {repr(self.agent)}: {snapshot_path}')
+		with snapshot_path.open('rb') as f:
+			payload = torch.load(f)
+		agent_payload = {}
+		for k, v in payload.items():
+			if k not in self.__dict__:
+				agent_payload[k] = v
+			elif k == '_global_epoch':
+				self._global_epoch = v
+				print(f'loaded epoch: {v}')
+				if self.cfg.use_wandb:
+					# add to config of wandb
+					wandb.config.update({'epoch': v})
+
+		# self.agent.load_snapshot_eval(agent_payload, bc)
+
+@hydra.main(config_path='cfgs', config_name='config_eval')
+def main(cfg):
+	from eval_robot import Workspace as W
+	root_dir = Path.cwd()
+	workspace = W(cfg)
+
+	workspace.eval()
+
+if __name__ == '__main__':
+	main()

205317/wandb/run-20241229_205323-2n31umej/files/config.yaml

@@ -0,0 +1,895 @@
+wandb_version: 1
+
+root_dir:
+  desc: null
+  value: /home/leonmkim/fish_leon
+replay_buffer_size:
+  desc: null
+  value: 150000
+replay_buffer_num_workers:
+  desc: null
+  value: 2
+nstep:
+  desc: null
+  value: 3
+batch_size:
+  desc: null
+  value: 128
+seed:
+  desc: null
+  value: 0
+dataset_shuffle_seed:
+  desc: null
+  value: 5
+device:
+  desc: null
+  value: cuda
+save_video:
+  desc: null
+  value: true
+save_train_video:
+  desc: null
+  value: true
+use_tb:
+  desc: null
+  value: true
+use_wandb:
+  desc: null
+  value: true
+wandb_run_id:
+  desc: null
+  value: '1045_0'
+wandb_notes:
+  desc: null
+  value: '1045_0_'
+eval:
+  desc: null
+  value: true
+true_action_history:
+  desc: null
+  value: false
+train_pad_after:
+  desc: null
+  value: 4
+process_contact_features:
+  desc: null
+  value: true
+obs_type:
+  desc: null
+  value: pixels
+use_color:
+  desc: null
+  value: true
+use_depth:
+  desc: null
+  value: true
+use_masks:
+  desc: null
+  value: false
+mask_list:
+  desc: null
+  value:
+  - EE_obj_mask
+mask_representation:
+  desc: null
+  value: channels
+crop_hw:
+  desc: null
+  value:
+  - 144
+  - 144
+crop_down_offset:
+  desc: null
+  value: 48
+color_crop_type:
+  desc: null
+  value: null
+depth_crop_type:
+  desc: null
+  value: null
+segmask_crop_type:
+  desc: null
+  value: null
+add_crop_binary_mask:
+  desc: null
+  value: false
+add_coord_conv_map:
+  desc: null
+  value: false
+use_context_color:
+  desc: null
+  value: false
+use_context_depth:
+  desc: null
+  value: false
+use_context_segmask:
+  desc: null
+  value: false
+context_color_crop_type:
+  desc: null
+  value: null
+context_depth_crop_type:
+  desc: null
+  value: null
+context_segmask_crop_type:
+  desc: null
+  value: null
+context_add_crop_binary_mask:
+  desc: null
+  value: false
+context_add_coord_conv_map:
+  desc: null
+  value: false
+use_contact_map:
+  desc: null
+  value: false
+use_sdf_maps:
+  desc: null
+  value: false
+use_normals_maps:
+  desc: null
+  value: false
+which_objects:
+  desc: null
+  value: both
+max_contact_prob:
+  desc: null
+  value: 0.1
+max_depth:
+  desc: null
+  value: 2.0
+grasped_dtc_max_value:
+  desc: null
+  value: 0.2
+env_dtc_max_value:
+  desc: null
+  value: 0.4
+grasped_normals_mask_max_dtc_value:
+  desc: null
+  value: 0.2
+env_normals_mask_max_dtc_value:
+  desc: null
+  value: 0.4
+clamp_dtc:
+  desc: null
+  value: true
+dtc_adaptive_normalization:
+  desc: null
+  value: false
+mask_normals_within_sdf:
+  desc: null
+  value: true
+adaptive_normals_mask:
+  desc: null
+  value: true
+learnable_contact_preprocess_params:
+  desc: null
+  value: true
+contact_model_name:
+  desc: null
+  value: local_multitask_outhd64all_home_crop_h144w144d48_mask_ctxtmask_seed_220979_epoch_9
+contact_estimation_model_ckpt_path:
+  desc: null
+  value: ~/fish_leon/contact_estimation/artifacts/175604_2/checkpoints/epoch=09-val_loss=0.00.ckpt
+encoder_type:
+  desc: null
+  value: small
+debug_timestamps:
+  desc: null
+  value: false
+open_loop:
+  desc: null
+  value: false
+action_trajectories:
+  desc: null
+  value: true
+stop_after_action:
+  desc: null
+  value: false
+interpolation_frequency:
+  desc: null
+  value: 25
+policy_frequency:
+  desc: null
+  value: 5
+wait_for_new_camera_frames:
+  desc: null
+  value: true
+baseline:
+  desc: null
+  value: false
+train_demo_idxs_list_or_num:
+  desc: null
+  value: -1
+log_train_every_steps:
+  desc: null
+  value: 25
+name_of_expert_demo:
+  desc: null
+  value: 120_240x320_all_twodim_left_to_right_annotated_start_idx_5hz_zstd7_EE_pxl_coords_expert_demos_imp_act
+expert_dataset_dirpath:
+  desc: null
+  value: /home/leonmkim/fish_leon/FISH/expert_demos/frankagym/FrankaInsertion-v1/120_240x320_all_twodim_left_to_right_annotated_start_idx_5hz_zstd7_EE_pxl_coords_expert_demos_imp_act
+store_dataset_in_memory:
+  desc: null
+  value: false
+expert_dataset:
+  desc: null
+  value: /home/leonmkim/fish_leon/FISH/expert_demos/frankagym/FrankaInsertion-v1/120_240x320_all_twodim_left_to_right_annotated_start_idx_5hz_zstd7_EE_pxl_coords_expert_demos_imp_act/demos.zarr
+action_key:
+  desc: null
+  value: action_trajectory_25hz
+semantic_demo_grouping_name:
+  desc: null
+  value: semantic_demo_grouping.yaml
+semantic_demo_grouping:
+  desc: null
+  value: /home/leonmkim/fish_leon/FISH/expert_demos/frankagym/FrankaInsertion-v1/120_240x320_all_twodim_left_to_right_annotated_start_idx_5hz_zstd7_EE_pxl_coords_expert_demos_imp_act/semantic_demo_grouping.yaml
+include_groups_list:
+  desc: null
+  value:
+  - greece_twodim_nominal
+  - greece_twodim_recovery
+expert_dataset_config:
+  desc: null
+  value: /home/leonmkim/fish_leon/FISH/expert_demos/frankagym/FrankaInsertion-v1/120_240x320_all_twodim_left_to_right_annotated_start_idx_5hz_zstd7_EE_pxl_coords_expert_demos_imp_act/demo_config.yaml
+name_of_valid_demo:
+  desc: null
+  value: 120_240x320_all_twodim_left_to_right_annotated_start_idx_5hz_zstd7_EE_pxl_coords_expert_demos_imp_act
+valid_dataset_dir:
+  desc: null
+  value: /home/leonmkim/fish_leon/FISH/expert_demos/frankagym/FrankaInsertion-v1/120_240x320_all_twodim_left_to_right_annotated_start_idx_5hz_zstd7_EE_pxl_coords_expert_demos_imp_act/demos.zarr
+valid_demo_idxs_list_or_num:
+  desc: null
+  value: null
+val_num_groups:
+  desc: null
+  value: 0
+load_bc:
+  desc: null
+  value: true
+checkpoint_epoch_list:
+  desc: null
+  value:
+  - 99
+  - 199
+  - 299
+  - 399
+  - 499
+  - 599
+  - 699
+  - 799
+  - 899
+  - 999
+  - 1249
+  - 1499
+  - 1749
+  - 1999
+  - 2999
+  - 3999
+  - 4999
+  - 5999
+  - 6999
+  - 7999
+  - 8999
+  - 9999
+snapshot_root_dir:
+  desc: null
+  value: /mnt/grasp_high_usage/leonmkim/contact_estimation/FISH
+save_snapshot:
+  desc: null
+  value: true
+save_last_snapshot:
+  desc: null
+  value: true
+save_snapshot_when_done:
+  desc: null
+  value: true
+top_k_checkpoints:
+  desc: null
+  value: 5
+save_snapshot_link_to_weights_dir:
+  desc: null
+  value: deprecated
+bc_regularize:
+  desc: null
+  value: false
+bc_weight_type:
+  desc: null
+  value: qfilter
+experiment_dir:
+  desc: null
+  value: ./exp_local/frankagym_pixels/FrankaInsertion-v1/1045_0
+agent:
+  desc: null
+  value:
+    _target_: agent.diffusion_policy.DiffusionPolicyAgent
+    name: diffusion_policy
+    load_checkpoint: true
+    device: cuda
+    n_obs_steps: 1
+    suite_name: frankagym
+    obs_type: pixels
+    enable_arm: true
+    enable_camera: true
+    use_tb: true
+    desired_image_shape:
+    - 13
+    - 180
+    - 240
+    orig_cam_shape:
+    - 3
+    - 240
+    - 320
+    config:
+      _target_: agent.diffusion_policy.DiffusionPolicyAgentConfig
+      compile: false
+      device: cuda
+      cam_resize_shape:
+      - 13
+      - 180
+      - 240
+      orig_cam_shape:
+      - 3
+      - 240
+      - 320
+      policy_frequency: 5
+      interpolation_frequency: 25
+      policy_cfg:
+        _target_: lerobot.common.policies.diffusion.configuration_diffusion.DiffusionConfig
+        n_obs_steps: 1
+        horizon: 36
+        n_action_steps: 36
+        output_shapes:
+          action:
+          - 7
+        input_normalization_modes:
+          observation.image: mean_std
+          observation.state: min_max
+          observation.action_history: min_max
+        output_normalization_modes:
+          action: min_max
+        vision_backbone: resnet18
+        pretrained_backbone_weights: null
+        transforms:
+        - _target_: torchaug.transforms.RandomAffine
+          degrees:
+          - -5
+          - 5
+          translate:
+          - 0.05
+          - 0.05
+          batch_transform: true
+          num_chunks: -1
+          batch_inplace: true
+        - _target_: torchaug.transforms.RandomColorJitter
+          brightness: 0.3
+          contrast: 0.4
+          saturation: 0.5
+          hue: 0.08
+          batch_transform: true
+          num_chunks: -1
+          batch_inplace: true
+        use_group_norm: true
+        spatial_softmax_num_keypoints: 32
+        action_history_encoder_config:
+          _target_: lerobot.common.policies.diffusion.configuration_diffusion.Unet1dEncoderConfig
+          in_channels: 7
+          out_channels: 32
+          history_length: 6
+          kernel_size: 5
+          downsample_kernel_size: 3
+          downsample_stride: 2
+          downsample_padding: 1
+        down_dims:
+        - 256
+        - 512
+        - 1024
+        kernel_size: 5
+        n_groups: 8
+        diffusion_step_embed_dim: 128
+        use_film_scale_modulation: true
+        noise_scheduler_type: DDIM
+        beta_schedule: squaredcos_cap_v2
+        beta_start: 0.0001
+        beta_end: 0.02
+        prediction_type: epsilon
+        clip_sample: true
+        clip_sample_range: 1.0
+        num_train_timesteps: 50
+        num_inference_steps: 10
+        do_mask_loss_for_padding: false
+        input_shapes:
+          observation.image:
+          - 13
+          - 180
+          - 240
+          context_observation.image:
+          - 13
+          - 180
+          - 240
+          observation.state:
+          - 8
+          observation.action_history:
+          - 7
+      train_cfg:
+        _target_: utils.TrainConfig
+        lr: 0.0001
+        lr_scheduler: cosine
+        lr_warmup_steps: 500
+        adam_betas:
+        - 0.95
+        - 0.999
+        adam_eps: 1.0e-08
+        adam_weight_decay: 1.0e-06
+        grad_clip_norm: 10
+        offline_steps: 1000000
+        use_amp: true
+      observation_cfg:
+        _target_: agent.encoder.VisualFeatureSet
+        use_depth: true
+        use_color: true
+        mask_input_dict:
+          _target_: agent.encoder.MaskInputDict
+          enable: false
+          representation: channels
+          mask_list:
+          - EE_obj_mask
+        crop_input_config:
+          _target_: agent.encoder.CropInputConfig
+          color_crop_type: null
+          depth_crop_type: null
+          segmask_crop_type: null
+          crop_hw:
+          - 144
+          - 144
+          crop_down_offset: 48
+          add_crop_binary_mask: false
+          add_coord_conv_map: false
+        context_input_config:
+          _target_: agent.encoder.ContextInputConfig
+          use_color: false
+          use_depth: false
+          mask_input_dict:
+            _target_: agent.encoder.MaskInputDict
+            enable: false
+            representation: channels
+            mask_list:
+            - EE_obj_mask
+          crop_input_config:
+            _target_: agent.encoder.CropInputConfig
+            color_crop_type: null
+            depth_crop_type: null
+            segmask_crop_type: null
+            crop_hw:
+            - 144
+            - 144
+            crop_down_offset: 48
+            add_crop_binary_mask: false
+            add_coord_conv_map: false
