resume: false
device: cuda
use_amp: false
seed: 0
dataset_repo_id: iantc104/gaze_real_occluded_insertion
video_backend: pyav
training:
  offline_steps: 100000
  num_workers: 4
  batch_size: 8
  eval_freq: -1
  log_freq: 100
  save_checkpoint: true
  save_freq: 10000
  online_steps: 0
  online_rollout_n_episodes: 1
  online_rollout_batch_size: 1
  online_steps_between_rollouts: 1
  online_sampling_ratio: 0.5
  online_env_seed: null
  online_buffer_capacity: null
  online_buffer_seed_size: 0
  do_online_rollout_async: false
  image_transforms:
    enable: false
    max_num_transforms: 3
    random_order: false
    brightness:
      weight: 1
      min_max:
      - 0.8
      - 1.2
    contrast:
      weight: 1
      min_max:
      - 0.8
      - 1.2
    saturation:
      weight: 1
      min_max:
      - 0.5
      - 1.5
    hue:
      weight: 1
      min_max:
      - -0.05
      - 0.05
    sharpness:
      weight: 1
      min_max:
      - 0.8
      - 1.2
  lr: 1.0e-05
  lr_backbone: 1.0e-05
  weight_decay: 0.0001
  grad_clip_norm: 10
  delta_timestamps:
    action:
    - 0.0
    - 0.030303030303030304
    - 0.06060606060606061
    - 0.09090909090909091
    - 0.12121212121212122
    - 0.15151515151515152
    - 0.18181818181818182
    - 0.21212121212121213
    - 0.24242424242424243
    - 0.2727272727272727
    - 0.30303030303030304
    - 0.3333333333333333
    - 0.36363636363636365
    - 0.3939393939393939
    - 0.42424242424242425
    - 0.45454545454545453
    - 0.48484848484848486
    - 0.5151515151515151
    - 0.5454545454545454
    - 0.5757575757575758
    - 0.6060606060606061
    - 0.6363636363636364
    - 0.6666666666666666
    - 0.696969696969697
    - 0.7272727272727273
    - 0.7575757575757576
    - 0.7878787878787878
    - 0.8181818181818182
    - 0.8484848484848485
    - 0.8787878787878788
    - 0.9090909090909091
    - 0.9393939393939394
    - 0.9696969696969697
    - 1.0
    - 1.0303030303030303
    - 1.0606060606060606
    - 1.0909090909090908
    - 1.121212121212121
    - 1.1515151515151516
    - 1.1818181818181819
    - 1.2121212121212122
    - 1.2424242424242424
    - 1.2727272727272727
    - 1.303030303030303
    - 1.3333333333333333
    - 1.3636363636363635
    - 1.393939393939394
    - 1.4242424242424243
    - 1.4545454545454546
    - 1.4848484848484849
eval:
  n_episodes: 50
  batch_size: 10
  use_async_envs: false
wandb:
  enable: true
  disable