+        mask_soft_approx_scheduler_config:
+          _target_: agent.encoder.MaskSoftApproxSchedulerConfig
+          num_steps: 40000
+          initial_value: 10.0
+          final_value: 1000.0
+          interpolation_scheme: cosine
+        use_contact_map: false
+        use_sdf_maps: false
+        use_normals_maps: false
+        which_objects: both
+        grasped_dtc_max_value: 0.2
+        env_dtc_max_value: 0.4
+        grasped_normals_mask_max_dtc_value: 0.2
+        env_normals_mask_max_dtc_value: 0.4
+        clamp_dtc: true
+        max_contact_prob: 0.1
+        mask_normals_within_sdf: true
+        dtc_adaptive_normalization: false
+        adaptive_normals_mask: true
+        max_depth: 2.0
+        image_shape:
+        - 13
+        - 180
+        - 240
+        learnable_contact_preprocess_params: true
+        learning_rate: 0.0001
+        weight_decay: 0.0
+        contact_model_name: local_multitask_outhd64all_home_crop_h144w144d48_mask_ctxtmask_seed_220979_epoch_9
+        zero_centered: false
+suite:
+  desc: null
+  value:
+    suite: frankagym
+    name: frankagym
+    frame_stack: 1
+    action_repeat: 1
+    discount: 0.99
+    hidden_dim: 1024
+    num_train_frames: 2010
+    num_seed_frames: 260
+    num_train_epochs: 5000
+    validate_every_epochs: 100
+    validate_diffusion_on_action_loss_every_epochs: 500
+    train_eval_diffusion_on_action_loss_every_epochs: 500
+    check_topk_every_epochs: 10
+    save_snapshot_every_epochs: 5000
+    eval_every_frames: 2000
+    num_eval_episodes: 5
+    save_snapshot: true
+    wait_for_user_to_start_episode: true
+    task_make_fn:
+      _target_: suite.frankagym.make
+      name: FrankaInsertion-v1
+      height: 240
+      width: 320
+      frame_stack: 1
+      action_repeat: 1
+      seed: 0
+      enable_arm: true
+      enable_gripper: true
+      start_with_gripper_open: true
+      enable_camera: true
+      path_to_depth_extrinsics: ~/fish_leon/FISH/cfgs/camera_poses/camera_poses_L515/20240904-122305/color_tf_world.npy
+      contact_estimation_model_ckpt_path: ~/fish_leon/contact_estimation/artifacts/175604_2/checkpoints/epoch=09-val_loss=0.00.ckpt
+      x_limit:
+      - 0.2
+      - 0.7
+      y_limit:
+      - -0.4
+      - 0.4
+      z_limit:
+      - -0.05
+      - 0.55
+      device: cuda
+      interpolation_frequency: 25
+      policy_frequency: 5
+      debug_timestamps: false
+      stop_after_action: false
+      open_loop: false
+      wait_for_new_camera_frames: true
+      action_key: action_trajectory_25hz
+      action_trajectory_horizon: 36
+      action_trajectories: true
+      path_to_zarr_dataset: /home/leonmkim/fish_leon/FISH/expert_demos/frankagym/FrankaInsertion-v1/120_240x320_all_twodim_left_to_right_annotated_start_idx_5hz_zstd7_EE_pxl_coords_expert_demos_imp_act/demos.zarr
+      agent_policy_cfg:
+        _target_: agent.diffusion_policy.DiffusionPolicyAgentConfig
+        compile: false
+        device: cuda
+        cam_resize_shape:
+        - 13
+        - 180
+        - 240
+        orig_cam_shape:
+        - 3
+        - 240
+        - 320
+        policy_frequency: 5
+        interpolation_frequency: 25
+        policy_cfg:
+          _target_: lerobot.common.policies.diffusion.configuration_diffusion.DiffusionConfig
+          n_obs_steps: 1
+          horizon: 36
+          n_action_steps: 36
+          output_shapes:
+            action:
+            - 7
+          input_normalization_modes:
+            observation.image: mean_std
+            observation.state: min_max
+            observation.action_history: min_max
+          output_normalization_modes:
+            action: min_max
+          vision_backbone: resnet18
+          pretrained_backbone_weights: null
+          transforms:
+          - _target_: torchaug.transforms.RandomAffine
+            degrees:
+            - -5
+            - 5
+            translate:
+            - 0.05
+            - 0.05
+            batch_transform: true
+            num_chunks: -1
+            batch_inplace: true
+          - _target_: torchaug.transforms.RandomColorJitter
+            brightness: 0.3
+            contrast: 0.4
+            saturation: 0.5
+            hue: 0.08
+            batch_transform: true
+            num_chunks: -1
+            batch_inplace: true
+          use_group_norm: true
+          spatial_softmax_num_keypoints: 32
+          action_history_encoder_config:
+            _target_: lerobot.common.policies.diffusion.configuration_diffusion.Unet1dEncoderConfig
+            in_channels: 7
+            out_channels: 32
+            history_length: 6
+            kernel_size: 5
+            downsample_kernel_size: 3
+            downsample_stride: 2
+            downsample_padding: 1
+          down_dims:
+          - 256
+          - 512
+          - 1024
+          kernel_size: 5
+          n_groups: 8
+          diffusion_step_embed_dim: 128
+          use_film_scale_modulation: true
+          noise_scheduler_type: DDIM
+          beta_schedule: squaredcos_cap_v2
+          beta_start: 0.0001
+          beta_end: 0.02
+          prediction_type: epsilon
+          clip_sample: true
+          clip_sample_range: 1.0
+          num_train_timesteps: 50
+          num_inference_steps: 10
+          do_mask_loss_for_padding: false
+          input_shapes:
+            observation.image:
+            - 13
+            - 180
+            - 240
+            context_observation.image:
+            - 13
+            - 180
+            - 240
+            observation.state:
+            - 8
+            observation.action_history:
+            - 7
+        train_cfg:
+          _target_: utils.TrainConfig
+          lr: 0.0001
+          lr_scheduler: cosine
+          lr_warmup_steps: 500
+          adam_betas:
+          - 0.95
+          - 0.999
+          adam_eps: 1.0e-08
+          adam_weight_decay: 1.0e-06
+          grad_clip_norm: 10
+          offline_steps: 1000000
+          use_amp: true
+        observation_cfg:
+          _target_: agent.encoder.VisualFeatureSet
+          use_depth: true
+          use_color: true
+          mask_input_dict:
+            _target_: agent.encoder.MaskInputDict
+            enable: false
+            representation: channels
+            mask_list:
+            - EE_obj_mask
+          crop_input_config:
+            _target_: agent.encoder.CropInputConfig
+            color_crop_type: null
+            depth_crop_type: null
+            segmask_crop_type: null
+            crop_hw:
+            - 144
+            - 144
+            crop_down_offset: 48
+            add_crop_binary_mask: false
+            add_coord_conv_map: false
+          context_input_config:
+            _target_: agent.encoder.ContextInputConfig
+            use_color: false
+            use_depth: false
+            mask_input_dict:
+              _target_: agent.encoder.MaskInputDict
+              enable: false
+              representation: channels
+              mask_list:
+              - EE_obj_mask
+            crop_input_config:
+              _target_: agent.encoder.CropInputConfig
+              color_crop_type: null
+              depth_crop_type: null
+              segmask_crop_type: null
+              crop_hw:
+              - 144
+              - 144
+              crop_down_offset: 48
+              add_crop_binary_mask: false
+              add_coord_conv_map: false
+          mask_soft_approx_scheduler_config:
+            _target_: agent.encoder.MaskSoftApproxSchedulerConfig
+            num_steps: 40000
+            initial_value: 10.0
+            final_value: 1000.0
+            interpolation_scheme: cosine
+          use_contact_map: false
+          use_sdf_maps: false
+          use_normals_maps: false
+          which_objects: both
+          grasped_dtc_max_value: 0.2
+          env_dtc_max_value: 0.4
+          grasped_normals_mask_max_dtc_value: 0.2
+          env_normals_mask_max_dtc_value: 0.4
+          clamp_dtc: true
+          max_contact_prob: 0.1
+          mask_normals_within_sdf: true
+          dtc_adaptive_normalization: false
+          adaptive_normals_mask: true
+          max_depth: 2.0
+          image_shape:
+          - 13
+          - 180
+          - 240
+          learnable_contact_preprocess_params: true
+          learning_rate: 0.0001
+          weight_decay: 0.0
+          contact_model_name: local_multitask_outhd64all_home_crop_h144w144d48_mask_ctxtmask_seed_220979_epoch_9
+          zero_centered: false
+      true_action_history: false
+num_train_frames_bc:
+  desc: null
+  value: 50000
+num_train_frames_drq:
+  desc: null
+  value: 1100000
+stddev_schedule_drq:
+  desc: null
+  value: linear(1.0,0.1,100000)
+task_name:
+  desc: null
+  value: FrankaInsertion-v1
+num_train_frames_vinn:
+  desc: null
+  value: 25000
+num_train_frames_diffusion:
+  desc: null
+  value: 1000000
+num_train_epochs_bc:
+  desc: null
+  value: 5000
+num_train_epochs_diffusion:
+  desc: null
+  value: 15000
+validate_every_epochs_bc:
+  desc: null
+  value: 5
+validate_every_epochs_diffusion:
+  desc: null
+  value: 250
+validate_diffusion_on_action_loss_every_epochs:
+  desc: null
+  value: 250
+train_eval_diffusion_on_action_loss_every_epochs:
+  desc: null
+  value: 250
+check_topk_every_epochs:
+  desc: null
+  value: 5
+check_topk_every_epochs_diffusion:
+  desc: null
+  value: 250
+save_snapshot_every_epochs_diffusion:
+  desc: null
+  value: 1500
+x_limit:
+  desc: null
+  value:
+  - 0.2
+  - 0.7
+y_limit:
+  desc: null
+  value:
+  - -0.4
+  - 0.4
+z_limit:
+  desc: null
+  value:
+  - -0.05
+  - 0.55
+home_displacement:
+  desc: null
+  value:
+  - 0.55
+  - 0.0
+  - 0.55
+  - 180.0
+  - 0.0
+  - 0.0
+enable_gripper:
+  desc: null
+  value: true
+start_with_gripper_open:
+  desc: null
+  value: true
+offset_mask:
+  desc: null
+  value:
+  - 1
+  - 1
+  - 1
+  - 1
+  - 1
+  - 1
+path_to_depth_extrinsics:
+  desc: null
+  value: ~/fish_leon/FISH/cfgs/camera_poses/camera_poses_L515/20240904-122305/color_tf_world.npy
+feature_type:
+  desc: null
+  value: 180x240_1_RGB_D_2.0_acthst_hst6_out32_dwnkrnl3_dwnstrd2_dwnpd1
+save_buffer:
+  desc: null
+  value: true
+num_eval:
+  desc: null
+  value: 5
+random_start:
+  desc: null
+  value: false
+eval_starts:
+  desc: null
+  value: /home/leonmkim/fish_leon/FISH/eval_starts/frankagym_pixels/FrankaInsertion-v1
+num_valid_demos:
+  desc: null
+  value: null
+load_checkpoint:
+  desc: null
+  value: true
+checkpoint_epoch:
+  desc: null
+  value: 12000
+load_residual_weight:
+  desc: null
+  value: false
+checkpoint_root_dir:
+  desc: null
+  value: /home/leonmkim/fish_leon/FISH
+checkpoint_weight_dir:
+  desc: null
+  value: /home/leonmkim/fish_leon/FISH/exp_local/frankagym_pixels/FrankaInsertion-v1/1045_0
+residual_weight:
+  desc: null
+  value: /home/leonmkim/fish_leon/FISH/weights/frankagym_pixels/FrankaInsertion-v1/weight.pt
+final_experiment_dir:
+  desc: null
+  value: ./exp_local/frankagym_pixels/FrankaInsertion-v1/1045_0/205317
+_wandb:
+  desc: null
+  value:
+    code_path: code/FISH/eval_robot.py
+    python_version: 3.10.14
+    cli_version: 0.17.5
+    framework: torch
+    is_jupyter_run: false
+    is_kaggle_kernel: false
+    start_time: 1735523603
+    t:
+      1:
+      - 1
+      - 41
+      - 49
+      - 50
+      - 55
+      - 83
+      2:
+      - 1
+      - 41
+      - 49
+      - 50
+      - 55
+      - 83
+      3:
+      - 16
+      - 23
+      - 35
+      4: 3.10.14
+      5: 0.17.5
+      8:
+      - 5
+      13: linux-x86_64
+grasped_obj_name:
+  desc: null
+  value: greece
+left_book_slot:
+  desc: null
+  value: twodim

205317/wandb/run-20241229_205323-2n31umej/files/diff.patch

@@ -0,0 +1,192 @@
+diff --git a/FISH/cfgs/config_eval.yaml b/FISH/cfgs/config_eval.yaml
+index 2303343..d3678fb 100644
+--- a/FISH/cfgs/config_eval.yaml
++++ b/FISH/cfgs/config_eval.yaml
+@@ -71,7 +71,7 @@ dtc_adaptive_normalization: false
+ mask_normals_within_sdf: true
+ adaptive_normals_mask: true
+ learnable_contact_preprocess_params: false
+-contact_model_name: 'local_multitask_outhd64all_home_crop_h144w144d48_ctxt_seed_183386_epoch_9' 
++contact_model_name: 'local_multitask_outhd64all_home_crop_h144w144d48_ctxt_seed_183386_epoch_9'
+ # contact_model_name: 'local_multitask_outhd64all_home_crop_h144w144d48_mask_ctxtmask_seed_220979_epoch_9'
+ contact_estimation_model_ckpt_path: '~/fish_leon/contact_estimation/artifacts/175604_2/checkpoints/epoch=09-val_loss=0.00.ckpt' # no mask no flow, just context
+ # contact_estimation_model_ckpt_path: '~/fish_leon/contact_estimation/artifacts/197406_2/checkpoints/epoch=08-val_loss=0.00.ckpt' # w/ mask no flow, just context
+@@ -123,6 +123,10 @@ bc_weight_type: 'qfilter' # linear, qfilter
+ 
+ # Load weights
+ load_checkpoint: ${agent.load_checkpoint}
++# 120 demos
++wandb_run_id: '1045_0' # greece
++
++
+ # 112_240x320_all_twodim_left_to_right_annotated_start_idx_5hz_zstd7_EE_pxl_coords_expert_demos_imp_act
+ # 14 demos
+ 
+@@ -144,7 +148,7 @@ load_checkpoint: ${agent.load_checkpoint}
+ # wandb_run_id: '523_0' # fowlers
+ # wandb_run_id: '524_0' # lib
+ # wandb_run_id: '525_0' # modelsys
+-wandb_run_id: '526_0' # electrodyn
++# wandb_run_id: '526_0' # electrodyn
+ 
+ # 54_240x320_hbm_twodim_fps_fix_annotated_start_idx_5hz_zstd7_EE_pxl_coords_expert_demos_imp_act
+ # 32 held out demos
+diff --git a/FISH/download_model_checkpoints.py b/FISH/download_model_checkpoints.py
+index 0f225ec..4d768b8 100644
+--- a/FISH/download_model_checkpoints.py
++++ b/FISH/download_model_checkpoints.py
+@@ -15,8 +15,8 @@ if not local_checkpoint_root_dir.exists():
+     local_checkpoint_root_dir.mkdir(parents=True)
+ 
+ # Download the model checkpoints
+-# remote_checkpoint_root_dir = Path('/mnt/grasp_high_usage/leonmkim/contact_estimation/FISH/exp_local/frankagym_pixels/FrankaInsertion-v1')
+-remote_checkpoint_root_dir = Path('/mnt/kostas-graid/datasets/extrinsic_contact_data/FISH/exp_local/frankagym_pixels/FrankaInsertion-v1')
++remote_checkpoint_root_dir = Path('/mnt/grasp_high_usage/leonmkim/contact_estimation/FISH/exp_local/frankagym_pixels/FrankaInsertion-v1')
++# remote_checkpoint_root_dir = Path('/mnt/kostas-graid/datasets/extrinsic_contact_data/FISH/exp_local/frankagym_pixels/FrankaInsertion-v1')
+ remote_checkpoint_root_dir = remote_checkpoint_root_dir.expanduser()
+ remote_username = 'leonmkim'
+ remote_host = 'grasp-login1'
+@@ -146,6 +146,11 @@ remote_host = 'grasp-login1'
+ #     '265627',
+ #     '265620',
+ # ]
++
++run_id_list = [
++    '1045_0',
++]
++
+ checkpoint_epoch = 12000
+ 
+ # use subprocess to download the model checkpoints in parallel
+Submodule ros_ws/src/contact_estimation_ros contains modified content
+diff --git a/ros_ws/src/contact_estimation_ros/fish_data_collection/src/rosbag_handler.py b/ros_ws/src/contact_estimation_ros/fish_data_collection/src/rosbag_handler.py
+index dcbcfa4..35dd7e8 100755
+--- a/ros_ws/src/contact_estimation_ros/fish_data_collection/src/rosbag_handler.py
++++ b/ros_ws/src/contact_estimation_ros/fish_data_collection/src/rosbag_handler.py
+@@ -15,6 +15,8 @@ import time
+ import datetime
+ import shutil
+ 
++from scipy.spatial.transform import Rotation as R
++
+ # solution to launching from main thread from: https://answers.ros.org/question/260212/start-launchfile-from-service/
+ # from Queue import Queue
+ 
+@@ -23,7 +25,7 @@ from fish_data_collection.srv import StartBagging, StartBaggingRequest, StartBag
+ from sensor_msgs.msg import CameraInfo
+ 
+ class RosbaggingHandler():
+-    def __init__(self, base_dset_dir, experiment_dirname, bag_launch_path, demo_dirname, camera_calibration_dir, 
++    def __init__(self, base_dset_dir, experiment_dirname, bag_launch_path, demo_dirname, 
+                  multicam=False, align_depth_to_color=False):
+         ## bag files and target files should be identified by episode id
+         self.align_depth_to_color = align_depth_to_color
+@@ -33,8 +35,6 @@ class RosbaggingHandler():
+         self.intrinsics_saved = False
+         self.extrinsics_saved = False
+ 
+-        self.camera_calibration_dir = camera_calibration_dir
+-
+         self.full_bag_dir = os.path.join(self.base_dset_dir, self.experiment_dirname, self.demo_dirname)
+         os.makedirs(self.full_bag_dir, exist_ok=True)
+ 
+@@ -63,7 +63,7 @@ class RosbaggingHandler():
+         self.abort_bagging_request = False
+ 
+         self.save_camera_intrinsics()
+-        self.save_camera_extrinsics(self.camera_calibration_dir)
++        self.save_camera_extrinsics()
+ 
+         while not rospy.is_shutdown():
+             if self.start_bagging_request:
+@@ -83,23 +83,56 @@ class RosbaggingHandler():
+         self.start_bagging_request = True
+         return response
+     
+-    def save_camera_extrinsics(self, path_to_calibration_dir):
++    def save_camera_extrinsics(self):
+         if not self.extrinsics_saved:
+-            # copy over the contents of the calibration directory, but not the directory itself, into the bag directory
+-            # list the contents of the calibration directory that end in .npy
+-            calibration_dir_contents = glob.glob(os.path.join(path_to_calibration_dir, '*.npy'))
+-            # calibration_dir_contents = os.listdir(path_to_calibration_dir)
+-            # copy the contents of the calibration directory into the bag directory
+-            for item in calibration_dir_contents:
+-                # make sure the item is a file
+-                if os.path.isfile(item):
+-                    # print('copying {} to {}'.format(item, os.path.join(self.full_bag_dir, os.path.basename(item))))
+-                    shutil.copy(item, os.path.join(self.full_bag_dir, os.path.basename(item)))
+-            # also save the path to the calibration directory as a txt file
+-            with open(os.path.join(self.full_bag_dir, 'calibration_dir.txt'), 'w') as f:
+-                f.write(path_to_calibration_dir)
++            # get cam extrinsic
++            # use ros tf2 to get transform from color_optical_frame to panda_link0
++            import tf2_ros
++            tfBuffer = tf2_ros.Buffer()
++            listener = tf2_ros.TransformListener(tfBuffer)
++
++            # get transform from color_optical_frame to panda_link0
++            # target, source where target is the parent frame and source is the child frame
++            color_tf_world_ros = tfBuffer.lookup_transform('camera_color_optical_frame', 'panda_link0', rospy.Time(0), rospy.Duration(1.0)).transform
++            self.color_tf_world = np.eye(4)
++            self.color_tf_world[0, 3] =color_tf_world_ros.translation.x
++            self.color_tf_world[1, 3] =color_tf_world_ros.translation.y
++            self.color_tf_world[2, 3] =color_tf_world_ros.translation.z
++            self.color_tf_world[:3, :3] = R.from_quat([color_tf_world_ros.rotation.x, color_tf_world_ros.rotation.y, color_tf_world_ros.rotation.z, color_tf_world_ros.rotation.w]).as_matrix()
++
++            # dump the info to the bag directory
++            np.save(os.path.join(self.full_bag_dir, 'color_tf_world.npy'), self.color_tf_world)
++            
++            depth_tf_world_ros = tfBuffer.lookup_transform('camera_depth_optical_frame', 'panda_link0', rospy.Time(0), rospy.Duration(1.0)).transform
++            self.depth_tf_world = np.eye(4)
++            self.depth_tf_world[0, 3] = depth_tf_world_ros.translation.x
++            self.depth_tf_world[1, 3] = depth_tf_world_ros.translation.y
++            self.depth_tf_world[2, 3] = depth_tf_world_ros.translation.z
++            self.depth_tf_world[:3, :3] = \
++            R.from_quat([
++                depth_tf_world_ros.rotation.x, 
++                depth_tf_world_ros.rotation.y, 
++                depth_tf_world_ros.rotation.z, 
++                depth_tf_world_ros.rotation.w,
++            ]).as_matrix()
++
++            np.save(os.path.join(self.full_bag_dir, 'depth_tf_world.npy'), self.depth_tf_world)
++
++            # # copy over the contents of the calibration directory, but not the directory itself, into the bag directory
++            # # list the contents of the calibration directory that end in .npy
++            # calibration_dir_contents = glob.glob(os.path.join(path_to_calibration_dir, '*.npy'))
++            # # calibration_dir_contents = os.listdir(path_to_calibration_dir)
++            # # copy the contents of the calibration directory into the bag directory
++            # for item in calibration_dir_contents:
++            #     # make sure the item is a file
++            #     if os.path.isfile(item):
++            #         # print('copying {} to {}'.format(item, os.path.join(self.full_bag_dir, os.path.basename(item))))
++            #         shutil.copy(item, os.path.join(self.full_bag_dir, os.path.basename(item)))
++            # # also save the path to the calibration directory as a txt file
++            # with open(os.path.join(self.full_bag_dir, 'calibration_dir.txt'), 'w') as f:
++            #     f.write(path_to_calibration_dir)
+         
+-            self.extrinsics_saved = True
++            # self.extrinsics_saved = True
+ 
+     def save_camera_intrinsics(self):
+         if not self.intrinsics_saved:
+@@ -198,9 +231,6 @@ class RosbaggingHandler():
+         if self.bag_recording:
+             self.current_recording_bag_path = bag_path
+         # response.success = True
+-
+-        
+-
+         # return response
+             
+     def stop_bagging_service_callback(self, req):
+@@ -304,7 +334,7 @@ if __name__ == '__main__':
+     assert os.path.exists(rosbag_launch_path), "rosbag launch file does not exist at path: {}".format(rosbag_launch_path)
+ 
+     camera_calibration_dir = '/home/leonmkim/FISH_franka_ros_ws/src/contact_estimation_ros/fish_data_collection/scripts/camera_poses_L515/20240313-153328'
+-    dc_handler = RosbaggingHandler(base_dset_dir, experiment_dirname, rosbag_launch_path, demo_dirname, camera_calibration_dir, multicam=multicam, align_depth_to_color=align_depth_to_color)
++    dc_handler = RosbaggingHandler(base_dset_dir, experiment_dirname, rosbag_launch_path, demo_dirname, multicam=multicam, align_depth_to_color=align_depth_to_color)
+ 
+     # rospy.spin()
+ 

205317/wandb/run-20241229_205323-2n31umej/files/media/videos/eval/0_eval_0_2dd247d1c4c04fa7cea7.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2dd247d1c4c04fa7cea7a03c76f9bcf0fac708360da8e1551f468b1b47fc66b9
+size 1171585

205317/wandb/run-20241229_205323-2n31umej/files/media/videos/eval/1_eval_1_f35843717fd499551c4b.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f35843717fd499551c4b0926cca736afc0945a08b7388201dffb14db7ec3fefa
+size 1166652

205317/wandb/run-20241229_205323-2n31umej/files/media/videos/eval/2_eval_2_c03a74d7b4bcab766fc2.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c03a74d7b4bcab766fc29ee131b9e9ac693b435c897bde76d32747b19fb58257
+size 1184865

205317/wandb/run-20241229_205323-2n31umej/files/media/videos/eval/3_eval_3_94417ff346f07f5fa23e.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:94417ff346f07f5fa23e9c53f6b0137c96c29dcb3d88023f1ed9c85959c133e3
+size 1210929

205317/wandb/run-20241229_205323-2n31umej/files/media/videos/eval/4_eval_4_20ea9dbf78cd7cd19362.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:20ea9dbf78cd7cd19362e85b72b43f16e349294c086d44f5e3adc1fd73292d66
+size 1225749

205317/wandb/run-20241229_205323-2n31umej/files/output.log

@@ -0,0 +1,225 @@
+
+loaded agent with feature_type: 180x240_1_RGB_D_2.0_acthst_hst6_out32_dwnkrnl3_dwnstrd2_dwnpd1
+[INFO] [1735523610.488230]: resetting environment
+[INFO] [1735523610.489337]: cleared current plan
+[INFO] [1735523610.491255]: moving to home
+[INFO] [1735523612.393259]: reached home
+[INFO] [1735523612.394469]: reset action history
+[INFO] [1735523615.152511]: environment reset
+Starting episode 0
+[INFO] [1735523615.155349]: resetting environment
+[INFO] [1735523615.156203]: cleared current plan
+[INFO] [1735523615.156975]: moving to home
+[INFO] [1735523616.159824]: reached home
+[INFO] [1735523616.162780]: reset action history
+[INFO] [1735523618.920455]: environment reset
+Press Enter to continue... after reseting env. To rate prev episode, press 0 for failure and 1 for success
+proceeding to start episode![INFO] [1735523677.158728]: resetting environment
+[INFO] [1735523677.160157]: cleared current plan
+[INFO] [1735523677.161284]: moving to home
+[INFO] [1735523678.163382]: reached home
+[INFO] [1735523678.164495]: reset action history
+[INFO] [1735523680.921619]: environment reset
+bag_path_name:=/home/leonmkim/fish_leon/FISH/exp_local/frankagym_pixels/FrankaInsertion-v1/1045_0/205317/episode_rosbags/episode_0_2024-12-29-20-54-40.bag
+... logging to /home/leonmkim/.ros/log/a8a09c64-c64e-11ef-96d2-5defea869005/roslaunch-leonmkim-ROG-Strix-G15CS-G15CS-1355277.log
+started roslaunch server http://158.130.50.37:34607/
+SUMMARY
+========
+PARAMETERS
+ * /rosdistro: noetic
+ * /rosversion: 1.16.0
+NODES
+  /
+    print_text (rostopic/rostopic)
+    pub_text (rostopic/rostopic)
+    rosbag_record (rosbag/record)
+ROS_MASTER_URI=http://localhost:11311
+process[pub_text-1]: started with pid [1355542]
+process[print_text-2]: started with pid [1355543]
+process[rosbag_record-3]: started with pid [1355567]
+started bagging!
+For topic gripper_width: timestamp difference is -181985368 for nearest: 1735523701867554436 - target: 1735523702049539804 at idx 249
+[INFO] [1735523707.719766]: Storing episode...
+[rosbag_record-3] killing on exit
+[pub_text-1] killing on exit[print_text-2] killing on exit
+[INFO] [1735523708.445290]: Stored episode 1.
+[INFO] [1735523708.445900]: Saving video...
+[INFO] [1735523708.649843]: Video saved!
+Starting episode 1
+[INFO] [1735523708.687548]: resetting environment
+[INFO] [1735523708.687901]: cleared current plan
+[INFO] [1735523708.688327]: moving to home
+[INFO] [1735523712.294043]: reached home
+[INFO] [1735523712.307056]: reset action history
+[INFO] [1735523715.017036]: environment reset
+Press Enter to continue... after reseting env. To rate prev episode, press 0 for failure and 1 for success
+proceeding to start episode!
+[INFO] [1735523725.239680]: resetting environment
+[INFO] [1735523725.240496]: cleared current plan
+[INFO] [1735523725.240901]: moving to home
+[INFO] [1735523726.242374]: reached home
+[INFO] [1735523726.242768]: reset action history
+[INFO] [1735523728.999282]: environment reset
+bag_path_name:=/home/leonmkim/fish_leon/FISH/exp_local/frankagym_pixels/FrankaInsertion-v1/1045_0/205317/episode_rosbags/episode_1_2024-12-29-20-55-29.bag
+... logging to /home/leonmkim/.ros/log/a8a09c64-c64e-11ef-96d2-5defea869005/roslaunch-leonmkim-ROG-Strix-G15CS-G15CS-1355277.log
+started roslaunch server http://158.130.50.37:36565/
+SUMMARY
+========
+PARAMETERS
+ * /rosdistro: noetic
+ * /rosversion: 1.16.0
+NODES
+  /
+    print_text (rostopic/rostopic)
+    pub_text (rostopic/rostopic)
+    rosbag_record (rosbag/record)
+ROS_MASTER_URI=http://localhost:11311
+process[pub_text-4]: started with pid [1355671]
+process[print_text-5]: started with pid [1355672]
+process[rosbag_record-6]: started with pid [1355696]
+started bagging!
+For topic gripper_width: timestamp difference is -119824366 for nearest: 1735523737367585226 - target: 1735523737487409592 at idx 126
+For topic gripper_width: timestamp difference is -117102585 for nearest: 1735523740767553367 - target: 1735523740884655952 at idx 169
+For topic gripper_width: timestamp difference is -114201270 for nearest: 1735523744367552318 - target: 1735523744481753588 at idx 218
+[INFO] [1735523755.752178]: Storing episode...
+[rosbag_record-6] killing on exit
+[print_text-5] killing on exit[pub_text-4] killing on exit
+[INFO] [1735523756.515642]: Stored episode 2.
+[INFO] [1735523756.515851]: Saving video...
+[INFO] [1735523756.765216]: Video saved!
+Starting episode 2
+[INFO] [1735523756.795257]: resetting environment
+[INFO] [1735523756.808481]: cleared current plan
+[INFO] [1735523756.813330]: moving to home
+[INFO] [1735523760.325088]: reached home
+[INFO] [1735523760.329300]: reset action history
+[INFO] [1735523763.101227]: environment reset
+Press Enter to continue... after reseting env. To rate prev episode, press 0 for failure and 1 for success
+proceeding to start episode![INFO] [1735523766.230032]: resetting environment
+[INFO] [1735523766.235887]: cleared current plan
+[INFO] [1735523766.242088]: moving to home
+[INFO] [1735523767.249254]: reached home
+[INFO] [1735523767.250467]: reset action history
+[INFO] [1735523770.007904]: environment reset
+bag_path_name:=/home/leonmkim/fish_leon/FISH/exp_local/frankagym_pixels/FrankaInsertion-v1/1045_0/205317/episode_rosbags/episode_2_2024-12-29-20-56-10.bag
+... logging to /home/leonmkim/.ros/log/a8a09c64-c64e-11ef-96d2-5defea869005/roslaunch-leonmkim-ROG-Strix-G15CS-G15CS-1355277.log
+started roslaunch server http://158.130.50.37:46491/
+SUMMARY
+========
+PARAMETERS
+ * /rosdistro: noetic
+ * /rosversion: 1.16.0
+NODES
+  /
+    print_text (rostopic/rostopic)
+    pub_text (rostopic/rostopic)
+    rosbag_record (rosbag/record)
+ROS_MASTER_URI=http://localhost:11311
+process[pub_text-7]: started with pid [1355801]
+process[print_text-8]: started with pid [1355802]
+process[rosbag_record-9]: started with pid [1355803]
+started bagging!
+For topic gripper_width: timestamp difference is -121928547 for nearest: 1735523770567553427 - target: 1735523770689481974 at idx 12
+For topic gripper_width: timestamp difference is -117060210 for nearest: 1735523781967578862 - target: 1735523782084639072 at idx 175
+For topic gripper_width: timestamp difference is -113255168 for nearest: 1735523786767553901 - target: 1735523786880809069 at idx 240
+For topic gripper_width: timestamp difference is -109881919 for nearest: 1735523790967551906 - target: 1735523791077433825 at idx 249
+[INFO] [1735523796.754220]: Storing episode...
+[rosbag_record-9] killing on exit
+[print_text-8] killing on exit[pub_text-7] killing on exit
+[INFO] [1735523797.544603]: Stored episode 3.
+[INFO] [1735523797.547160]: Saving video...
+[INFO] [1735523797.840266]: Video saved!
+Starting episode 3
+[INFO] [1735523797.913654]: resetting environment
+[INFO] [1735523797.923165]: cleared current plan
+[INFO] [1735523797.934853]: moving to home
+[INFO] [1735523801.149350]: reached home
+[INFO] [1735523801.153742]: reset action history
+[INFO] [1735523804.016112]: environment reset
+Press Enter to continue... after reseting env. To rate prev episode, press 0 for failure and 1 for success
+proceeding to start episode![INFO] [1735523806.043821]: resetting environment
+[INFO] [1735523806.045080]: cleared current plan
+[INFO] [1735523806.046980]: moving to home
+[INFO] [1735523807.052008]: reached home
+[INFO] [1735523807.060346]: reset action history
+[INFO] [1735523809.887860]: environment reset
+bag_path_name:=/home/leonmkim/fish_leon/FISH/exp_local/frankagym_pixels/FrankaInsertion-v1/1045_0/205317/episode_rosbags/episode_3_2024-12-29-20-56-49.bag
+... logging to /home/leonmkim/.ros/log/a8a09c64-c64e-11ef-96d2-5defea869005/roslaunch-leonmkim-ROG-Strix-G15CS-G15CS-1355277.log
+started roslaunch server http://158.130.50.37:44079/
+SUMMARY
+========
+PARAMETERS
+ * /rosdistro: noetic
+ * /rosversion: 1.16.0
+NODES
+  /
+    print_text (rostopic/rostopic)
+    pub_text (rostopic/rostopic)
+    rosbag_record (rosbag/record)
+ROS_MASTER_URI=http://localhost:11311
+process[pub_text-10]: started with pid [1355952]
+process[print_text-11]: started with pid [1355953]
+process[rosbag_record-12]: started with pid [1355954]
+started bagging!
+For topic gripper_width: timestamp difference is -114542308 for nearest: 1735523826167607007 - target: 1735523826282149315 at idx 212
+For topic gripper_width: timestamp difference is -114427458 for nearest: 1735523826367560972 - target: 1735523826481988430 at idx 213
+For topic gripper_width: timestamp difference is -113633562 for nearest: 1735523827367551159 - target: 1735523827481184721 at idx 224
+For topic gripper_width: timestamp difference is -109286799 for nearest: 1735523832767577396 - target: 1735523832876864195 at idx 249
+[INFO] [1735523837.946021]: Storing episode...
+[pub_text-10] killing on exit
+[rosbag_record-12] killing on exit[print_text-11] killing on exit
+[INFO] [1735523838.723363]: Stored episode 4.
+[INFO] [1735523838.731264]: Saving video...
+[INFO] [1735523839.282027]: Video saved!
+Starting episode 4
+[INFO] [1735523839.351837]: resetting environment
+[INFO] [1735523839.356024]: cleared current plan
+[INFO] [1735523839.371458]: moving to home
+[WARN] [1735523839.299034]: Plan exhausted
+[WARN] [1735523839.338714]: Plan exhausted
+[INFO] [1735523842.987996]: reached home
+[INFO] [1735523843.005783]: reset action history
+[INFO] [1735523845.794588]: environment reset
+Press Enter to continue... after reseting env. To rate prev episode, press 0 for failure and 1 for success
+proceeding to start episode![INFO] [1735523847.782817]: resetting environment
+[INFO] [1735523847.787585]: cleared current plan
+[INFO] [1735523847.792147]: moving to home
+[INFO] [1735523848.802336]: reached home
+[INFO] [1735523848.808479]: reset action history
+[INFO] [1735523851.596875]: environment reset
+bag_path_name:=/home/leonmkim/fish_leon/FISH/exp_local/frankagym_pixels/FrankaInsertion-v1/1045_0/205317/episode_rosbags/episode_4_2024-12-29-20-57-31.bag
+... logging to /home/leonmkim/.ros/log/a8a09c64-c64e-11ef-96d2-5defea869005/roslaunch-leonmkim-ROG-Strix-G15CS-G15CS-1355277.log
+started roslaunch server http://158.130.50.37:35151/
+SUMMARY
+========
+PARAMETERS
+ * /rosdistro: noetic
+ * /rosversion: 1.16.0
+NODES
+  /
+    print_text (rostopic/rostopic)
+    pub_text (rostopic/rostopic)
+    rosbag_record (rosbag/record)
+ROS_MASTER_URI=http://localhost:11311
+process[pub_text-13]: started with pid [1356084]
+process[print_text-14]: started with pid [1356085]
+process[rosbag_record-15]: started with pid [1356086]
+started bagging!
+[INFO] [1735523880.182312]: Storing episode...
+[rosbag_record-15] killing on exit
+[pub_text-13] killing on exit[print_text-14] killing on exit
+[INFO] [1735523880.957556]: Stored episode 5.
+[INFO] [1735523880.967952]: Saving video...
+[INFO] [1735523881.596726]: Video saved!
+[INFO] [1735523881.664569]: resetting environment
+[INFO] [1735523881.677550]: cleared current plan
+[INFO] [1735523881.697741]: moving to home
+[WARN] [1735523881.539209]: Plan exhausted
+[WARN] [1735523881.582183]: Plan exhausted
+[WARN] [1735523881.619458]: Plan exhausted
+[WARN] [1735523881.658995]: Plan exhausted
+[INFO] [1735523885.215016]: reached home
+[INFO] [1735523885.231086]: reset action history
+[INFO] [1735523888.142847]: environment reset
+Evaluation finished. To wrap up, rate prev episode, press 0 for failure and 1 for success
+proceeding to start episode!

205317/wandb/run-20241229_205323-2n31umej/files/requirements.txt

@@ -0,0 +1,339 @@
+Cython==3.0.10
+Farama-Notifications==0.0.4
+GitPython==3.1.43
+Jinja2==3.1.4
+Markdown==3.6
+MarkupSafe==2.1.5
+POT==0.7.0
+PyOpenGL==3.1.7
+PySocks==1.7.1
+PyYAML==6.0.1
+Pygments==2.18.0
+Rtree==1.3.0
+Werkzeug==3.0.3
+absl-py==2.1.0
+accelerate==0.33.0
+actionlib-msgs==1.13.0.post3
+actionlib==1.12.0
+actionlib==1.14.0
+aiohttp==3.9.5
+aiosignal==1.3.1
+angles==1.9.13
+antlr4-python3-runtime==4.9.3
+anyio==4.4.0
+asciitree==0.3.3
+async-timeout==4.0.3
+attrs==23.2.0
+autoprop==4.1.0
+beartype==0.18.5
+beautifulsoup4==4.12.3
+bondpy==1.8.6
+byol-pytorch==0.8.0
+cachetools==5.4.0
+camera-calibration-parsers==1.12.0
+camera-calibration==1.17.0
+cascadio==0.0.13
+catkin-pkg==1.0.0
+catkin==0.7.18
+catkin==0.8.10
+certifi==2024.7.4
+cffi==1.16.0
+chardet==5.2.0
+charset-normalizer==3.3.2
+click==8.1.7
+cloudpickle==3.0.0
+cmake==3.30.1
+colorlog==6.8.2
+contourpy==1.2.1
+controller-manager-msgs==0.20.0
+controller-manager==0.20.0
+coverage==7.6.0
+coveralls==4.0.1
+cv-bridge==1.16.2
+cycler==0.12.1
+datasets==2.20.0
+decorator==4.4.2
+deepdiff==7.0.1
+defusedxml==0.7.1
+diagnostic-analysis==1.11.0
+diagnostic-common-diagnostics==1.11.0
+diagnostic-updater==1.11.0
+diffusers==0.27.2
+dill==0.3.8
+distro==1.9.0
+dm-control==1.0.8
+dm-env==1.6
+dm-tree==0.1.8
+docker-pycreds==0.4.0
+docopt==0.6.2
+docutils==0.21.2
+dynamic-reconfigure==1.7.3
+einops==0.8.0
+embreex==2.17.7.post5
+empy==3.3.4
+etils==1.7.0
+exceptiongroup==1.2.2
+ezdxf==1.3.2
+fasteners==0.19
+filelock==3.15.4
+fonttools==4.53.1
+freetype-py==2.4.0
+frozenlist==1.4.1
+fsspec==2024.5.0
+gazebo_plugins==2.9.2
+gazebo_ros==2.9.2
+gdown==5.2.0
+gencpp==0.7.0
+geneus==3.0.0
+genlisp==0.4.18
+genmsg==0.5.12
+genmsg==0.6.0
+gennodejs==2.0.2
+genpy==0.6.14
+genpy==0.6.15
+geometry-msgs==1.13.0.post2
+gitdb==4.0.11
+glfw==2.7.0
+glooey==0.3.6
+gmsh==4.12.2
+gnupg==2.3.1
+google-auth-oauthlib==1.0.0
+google-auth==2.32.0
+grpcio==1.65.1
+gym-envs==0.0.1
+gym-notices==0.0.8
+gym==0.22.0
+gymnasium==0.29.1
+h11==0.14.0
+h5py==3.11.0
+hf_transfer==0.1.8
+httpcore==1.0.5
+httpx==0.27.0
+huggingface-hub==0.23.5
+hydra-core==1.3.2
+hydra-submitit-launcher==1.2.0
+idna==3.7
+image-geometry==1.16.2
+imageio-ffmpeg==0.5.1
+imageio==2.34.2
+importlib_metadata==8.2.0
+importlib_resources==6.4.0
+iniconfig==2.0.0
+interactive-markers==1.12.0
+joint-state-publisher-gui==1.15.1
+joint-state-publisher==1.15.1
+jsonschema-specifications==2023.12.1
+jsonschema==4.23.0
+kiwisolver==1.4.5
+kornia==0.7.3
+kornia_rs==0.1.5
+labmaze==1.0.6
+laser_geometry==1.6.7
+lazy_loader==0.4
+lerobot==0.1.0
+lightning-utilities==0.11.6
+llvmlite==0.43.0
+lxml==5.2.2
+manifold3d==2.5.1
+mapbox-earcut==1.0.1
+markdown-it-py==3.0.0
+matplotlib==3.9.1
+mdurl==0.1.2
+meshio==5.3.5
+message-filters==1.16.0
+more-itertools==10.3.0
+moviepy==1.0.3
+mpmath==1.3.0
+mujoco==3.2.0
+multidict==6.0.5
+multiprocess==0.70.16
+natsort==8.4.0
+netifaces==0.11.0
+networkx==3.3
+nodeenv==1.9.1
+numba==0.60.0
+numcodecs==0.13.0
+numpy==1.26.4
+nvidia-cublas-cu12==12.1.3.1
+nvidia-cuda-cupti-cu12==12.1.105
+nvidia-cuda-nvrtc-cu12==12.1.105
+nvidia-cuda-runtime-cu12==12.1.105
+nvidia-cudnn-cu12==9.1.0.70
+nvidia-cufft-cu12==11.0.2.54
+nvidia-curand-cu12==10.3.2.106
+nvidia-cusolver-cu12==11.4.5.107
+nvidia-cusparse-cu12==12.1.0.106
+nvidia-nccl-cu12==2.20.5
+nvidia-nvjitlink-cu12==12.5.82
+nvidia-nvtx-cu12==12.1.105
+oauthlib==3.2.2
+omegaconf==2.3.0
+openctm==0.0.5
+opencv-python==4.10.0.84
+ordered-set==4.1.0
+packaging==24.1
+pandas==2.2.2
+pillow==10.4.0
+pip==24.2
+platformdirs==4.2.2
+pluggy==1.5.0
+proglog==0.1.10
+protobuf==5.27.2
+psutil==6.0.0
+pyarrow-hotfix==0.6
+pyarrow==17.0.0
+pyasn1==0.6.0
+pyasn1_modules==0.4.0
+pyav==12.3.0
+pycollada==0.8
+pycparser==2.22
+pycryptodomex==3.21.0
+pyglet==1.5.29
+pyinstrument==4.6.2
+pymunk==6.8.1
+pyparsing==2.4.7
+pyrealsense2==2.54.2.5684
+pyribbit==0.1.46
+pyright==1.1.373
+pytest-beartype==0.0.2
+pytest-cov==5.0.0
+pytest==8.3.1
+python-dateutil==2.9.0.post0
+python-fcl==0.7.0.6
+python-qt-binding==0.4.4
+pytorch-lightning==2.4.0
+pytz==2024.1
+qt-dotgraph==0.4.2
+qt-gui-cpp==0.4.2
+qt-gui-py-common==0.4.2
+qt-gui==0.4.2
+referencing==0.35.1
+regex==2024.5.15
+requests-oauthlib==2.0.0
+requests==2.32.3
+rerun-sdk==0.17.0
+resource_retriever==1.12.7
+rich==13.7.1
+ros-numpy==0.0.5
+rosbag==1.16.0
+rosboost-cfg==1.15.8
+rosclean==1.15.8
+roscpp==1.15.11
+roscreate==1.15.8
+rosgraph-msgs==1.11.3.post2
+rosgraph==1.15.11
+rosgraph==1.16.0
+roslaunch==1.16.0
+roslib==1.14.7.post0
+roslib==1.15.8
+roslint==0.12.0
+roslz4==1.16.0
+rosmake==1.15.8
+rosmaster==1.16.0
+rosmsg==1.16.0
+rosnode==1.16.0
+rosparam==1.16.0
+rospkg==1.5.1
+rospy==1.15.11
+rospy==1.16.0
+rosservice==1.16.0
+rostest==1.16.0
+rostopic==1.16.0
+rosunit==1.15.8
+roswtf==1.16.0
+rpds-py==0.19.1
+rqt-console==0.4.12
+rqt-image-view==0.4.17
+rqt-logger-level==0.4.12
+rqt-moveit==0.5.11
+rqt-reconfigure==0.5.5
+rqt-robot-dashboard==0.5.8
+rqt-robot-monitor==0.5.15
+rqt-runtime-monitor==0.5.10
+rqt-rviz==0.7.0
+rqt-tf-tree==0.6.4
+rqt_action==0.4.9
+rqt_bag==0.5.1
+rqt_bag_plugins==0.5.1
+rqt_dep==0.4.12
+rqt_graph==0.4.14
+rqt_gui==0.5.3
+rqt_gui_py==0.5.3
+rqt_launch==0.4.9
+rqt_msg==0.4.10
+rqt_nav_view==0.5.7
+rqt_plot==0.4.13
+rqt_pose_view==0.5.11
+rqt_publisher==0.4.10
+rqt_py_common==0.5.3
+rqt_py_console==0.4.10
+rqt_robot_steering==0.5.12
+rqt_service_caller==0.4.10
+rqt_shell==0.4.11
+rqt_srv==0.4.9
+rqt_top==0.4.10
+rqt_topic==0.4.13
+rqt_web==0.4.10
+rsa==4.9
+ruff==0.5.4
+rviz==1.14.25
+safetensors==0.4.3
+scikit-image==0.24.0
+scikit-video==1.1.11
+scipy==1.14.0
+seaborn==0.13.2
+sensor-msgs==1.13.1
+sentry-sdk==2.11.0
+setproctitle==1.3.3
+setuptools==65.5.0
+shapely==2.0.5
+signature_dispatch==1.0.1
+six==1.16.0
+smach-ros==2.5.2
+smach==2.5.2
+smclib==1.8.6
+smmap==5.0.1
+sniffio==1.3.1
+soupsieve==2.5
+std-msgs==0.5.13.post0
+submitit==1.5.1
+svg.path==6.3
+sympy==1.13.1
+tensorboard-data-server==0.7.2
+tensorboard==2.14.0
+termcolor==2.4.0
+tf-conversions==1.13.2
+tf2-geometry-msgs==0.7.7
+tf2-kdl==0.7.7
+tf2-msgs==0.7.2.post3
+tf2-py==0.7.7
+tf2-ros==0.6.5
+tf2-ros==0.7.7
+tf2_py==0.6.5.post1
+tf==1.13.2
+tifffile==2024.7.24
+tomli==2.0.1
+topic-tools==1.16.0
+torch==2.4.0
+torchaug==0.5.2
+torchmetrics==1.4.0.post0
+torchvision==0.19.0
+tqdm==4.66.4
+trimesh==4.4.3
+triton==3.0.0
+typeguard==3.0.2
+typing_extensions==4.12.2
+tzdata==2024.1
+urchin==0.0.27
+urllib3==2.2.2
+vecrec==0.3.1
+vhacdx==0.0.8.post1
+wandb==0.17.5
+wheel==0.43.0
+xacro==1.14.18
+xatlas==0.0.9
+xxhash==3.4.1
+yarl==1.9.4
+zarr==2.18.2
+zipp==3.19.2
+zstandard==0.23.0

205317/wandb/run-20241229_205323-2n31umej/files/wandb-metadata.json

@@ -0,0 +1,92 @@
+{
+    "os": "Linux-5.15.0-125-generic-x86_64-with-glibc2.31",
+    "python": "3.10.14",
+    "heartbeatAt": "2024-12-30T01:53:24.140645",
+    "startedAt": "2024-12-30T01:53:23.741358",
+    "docker": null,
+    "cuda": null,
+    "args": [
+        "agent=diffusion",
+        "suite=frankagym",
+        "suite/frankagym_task@_global_=insertion"
+    ],
+    "state": "running",
+    "program": "/home/leonmkim/fish_leon/FISH/eval_robot.py",
+    "codePathLocal": null,
+    "codePath": "FISH/eval_robot.py",
+    "git": {
+        "remote": "https://github.com/leonmkim/fish_leon.git",
+        "commit": "23581857e509febdae7f0ac45cc9cf4f9081f85c"
+    },
+    "email": "leonmkim@seas.upenn.edu",
+    "root": "/home/leonmkim/fish_leon",
+    "host": "leonmkim-ROG-Strix-G15CS-G15CS",
+    "username": "leonmkim",
+    "executable": "/home/leonmkim/.pyenv/versions/lerobot/bin/python",
+    "cpu_count": 8,
+    "cpu_count_logical": 8,
+    "cpu_freq": {
+        "current": 3933.782375,
+        "min": 800.0,
+        "max": 4700.0
+    },
+    "cpu_freq_per_core": [
+        {
+            "current": 3000.0,
+            "min": 800.0,
+            "max": 4700.0
+        },
+        {
+            "current": 3000.0,
+            "min": 800.0,
+            "max": 4700.0
+        },
+        {
+            "current": 4493.128,
+            "min": 800.0,
+            "max": 4700.0
+        },
+        {
+            "current": 3000.0,
+            "min": 800.0,
+            "max": 4700.0
+        },
+        {
+            "current": 3000.0,
+            "min": 800.0,
+            "max": 4700.0
+        },
+        {
+            "current": 3000.0,
+            "min": 800.0,
+            "max": 4700.0
+        },
+        {
+            "current": 3000.0,
+            "min": 800.0,
+            "max": 4700.0
+        },
+        {
+            "current": 4496.933,
+            "min": 800.0,
+            "max": 4700.0
+        }
+    ],
+    "disk": {
+        "/": {
+            "total": 915.3232879638672,
+            "used": 463.0247688293457
+        }
+    },
+    "gpu": "NVIDIA GeForce RTX 2070 SUPER",
+    "gpu_count": 1,
+    "gpu_devices": [
+        {
+            "name": "NVIDIA GeForce RTX 2070 SUPER",
+            "memory_total": 8589934592
+        }
+    ],
+    "memory": {
+        "total": 62.71595764160156
+    }
+}

205317/wandb/run-20241229_205323-2n31umej/files/wandb-summary.json

@@ -0,0 +1 @@
+{"eval/0_eval": {"_type": "video-file", "sha256": "2dd247d1c4c04fa7cea7a03c76f9bcf0fac708360da8e1551f468b1b47fc66b9", "size": 1171585, "path": "media/videos/eval/0_eval_0_2dd247d1c4c04fa7cea7.mp4"}, "global_step": 660, "_timestamp": 1735523890.298451, "_runtime": 286.548122882843, "_step": 9, "eval/1_eval": {"_type": "video-file", "sha256": "f35843717fd499551c4b0926cca736afc0945a08b7388201dffb14db7ec3fefa", "size": 1166652, "path": "media/videos/eval/1_eval_1_f35843717fd499551c4b.mp4"}, "eval/num_success": 4.0, "episode": 5.0, "eval/success_rate": 0.800000011920929, "eval/2_eval": {"_type": "video-file", "sha256": "c03a74d7b4bcab766fc29ee131b9e9ac693b435c897bde76d32747b19fb58257", "size": 1184865, "path": "media/videos/eval/2_eval_2_c03a74d7b4bcab766fc2.mp4"}, "eval/3_eval": {"_type": "video-file", "sha256": "94417ff346f07f5fa23e9c53f6b0137c96c29dcb3d88023f1ed9c85959c133e3", "size": 1210929, "path": "media/videos/eval/3_eval_3_94417ff346f07f5fa23e.mp4"}, "eval/4_eval": {"_type": "video-file", "sha256": "20ea9dbf78cd7cd19362e85b72b43f16e349294c086d44f5e3adc1fd73292d66", "size": 1225749, "path": "media/videos/eval/4_eval_4_20ea9dbf78cd7cd19362.mp4"}, "_wandb": {"runtime": 286}}

205317/wandb/run-20241229_205323-2n31umej/logs/debug.log

@@ -0,0 +1,31 @@
+2024-12-29 20:53:23,743 INFO    MainThread:1355277 [wandb_setup.py:_flush():76] Current SDK version is 0.17.5
+2024-12-29 20:53:23,743 INFO    MainThread:1355277 [wandb_setup.py:_flush():76] Configure stats pid to 1355277
+2024-12-29 20:53:23,743 INFO    MainThread:1355277 [wandb_setup.py:_flush():76] Loading settings from /home/leonmkim/.config/wandb/settings
+2024-12-29 20:53:23,743 INFO    MainThread:1355277 [wandb_setup.py:_flush():76] Loading settings from /home/leonmkim/fish_leon/FISH/exp_local/frankagym_pixels/FrankaInsertion-v1/1045_0/205317/wandb/settings
+2024-12-29 20:53:23,743 INFO    MainThread:1355277 [wandb_setup.py:_flush():76] Loading settings from environment variables: {}
+2024-12-29 20:53:23,743 INFO    MainThread:1355277 [wandb_setup.py:_flush():76] Applying setup settings: {'_disable_service': False}
+2024-12-29 20:53:23,743 INFO    MainThread:1355277 [wandb_setup.py:_flush():76] Inferring run settings from compute environment: {'program_relpath': 'FISH/eval_robot.py', 'program_abspath': '/home/leonmkim/fish_leon/FISH/eval_robot.py', 'program': '/home/leonmkim/fish_leon/FISH/eval_robot.py'}
+2024-12-29 20:53:23,743 INFO    MainThread:1355277 [wandb_setup.py:_flush():76] Applying login settings: {}
+2024-12-29 20:53:23,743 INFO    MainThread:1355277 [wandb_init.py:_log_setup():529] Logging user logs to /home/leonmkim/fish_leon/FISH/exp_local/frankagym_pixels/FrankaInsertion-v1/1045_0/205317/wandb/run-20241229_205323-2n31umej/logs/debug.log
+2024-12-29 20:53:23,743 INFO    MainThread:1355277 [wandb_init.py:_log_setup():530] Logging internal logs to /home/leonmkim/fish_leon/FISH/exp_local/frankagym_pixels/FrankaInsertion-v1/1045_0/205317/wandb/run-20241229_205323-2n31umej/logs/debug-internal.log
+2024-12-29 20:53:23,744 INFO    MainThread:1355277 [wandb_init.py:init():569] calling init triggers
+2024-12-29 20:53:23,744 INFO    MainThread:1355277 [wandb_init.py:init():576] wandb.init called with sweep_config: {}
+config: {'root_dir': '/home/leonmkim/fish_leon', 'replay_buffer_size': 150000, 'replay_buffer_num_workers': 2, 'nstep': 3, 'batch_size': 128, 'seed': 0, 'dataset_shuffle_seed': 5, 'device': 'cuda', 'save_video': True, 'save_train_video': True, 'use_tb': True, 'use_wandb': True, 'wandb_run_id': '1045_0', 'wandb_notes': '1045_0_', 'eval': True, 'true_action_history': False, 'train_pad_after': 4, 'process_contact_features': True, 'obs_type': 'pixels', 'use_color': True, 'use_depth': True, 'use_masks': False, 'mask_list': ['EE_obj_mask'], 'mask_representation': 'channels', 'crop_hw': [144, 144], 'crop_down_offset': 48, 'color_crop_type': None, 'depth_crop_type': None, 'segmask_crop_type': None, 'add_crop_binary_mask': False, 'add_coord_conv_map': False, 'use_context_color': False, 'use_context_depth': False, 'use_context_segmask': False, 'context_color_crop_type': None, 'context_depth_crop_type': None, 'context_segmask_crop_type': None, 'context_add_crop_binary_mask': False, 'context_add_coord_conv_map': False, 'use_contact_map': False, 'use_sdf_maps': False, 'use_normals_maps': False, 'which_objects': 'both', 'max_contact_prob': 0.1, 'max_depth': 2.0, 'grasped_dtc_max_value': 0.2, 'env_dtc_max_value': 0.4, 'grasped_normals_mask_max_dtc_value': 0.2, 'env_normals_mask_max_dtc_value': 0.4, 'clamp_dtc': True, 'dtc_adaptive_normalization': False, 'mask_normals_within_sdf': True, 'adaptive_normals_mask': True, 'learnable_contact_preprocess_params': True, 'contact_model_name': 'local_multitask_outhd64all_home_crop_h144w144d48_mask_ctxtmask_seed_220979_epoch_9', 'contact_estimation_model_ckpt_path': '~/fish_leon/contact_estimation/artifacts/175604_2/checkpoints/epoch=09-val_loss=0.00.ckpt', 'encoder_type': 'small', 'debug_timestamps': False, 'open_loop': False, 'action_trajectories': True, 'stop_after_action': False, 'interpolation_frequency': 25, 'policy_frequency': 5, 'wait_for_new_camera_frames': True, 'baseline': False, 'train_demo_idxs_list_or_num': -1, 'log_train_every_steps': 25, 'name_of_expert_demo': '120_240x320_all_twodim_left_to_right_annotated_start_idx_5hz_zstd7_EE_pxl_coords_expert_demos_imp_act', 'expert_dataset_dirpath': '/home/leonmkim/fish_leon/FISH/expert_demos/frankagym/FrankaInsertion-v1/120_240x320_all_twodim_left_to_right_annotated_start_idx_5hz_zstd7_EE_pxl_coords_expert_demos_imp_act', 'store_dataset_in_memory': False, 'expert_dataset': '/home/leonmkim/fish_leon/FISH/expert_demos/frankagym/FrankaInsertion-v1/120_240x320_all_twodim_left_to_right_annotated_start_idx_5hz_zstd7_EE_pxl_coords_expert_demos_imp_act/demos.zarr', 'action_key': 'action_trajectory_25hz', 'semantic_demo_grouping_name': 'semantic_demo_grouping.yaml', 'semantic_demo_grouping': '/home/leonmkim/fish_leon/FISH/expert_demos/frankagym/FrankaInsertion-v1/120_240x320_all_twodim_left_to_right_annotated_start_idx_5hz_zstd7_EE_pxl_coords_expert_demos_imp_act/semantic_demo_grouping.yaml', 'include_groups_list': ['greece_twodim_nominal', 'greece_twodim_recovery'], 'expert_dataset_config': '/home/leonmkim/fish_leon/FISH/expert_demos/frankagym/FrankaInsertion-v1/120_240x320_all_twodim_left_to_right_annotated_start_idx_5hz_zstd7_EE_pxl_coords_expert_demos_imp_act/demo_config.yaml', 'name_of_valid_demo': '120_240x320_all_twodim_left_to_right_annotated_start_idx_5hz_zstd7_EE_pxl_coords_expert_demos_imp_act', 'valid_dataset_dir': '/home/leonmkim/fish_leon/FISH/expert_demos/frankagym/FrankaInsertion-v1/120_240x320_all_twodim_left_to_right_annotated_start_idx_5hz_zstd7_EE_pxl_coords_expert_demos_imp_act/demos.zarr', 'valid_demo_idxs_list_or_num': None, 'val_num_groups': 0, 'load_bc': True, 'checkpoint_epoch_list': [99, 199, 299, 399, 499, 599, 699, 799, 899, 999, 1249, 1499, 1749, 1999, 2999, 3999, 4999, 5999, 6999, 7999, 8999, 9999], 'snapshot_root_dir': '/mnt/grasp_high_usage/leonmkim/contact_estimation/FISH', 'save_snapshot': True, 'save_last_snapshot': True, 'save_snapshot_when_done': True, 'top_k_checkpoints': 5, 'save_snapshot_link_to_weights_dir': 'deprecated', 'bc_regularize': False, 'bc_weight_type': 'qfilter', 'experiment_dir': './exp_local/frankagym_pixels/FrankaInsertion-v1/1045_0', 'agent': {'_target_': 'agent.diffusion_policy.DiffusionPolicyAgent', 'name': 'diffusion_policy', 'load_checkpoint': True, 'device': 'cuda', 'n_obs_steps': 1, 'suite_name': 'frankagym', 'obs_type': 'pixels', 'enable_arm': True, 'enable_camera': True, 'use_tb': True, 'desired_image_shape': [13, 180, 240], 'orig_cam_shape': [3, 240, 320], 'config': {'_target_': 'agent.diffusion_policy.DiffusionPolicyAgentConfig', 'compile': False, 'device': 'cuda', 'cam_resize_shape': [13, 180, 240], 'orig_cam_shape': [3, 240, 320], 'policy_frequency': 5, 'interpolation_frequency': 25, 'policy_cfg': {'_target_': 'lerobot.common.policies.diffusion.configuration_diffusion.DiffusionConfig', 'n_obs_steps': 1, 'horizon': 36, 'n_action_steps': 36, 'output_shapes': {'action': [7]}, 'input_normalization_modes': {'observation.image': 'mean_std', 'observation.state': 'min_max', 'observation.action_history': 'min_max'}, 'output_normalization_modes': {'action': 'min_max'}, 'vision_backbone': 'resnet18', 'pretrained_backbone_weights': None, 'transforms': [{'_target_': 'torchaug.transforms.RandomAffine', 'degrees': [-5, 5], 'translate': [0.05, 0.05], 'batch_transform': True, 'num_chunks': -1, 'batch_inplace': True}, {'_target_': 'torchaug.transforms.RandomColorJitter', 'brightness': 0.3, 'contrast': 0.4, 'saturation': 0.5, 'hue': 0.08, 'batch_transform': True, 'num_chunks': -1, 'batch_inplace': True}], 'use_group_norm': True, 'spatial_softmax_num_keypoints': 32, 'action_history_encoder_config': {'_target_': 'lerobot.common.policies.diffusion.configuration_diffusion.Unet1dEncoderConfig', 'in_channels': 7, 'out_channels': 32, 'history_length': 6, 'kernel_size': 5, 'downsample_kernel_size': 3, 'downsample_stride': 2, 'downsample_padding': 1}, 'down_dims': [256, 512, 1024], 'kernel_size': 5, 'n_groups': 8, 'diffusion_step_embed_dim': 128, 'use_film_scale_modulation': True, 'noise_scheduler_type': 'DDIM', 'beta_schedule': 'squaredcos_cap_v2', 'beta_start': 0.0001, 'beta_end': 0.02, 'prediction_type': 'epsilon', 'clip_sample': True, 'clip_sample_range': 1.0, 'num_train_timesteps': 50, 'num_inference_steps': 10, 'do_mask_loss_for_padding': False, 'input_shapes': {'observation.image': [13, 180, 240], 'context_observation.image': [13, 180, 240], 'observation.state': [8], 'observation.action_history': [7]}}, 'train_cfg': {'_target_': 'utils.TrainConfig', 'lr': 0.0001, 'lr_scheduler': 'cosine', 'lr_warmup_steps': 500, 'adam_betas': [0.95, 0.999], 'adam_eps': 1e-08, 'adam_weight_decay': 1e-06, 'grad_clip_norm': 10, 'offline_steps': 1000000, 'use_amp': True}, 'observation_cfg': {'_target_': 'agent.encoder.VisualFeatureSet', 'use_depth': True, 'use_color': True, 'mask_input_dict': {'_target_': 'agent.encoder.MaskInputDict', 'enable': False, 'representation': 'channels', 'mask_list': ['EE_obj_mask']}, 'crop_input_config': {'_target_': 'agent.encoder.CropInputConfig', 'color_crop_type': None, 'depth_crop_type': None, 'segmask_crop_type': None, 'crop_hw': [144, 144], 'crop_down_offset': 48, 'add_crop_binary_mask': False, 'add_coord_conv_map': False}, 'context_input_config': {'_target_': 'agent.encoder.ContextInputConfig', 'use_color': False, 'use_depth': False, 'mask_input_dict': {'_target_': 'agent.encoder.MaskInputDict', 'enable': False, 'representation': 'channels', 'mask_list': ['EE_obj_mask']}, 'crop_input_config': {'_target_': 'agent.encoder.CropInputConfig', 'color_crop_type': None, 'depth_crop_type': None, 'segmask_crop_type': None, 'crop_hw': [144, 144], 'crop_down_offset': 48, 'add_crop_binary_mask': False, 'add_coord_conv_map': False}}, 'mask_soft_approx_scheduler_config': {'_target_': 'agent.encoder.MaskSoftApproxSchedulerConfig', 'num_steps': 40000, 'initial_value': 10.0, 'final_value': 1000.0, 'interpolation_scheme': 'cosine'}, 'use_contact_map': False, 'use_sdf_maps': False, 'use_normals_maps': False, 'which_objects': 'both', 'grasped_dtc_max_value': 0.2, 'env_dtc_max_value': 0.4, 'grasped_normals_mask_max_dtc_value': 0.2, 'env_normals_mask_max_dtc_value': 0.4, 'clamp_dtc': True, 'max_contact_prob': 0.1, 'mask_normals_within_sdf': True, 'dtc_adaptive_normalization': False, 'adaptive_normals_mask': True, 'max_depth': 2.0, 'image_shape': [13, 180, 240], 'learnable_contact_preprocess_params': True, 'learning_rate': 0.0001, 'weight_decay': 0.0, 'contact_model_name': 'local_multitask_outhd64all_home_crop_h144w144d48_mask_ctxtmask_seed_220979_epoch_9', 'zero_centered': False}}}, 'suite': {'suite': 'frankagym', 'name': 'frankagym', 'frame_stack': 1, 'action_repeat': 1, 'discount': 0.99, 'hidden_dim': 1024, 'num_train_frames': 2010, 'num_seed_frames': 260, 'num_train_epochs': 5000, 'validate_every_epochs': 100, 'validate_diffusion_on_action_loss_every_epochs': 500, 'train_eval_diffusion_on_action_loss_every_epochs': 500, 'check_topk_every_epochs': 10, 'save_snapshot_every_epochs': 5000, 'eval_every_frames': 2000, 'num_eval_episodes': 5, 'save_snapshot': True, 'wait_for_user_to_start_episode': True, 'task_make_fn': {'_target_': 'suite.frankagym.make', 'name': 'FrankaInsertion-v1', 'height': 240, 'width': 320, 'frame_stack': 1, 'action_repeat': 1, 'seed': 0, 'enable_arm': True, 'enable_gripper': True, 'start_with_gripper_open': True, 'enable_camera': True, 'path_to_depth_extrinsics': '~/fish_leon/FISH/cfgs/camera_poses/camera_poses_L515/20240904-122305/color_tf_world.npy', 'contact_estimation_model_ckpt_path': '~/fish_leon/contact_estimation/artifacts/175604_2/checkpoints/epoch=09-val_loss=0.00.ckpt', 'x_limit': [0.2, 0.7], 'y_limit': [-0.4, 0.4], 'z_limit': [-0.05, 0.55], 'device': 'cuda', 'interpolation_frequency': 25, 'policy_frequency': 5, 'debug_timestamps': False, 'stop_after_action': False, 'open_loop': False, 'wait_for_new_camera_frames': True, 'action_key': 'action_trajectory_25hz', 'action_trajectory_horizon': 36, 'action_trajectories': True, 'path_to_zarr_dataset': '/home/leonmkim/fish_leon/FISH/expert_demos/frankagym/FrankaInsertion-v1/120_240x320_all_twodim_left_to_right_annotated_start_idx_5hz_zstd7_EE_pxl_coords_expert_demos_imp_act/demos.zarr', 'agent_policy_cfg': {'_target_': 'agent.diffusion_policy.DiffusionPolicyAgentConfig', 'compile': False, 'device': 'cuda', 'cam_resize_shape': [13, 180, 240], 'orig_cam_shape': [3, 240, 320], 'policy_frequency': 5, 'interpolation_frequency': 25, 'policy_cfg': {'_target_': 'lerobot.common.policies.diffusion.configuration_diffusion.DiffusionConfig', 'n_obs_steps': 1, 'horizon': 36, 'n_action_steps': 36, 'output_shapes': {'action': [7]}, 'input_normalization_modes': {'observation.image': 'mean_std', 'observation.state': 'min_max', 'observation.action_history': 'min_max'}, 'output_normalization_modes': {'action': 'min_max'}, 'vision_backbone': 'resnet18', 'pretrained_backbone_weights': None, 'transforms': [{'_target_': 'torchaug.transforms.RandomAffine', 'degrees': [-5, 5], 'translate': [0.05, 0.05], 'batch_transform': True, 'num_chunks': -1, 'batch_inplace': True}, {'_target_': 'torchaug.transforms.RandomColorJitter', 'brightness': 0.3, 'contrast': 0.4, 'saturation': 0.5, 'hue': 0.08, 'batch_transform': True, 'num_chunks': -1, 'batch_inplace': True}], 'use_group_norm': True, 'spatial_softmax_num_keypoints': 32, 'action_history_encoder_config': {'_target_': 'lerobot.common.policies.diffusion.configuration_diffusion.Unet1dEncoderConfig', 'in_channels': 7, 'out_channels': 32, 'history_length': 6, 'kernel_size': 5, 'downsample_kernel_size': 3, 'downsample_stride': 2, 'downsample_padding': 1}, 'down_dims': [256, 512, 1024], 'kernel_size': 5, 'n_groups': 8, 'diffusion_step_embed_dim': 128, 'use_film_scale_modulation': True, 'noise_scheduler_type': 'DDIM', 'beta_schedule': 'squaredcos_cap_v2', 'beta_start': 0.0001, 'beta_end': 0.02, 'prediction_type': 'epsilon', 'clip_sample': True, 'clip_sample_range': 1.0, 'num_train_timesteps': 50, 'num_inference_steps': 10, 'do_mask_loss_for_padding': False, 'input_shapes': {'observation.image': [13, 180, 240], 'context_observation.image': [13, 180, 240], 'observation.state': [8], 'observation.action_history': [7]}}, 'train_cfg': {'_target_': 'utils.TrainConfig', 'lr': 0.0001, 'lr_scheduler': 'cosine', 'lr_warmup_steps': 500, 'adam_betas': [0.95, 0.999], 'adam_eps': 1e-08, 'adam_weight_decay': 1e-06, 'grad_clip_norm': 10, 'offline_steps': 1000000, 'use_amp': True}, 'observation_cfg': {'_target_': 'agent.encoder.VisualFeatureSet', 'use_depth': True, 'use_color': True, 'mask_input_dict': {'_target_': 'agent.encoder.MaskInputDict', 'enable': False, 'representation': 'channels', 'mask_list': ['EE_obj_mask']}, 'crop_input_config': {'_target_': 'agent.encoder.CropInputConfig', 'color_crop_type': None, 'depth_crop_type': None, 'segmask_crop_type': None, 'crop_hw': [144, 144], 'crop_down_offset': 48, 'add_crop_binary_mask': False, 'add_coord_conv_map': False}, 'context_input_config': {'_target_': 'agent.encoder.ContextInputConfig', 'use_color': False, 'use_depth': False, 'mask_input_dict': {'_target_': 'agent.encoder.MaskInputDict', 'enable': False, 'representation': 'channels', 'mask_list': ['EE_obj_mask']}, 'crop_input_config': {'_target_': 'agent.encoder.CropInputConfig', 'color_crop_type': None, 'depth_crop_type': None, 'segmask_crop_type': None, 'crop_hw': [144, 144], 'crop_down_offset': 48, 'add_crop_binary_mask': False, 'add_coord_conv_map': False}}, 'mask_soft_approx_scheduler_config': {'_target_': 'agent.encoder.MaskSoftApproxSchedulerConfig', 'num_steps': 40000, 'initial_value': 10.0, 'final_value': 1000.0, 'interpolation_scheme': 'cosine'}, 'use_contact_map': False, 'use_sdf_maps': False, 'use_normals_maps': False, 'which_objects': 'both', 'grasped_dtc_max_value': 0.2, 'env_dtc_max_value': 0.4, 'grasped_normals_mask_max_dtc_value': 0.2, 'env_normals_mask_max_dtc_value': 0.4, 'clamp_dtc': True, 'max_contact_prob': 0.1, 'mask_normals_within_sdf': True, 'dtc_adaptive_normalization': False, 'adaptive_normals_mask': True, 'max_depth': 2.0, 'image_shape': [13, 180, 240], 'learnable_contact_preprocess_params': True, 'learning_rate': 0.0001, 'weight_decay': 0.0, 'contact_model_name': 'local_multitask_outhd64all_home_crop_h144w144d48_mask_ctxtmask_seed_220979_epoch_9', 'zero_centered': False}}, 'true_action_history': False}}, 'num_train_frames_bc': 50000, 'num_train_frames_drq': 1100000, 'stddev_schedule_drq': 'linear(1.0,0.1,100000)', 'task_name': 'FrankaInsertion-v1', 'num_train_frames_vinn': 25000, 'num_train_frames_diffusion': 1000000, 'num_train_epochs_bc': 5000, 'num_train_epochs_diffusion': 15000, 'validate_every_epochs_bc': 5, 'validate_every_epochs_diffusion': 250, 'validate_diffusion_on_action_loss_every_epochs': 250, 'train_eval_diffusion_on_action_loss_every_epochs': 250, 'check_topk_every_epochs': 5, 'check_topk_every_epochs_diffusion': 250, 'save_snapshot_every_epochs_diffusion': 1500, 'x_limit': [0.2, 0.7], 'y_limit': [-0.4, 0.4], 'z_limit': [-0.05, 0.55], 'home_displacement': [0.55, 0.0, 0.55, 180.0, 0.0, 0.0], 'enable_gripper': True, 'start_with_gripper_open': True, 'offset_mask': [1, 1, 1, 1, 1, 1], 'path_to_depth_extrinsics': '~/fish_leon/FISH/cfgs/camera_poses/camera_poses_L515/20240904-122305/color_tf_world.npy', 'feature_type': '180x240_1_RGB_D_2.0_acthst_hst6_out32_dwnkrnl3_dwnstrd2_dwnpd1', 'save_buffer': True, 'num_eval': 5, 'random_start': False, 'eval_starts': '/home/leonmkim/fish_leon/FISH/eval_starts/frankagym_pixels/FrankaInsertion-v1', 'num_valid_demos': None, 'load_checkpoint': True, 'checkpoint_epoch': 12000, 'load_residual_weight': False, 'checkpoint_root_dir': '/home/leonmkim/fish_leon/FISH', 'checkpoint_weight_dir': '/home/leonmkim/fish_leon/FISH/exp_local/frankagym_pixels/FrankaInsertion-v1/1045_0', 'residual_weight': '/home/leonmkim/fish_leon/FISH/weights/frankagym_pixels/FrankaInsertion-v1/weight.pt', 'final_experiment_dir': './exp_local/frankagym_pixels/FrankaInsertion-v1/1045_0/205317'}
+2024-12-29 20:53:23,744 INFO    MainThread:1355277 [wandb_init.py:init():619] starting backend
+2024-12-29 20:53:23,744 INFO    MainThread:1355277 [wandb_init.py:init():623] setting up manager
+2024-12-29 20:53:23,748 INFO    MainThread:1355277 [backend.py:_multiprocessing_setup():105] multiprocessing start_methods=fork,spawn,forkserver, using: spawn
+2024-12-29 20:53:23,750 INFO    MainThread:1355277 [wandb_init.py:init():631] backend started and connected
+2024-12-29 20:53:23,761 INFO    MainThread:1355277 [wandb_init.py:init():720] updated telemetry
+2024-12-29 20:53:23,767 INFO    MainThread:1355277 [wandb_init.py:init():753] communicating run to backend with 90.0 second timeout
+2024-12-29 20:53:24,005 INFO    MainThread:1355277 [wandb_run.py:_on_init():2435] communicating current version
+2024-12-29 20:53:24,117 INFO    MainThread:1355277 [wandb_run.py:_on_init():2444] got version response upgrade_message: "wandb version 0.19.1 is available!  To upgrade, please run:\n $ pip install wandb --upgrade"
+
+2024-12-29 20:53:24,117 INFO    MainThread:1355277 [wandb_init.py:init():804] starting run threads in backend
+2024-12-29 20:53:24,460 INFO    MainThread:1355277 [wandb_run.py:_console_start():2413] atexit reg
+2024-12-29 20:53:24,460 INFO    MainThread:1355277 [wandb_run.py:_redirect():2255] redirect: wrap_raw
+2024-12-29 20:53:24,460 INFO    MainThread:1355277 [wandb_run.py:_redirect():2320] Wrapping output streams.
+2024-12-29 20:53:24,460 INFO    MainThread:1355277 [wandb_run.py:_redirect():2345] Redirects installed.
+2024-12-29 20:53:24,461 INFO    MainThread:1355277 [wandb_init.py:init():847] run started, returning control to user process
+2024-12-29 20:53:24,462 INFO    MainThread:1355277 [wandb_run.py:_tensorboard_callback():1544] tensorboard callback: /home/leonmkim/fish_leon/FISH/exp_local/frankagym_pixels/FrankaInsertion-v1/1045_0/205317/tb, True
+2024-12-29 20:53:30,487 INFO    MainThread:1355277 [wandb_run.py:_config_callback():1382] config_cb None None {'grasped_obj_name': 'greece', 'left_book_slot': 'twodim'}
+2024-12-29 20:58:20,684 WARNING MsgRouterThr:1355277 [router.py:message_loop():77] message_loop has been closed

config.yaml

@@ -0,0 +1,548 @@
+root_dir: /mnt/kostas-graid/datasets/extrinsic_contact_data
+replay_buffer_size: 150000
+replay_buffer_num_workers: 2
+nstep: 3
+batch_size: 128
+seed: 0
+dataset_shuffle_seed: 5
+device: cuda
+save_video: true
+save_train_video: true
+use_tb: true
+use_wandb: true
+wandb_run_id: '1045_0'
+wandb_notes: '1045_0_'
+eval: false
+true_action_history: false
+train_pad_after: 4
+process_contact_features: ${eval}
+obs_type: pixels
+use_color: true
+use_depth: true
+use_masks: false
+mask_list:
+- EE_obj_mask
+mask_representation: channels
+crop_hw:
+- 144
+- 144
+crop_down_offset: 48
+color_crop_type: null
+depth_crop_type: null
+segmask_crop_type: null
+add_crop_binary_mask: false
+add_coord_conv_map: false
+use_context_color: false
+use_context_depth: false
+use_context_segmask: false
+context_color_crop_type: null
+context_depth_crop_type: null
+context_segmask_crop_type: null
+context_add_crop_binary_mask: false
+context_add_coord_conv_map: false
+use_contact_map: false
+use_sdf_maps: false
+use_normals_maps: false
+which_objects: both
+max_contact_prob: 0.1
+max_depth: 2.0
+grasped_dtc_max_value: 0.2
+env_dtc_max_value: 0.4
+grasped_normals_mask_max_dtc_value: 0.2
+env_normals_mask_max_dtc_value: 0.4
+clamp_dtc: true
+dtc_adaptive_normalization: false
+mask_normals_within_sdf: true
+adaptive_normals_mask: true
+learnable_contact_preprocess_params: true
+contact_model_name: local_multitask_outhd64all_home_crop_h144w144d48_mask_ctxtmask_seed_220979_epoch_9
+contact_estimation_model_ckpt_path: ~/fish_leon/contact_estimation/artifacts/175604_2/checkpoints/epoch=09-val_loss=0.00.ckpt
+encoder_type: small
+debug_timestamps: false
+open_loop: false
+action_trajectories: true
+stop_after_action: false
+interpolation_frequency: 25
+policy_frequency: 5
+wait_for_new_camera_frames: true
+baseline: false
+train_demo_idxs_list_or_num: -1
+log_train_every_steps: 25
+name_of_expert_demo: 120_240x320_all_twodim_left_to_right_annotated_start_idx_5hz_zstd7_EE_pxl_coords_expert_demos_imp_act
+expert_dataset_dirpath: ${root_dir}/FISH/expert_demos/${suite.name}/${task_name}/${name_of_expert_demo}
+store_dataset_in_memory: false
+expert_dataset: ${expert_dataset_dirpath}/demos.zarr
+action_key: ${oc.if_else:${action_trajectories}, 'action_trajectory_${interpolation_frequency}hz',
+  'action'}
+semantic_demo_grouping_name: semantic_demo_grouping.yaml
+semantic_demo_grouping: ${expert_dataset_dirpath}/${semantic_demo_grouping_name}
+include_groups_list:
+- greece_twodim_nominal
+- greece_twodim_recovery
+expert_dataset_config: ${expert_dataset_dirpath}/demo_config.yaml
+name_of_valid_demo: 120_240x320_all_twodim_left_to_right_annotated_start_idx_5hz_zstd7_EE_pxl_coords_expert_demos_imp_act
+valid_dataset_dir: ${root_dir}/FISH/expert_demos/${suite.name}/${task_name}/${name_of_valid_demo}/demos.zarr
+valid_demo_idxs_list_or_num: null
+val_num_groups: 0
+load_bc: ${agent.load_checkpoint}
+checkpoint_epoch_list:
+- 99
+- 199
+- 299
+- 399
+- 499
+- 599
+- 699
+- 799
+- 899
+- 999
+- 1249
+- 1499
+- 1749
+- 1999
+- 2999
+- 3999
+- 4999
+- 5999
+- 6999
+- 7999
+- 8999
+- 9999
+snapshot_root_dir: /mnt/grasp_high_usage/leonmkim/contact_estimation/FISH
+save_snapshot: true
+save_last_snapshot: true
+save_snapshot_when_done: true
+top_k_checkpoints: 5
+save_snapshot_link_to_weights_dir: deprecated
+bc_regularize: false
+bc_weight_type: qfilter
+experiment_dir: ./exp_local/${suite.name}_${obs_type}/${task_name}/${wandb_run_id}
+agent:
+  _target_: agent.diffusion_policy.DiffusionPolicyAgent
+  name: diffusion_policy
+  load_checkpoint: ${eval}
+  device: ${device}
+  n_obs_steps: ${.config.policy_cfg.n_obs_steps}
+  suite_name: ${suite.name}
+  obs_type: ${obs_type}
+  enable_arm: ${eval}
+  enable_camera: ${eval}
+  use_tb: ${use_tb}
+  desired_image_shape:
+  - 13
+  - 180
+  - 240
+  orig_cam_shape:
+  - 3
+  - 240
+  - 320
+  config:
+    _target_: agent.diffusion_policy.DiffusionPolicyAgentConfig
+    compile: false
+    device: ${device}
+    cam_resize_shape: ${agent.desired_image_shape}
+    orig_cam_shape: ${agent.orig_cam_shape}
+    policy_frequency: ${policy_frequency}
+    interpolation_frequency: ${interpolation_frequency}
+    policy_cfg:
+      _target_: lerobot.common.policies.diffusion.configuration_diffusion.DiffusionConfig
+      n_obs_steps: 1
+      horizon: 36
+      n_action_steps: ${agent.config.policy_cfg.horizon}
+      output_shapes:
+        action:
+        - 7
+      input_normalization_modes:
+        observation.image: mean_std
+        observation.state: min_max
+        observation.action_history: min_max
+      output_normalization_modes:
+        action: min_max
+      vision_backbone: resnet18
+      pretrained_backbone_weights: null
+      transforms:
+      - _target_: torchaug.transforms.RandomAffine
+        degrees:
+        - -5
+        - 5
+        translate:
+        - 0.05
+        - 0.05
+        batch_transform: true
+        num_chunks: -1
+        batch_inplace: true
+      - _target_: torchaug.transforms.RandomColorJitter
+        brightness: 0.3
+        contrast: 0.4
+        saturation: 0.5
+        hue: 0.08
+        batch_transform: true
+        num_chunks: -1
+        batch_inplace: true
+      use_group_norm: true
+      spatial_softmax_num_keypoints: 32
+      action_history_encoder_config:
+        _target_: lerobot.common.policies.diffusion.configuration_diffusion.Unet1dEncoderConfig
+        in_channels: 7
+        out_channels: 32
+        history_length: 6
+        kernel_size: ${agent.config.policy_cfg.kernel_size}
+        downsample_kernel_size: 3
+        downsample_stride: 2
+        downsample_padding: 1
+      down_dims:
+      - 256
+      - 512
+      - 1024
+      kernel_size: 5
+      n_groups: 8
+      diffusion_step_embed_dim: 128
+      use_film_scale_modulation: true
+      noise_scheduler_type: DDIM
+      beta_schedule: squaredcos_cap_v2
+      beta_start: 0.0001
+      beta_end: 0.02
+      prediction_type: epsilon
+      clip_sample: true
+      clip_sample_range: 1.0
+      num_train_timesteps: 50
+      num_inference_steps: 10
+      do_mask_loss_for_padding: false
+      input_shapes:
+        observation.image:
+        - 13
+        - 180
+        - 240
+        context_observation.image:
+        - 13
+        - 180
+        - 240
+        observation.state:
+        - 8
+        observation.action_history:
+        - 7
+    train_cfg:
+      _target_: utils.TrainConfig
+      lr: 0.0001
+      lr_scheduler: cosine
+      lr_warmup_steps: 500
+      adam_betas:
+      - 0.95
+      - 0.999
+      adam_eps: 1.0e-08
+      adam_weight_decay: 1.0e-06
+      grad_clip_norm: 10
+      offline_steps: ${num_train_frames_diffusion}
+      use_amp: true
+    observation_cfg:
+      _target_: agent.encoder.VisualFeatureSet
+      use_depth: ${use_depth}
+      use_color: ${use_color}
+      mask_input_dict:
+        _target_: agent.encoder.MaskInputDict
+        enable: ${use_masks}
+        representation: ${mask_representation}
+        mask_list: ${mask_list}
+      crop_input_config:
+        _target_: agent.encoder.CropInputConfig
+        color_crop_type: ${color_crop_type}
+        depth_crop_type: ${depth_crop_type}
+        segmask_crop_type: ${segmask_crop_type}
+        crop_hw: ${crop_hw}
+        crop_down_offset: ${crop_down_offset}
+        add_crop_binary_mask: ${add_crop_binary_mask}
+        add_coord_conv_map: ${add_coord_conv_map}
+      context_input_config:
+        _target_: agent.encoder.ContextInputConfig
+        use_color: ${use_context_color}
+        use_depth: ${use_context_depth}
+        mask_input_dict:
+          _target_: agent.encoder.MaskInputDict
+          enable: ${use_context_segmask}
+          representation: ${mask_representation}
+          mask_list: ${mask_list}
+        crop_input_config:
+          _target_: agent.encoder.CropInputConfig
+          color_crop_type: ${context_color_crop_type}
+          depth_crop_type: ${context_depth_crop_type}
+          segmask_crop_type: ${context_segmask_crop_type}
+          crop_hw: ${crop_hw}
+          crop_down_offset: ${crop_down_offset}
+          add_crop_binary_mask: ${context_add_crop_binary_mask}
+          add_coord_conv_map: ${context_add_coord_conv_map}
+      mask_soft_approx_scheduler_config:
+        _target_: agent.encoder.MaskSoftApproxSchedulerConfig
+        num_steps: 40000
+        initial_value: 10.0
+        final_value: 1000.0
+        interpolation_scheme: cosine
+      use_contact_map: ${use_contact_map}
+      use_sdf_maps: ${use_sdf_maps}
+      use_normals_maps: ${use_normals_maps}
+      which_objects: ${which_objects}
+      grasped_dtc_max_value: ${grasped_dtc_max_value}
+      env_dtc_max_value: ${env_dtc_max_value}
+      grasped_normals_mask_max_dtc_value: ${grasped_normals_mask_max_dtc_value}
+      env_normals_mask_max_dtc_value: ${env_normals_mask_max_dtc_value}
+      clamp_dtc: ${clamp_dtc}
+      max_contact_prob: ${max_contact_prob}
+      mask_normals_within_sdf: ${mask_normals_within_sdf}
+      dtc_adaptive_normalization: ${dtc_adaptive_normalization}
+      adaptive_normals_mask: ${adaptive_normals_mask}
+      max_depth: ${max_depth}
+      image_shape: ${agent.desired_image_shape}
+      learnable_contact_preprocess_params: ${learnable_contact_preprocess_params}
+      learning_rate: 0.0001
+      weight_decay: 0.0
+      contact_model_name: ${contact_model_name}
+      zero_centered: false
+suite:
+  suite: frankagym
+  name: frankagym
+  frame_stack: ${agent.n_obs_steps}
+  action_repeat: 1
+  discount: 0.99
+  hidden_dim: 1024
+  num_train_frames: 1000000
+  num_seed_frames: 0
+  num_train_epochs: 15000
+  validate_every_epochs: 250
+  validate_diffusion_on_action_loss_every_epochs: 250
+  train_eval_diffusion_on_action_loss_every_epochs: 250
+  check_topk_every_epochs: 250
+  save_snapshot_every_epochs: 1500
+  eval_every_frames: 2000
+  num_eval_episodes: 5
+  save_snapshot: true
+  wait_for_user_to_start_episode: true
+  task_make_fn:
+    _target_: suite.frankagym.make
+    name: ${task_name}
+    height: 240
+    width: 320
+    frame_stack: ${suite.frame_stack}
+    action_repeat: ${suite.action_repeat}
+    seed: ${seed}
+    enable_arm: ${agent.enable_arm}
+    enable_gripper: ${enable_gripper}
+    start_with_gripper_open: ${start_with_gripper_open}
+    enable_camera: ${agent.enable_camera}
+    path_to_depth_extrinsics: ${path_to_depth_extrinsics}
+    contact_estimation_model_ckpt_path: ${contact_estimation_model_ckpt_path}
+    x_limit: ${x_limit}
+    y_limit: ${y_limit}
+    z_limit: ${z_limit}
+    device: ${device}
+    interpolation_frequency: ${interpolation_frequency}
+    policy_frequency: ${policy_frequency}
+    debug_timestamps: ${debug_timestamps}
+    stop_after_action: ${stop_after_action}
+    open_loop: ${open_loop}
+    wait_for_new_camera_frames: ${wait_for_new_camera_frames}
+    action_key: ${action_key}
+    action_trajectory_horizon: ${agent.config.policy_cfg.horizon}
+    action_trajectories: ${action_trajectories}
+    path_to_zarr_dataset: ${expert_dataset}
+    agent_policy_cfg:
+      _target_: agent.diffusion_policy.DiffusionPolicyAgentConfig
+      compile: false
+      device: ${device}
+      cam_resize_shape: ${agent.desired_image_shape}
+      orig_cam_shape: ${agent.orig_cam_shape}
+      policy_frequency: ${policy_frequency}
+      interpolation_frequency: ${interpolation_frequency}
+      policy_cfg:
+        _target_: lerobot.common.policies.diffusion.configuration_diffusion.DiffusionConfig
+        n_obs_steps: 1
+        horizon: 36
+        n_action_steps: ${agent.config.policy_cfg.horizon}
+        output_shapes:
+          action:
+          - 7
+        input_normalization_modes:
+          observation.image: mean_std
+          observation.state: min_max
+          observation.action_history: min_max
+        output_normalization_modes:
+          action: min_max
+        vision_backbone: resnet18
+        pretrained_backbone_weights: null
+        transforms:
+        - _target_: torchaug.transforms.RandomAffine
+          degrees:
+          - -5
+          - 5
+          translate:
+          - 0.05
+          - 0.05
+          batch_transform: true
+          num_chunks: -1
+          batch_inplace: true
+        - _target_: torchaug.transforms.RandomColorJitter
+          brightness: 0.3
+          contrast: 0.4
+          saturation: 0.5
+          hue: 0.08
+          batch_transform: true
+          num_chunks: -1
+          batch_inplace: true
+        use_group_norm: true
+        spatial_softmax_num_keypoints: 32
+        action_history_encoder_config:
+          _target_: lerobot.common.policies.diffusion.configuration_diffusion.Unet1dEncoderConfig
+          in_channels: 7
+          out_channels: 32
+          history_length: 6
+          kernel_size: ${agent.config.policy_cfg.kernel_size}
+          downsample_kernel_size: 3
+          downsample_stride: 2
+          downsample_padding: 1
+        down_dims:
+        - 256
+        - 512
+        - 1024
+        kernel_size: 5
+        n_groups: 8
+        diffusion_step_embed_dim: 128
+        use_film_scale_modulation: true
+        noise_scheduler_type: DDIM
+        beta_schedule: squaredcos_cap_v2
+        beta_start: 0.0001
+        beta_end: 0.02
+        prediction_type: epsilon
+        clip_sample: true
+        clip_sample_range: 1.0
+        num_train_timesteps: 50
+        num_inference_steps: 10
+        do_mask_loss_for_padding: false
+        input_shapes:
+          observation.image:
+          - 13
+          - 180
+          - 240
+          context_observation.image:
+          - 13
+          - 180
+          - 240
+          observation.state:
+          - 8
+          observation.action_history:
+          - 7
+      train_cfg:
+        _target_: utils.TrainConfig
+        lr: 0.0001
+        lr_scheduler: cosine
+        lr_warmup_steps: 500
+        adam_betas:
+        - 0.95
+        - 0.999
+        adam_eps: 1.0e-08
+        adam_weight_decay: 1.0e-06
+        grad_clip_norm: 10
+        offline_steps: ${num_train_frames_diffusion}
+        use_amp: true
+      observation_cfg:
+        _target_: agent.encoder.VisualFeatureSet
+        use_depth: ${use_depth}
+        use_color: ${use_color}
+        mask_input_dict:
+          _target_: agent.encoder.MaskInputDict
+          enable: ${use_masks}
+          representation: ${mask_representation}
+          mask_list: ${mask_list}
+        crop_input_config:
+          _target_: agent.encoder.CropInputConfig
+          color_crop_type: ${color_crop_type}
+          depth_crop_type: ${depth_crop_type}
+          segmask_crop_type: ${segmask_crop_type}
+          crop_hw: ${crop_hw}
+          crop_down_offset: ${crop_down_offset}
+          add_crop_binary_mask: ${add_crop_binary_mask}
+          add_coord_conv_map: ${add_coord_conv_map}
+        context_input_config:
+          _target_: agent.encoder.ContextInputConfig
+          use_color: ${use_context_color}
+          use_depth: ${use_context_depth}
+          mask_input_dict:
+            _target_: agent.encoder.MaskInputDict
+            enable: ${use_context_segmask}
+            representation: ${mask_representation}
+            mask_list: ${mask_list}
+          crop_input_config:
+            _target_: agent.encoder.CropInputConfig
+            color_crop_type: ${context_color_crop_type}
+            depth_crop_type: ${context_depth_crop_type}
+            segmask_crop_type: ${context_segmask_crop_type}
+            crop_hw: ${crop_hw}
+            crop_down_offset: ${crop_down_offset}
+            add_crop_binary_mask: ${context_add_crop_binary_mask}
+            add_coord_conv_map: ${context_add_coord_conv_map}
+        mask_soft_approx_scheduler_config:
+          _target_: agent.encoder.MaskSoftApproxSchedulerConfig
+          num_steps: 40000
+          initial_value: 10.0
+          final_value: 1000.0
+          interpolation_scheme: cosine
+        use_contact_map: ${use_contact_map}
+        use_sdf_maps: ${use_sdf_maps}
+        use_normals_maps: ${use_normals_maps}
+        which_objects: ${which_objects}
+        grasped_dtc_max_value: ${grasped_dtc_max_value}
+        env_dtc_max_value: ${env_dtc_max_value}
+        grasped_normals_mask_max_dtc_value: ${grasped_normals_mask_max_dtc_value}
+        env_normals_mask_max_dtc_value: ${env_normals_mask_max_dtc_value}
+        clamp_dtc: ${clamp_dtc}
+        max_contact_prob: ${max_contact_prob}
+        mask_normals_within_sdf: ${mask_normals_within_sdf}
+        dtc_adaptive_normalization: ${dtc_adaptive_normalization}
+        adaptive_normals_mask: ${adaptive_normals_mask}
+        max_depth: ${max_depth}
+        image_shape: ${agent.desired_image_shape}
+        learnable_contact_preprocess_params: ${learnable_contact_preprocess_params}
+        learning_rate: 0.0001
+        weight_decay: 0.0
+        contact_model_name: ${contact_model_name}
+        zero_centered: false
+    true_action_history: ${true_action_history}
+num_train_frames_bc: 50000
+num_train_frames_drq: 1100000
+stddev_schedule_drq: linear(1.0,0.1,100000)
+task_name: FrankaInsertion-v1
+num_train_frames_vinn: 25000
+num_train_frames_diffusion: 1000000
+num_train_epochs_bc: 5000
+num_train_epochs_diffusion: 15000
+validate_every_epochs_bc: 5
+validate_every_epochs_diffusion: 250
+validate_diffusion_on_action_loss_every_epochs: 250
+train_eval_diffusion_on_action_loss_every_epochs: 250
+check_topk_every_epochs: 5
+check_topk_every_epochs_diffusion: 250
+save_snapshot_every_epochs_diffusion: 1500
+x_limit:
+- 0.2
+- 0.7
+y_limit:
+- -0.4
+- 0.4
+z_limit:
+- -0.05
+- 0.55
+home_displacement:
+- 0.55
+- 0.0
+- 0.55
+- 180.0
+- 0.0
+- 0.0
+enable_gripper: true
+start_with_gripper_open: true
+offset_mask:
+- 1
+- 1
+- 1
+- 1
+- 1
+- 1
+path_to_depth_extrinsics: ~/fish_leon/FISH/cfgs/camera_poses/camera_poses_L515/20240904-122305/color_tf_world.npy
+feature_type: 180x240_1_RGB_D_2.0_acthst_hst6_out32_dwnkrnl3_dwnstrd2_dwnpd1

snapshot_10500.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a969ffbdc5d4f82b42645ee428ed42ce53516504ab5469a4a8b03ada3af4386d
+size 910064728

snapshot_12000.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c57f96ed3083c018508a64fca7c89aa2aa61eb201819d2b71e84551952c8413f
+size 910064728

snapshot_12999.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:83031bd9a71dace532377d29868da3976cae276f497779868fb835c55d98a62a
+size 910064728

snapshot_13500.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a6dc9172d30a9852821d1a0d74fc847537782edceca8e6d76dd2af5224f4816e
+size 910064728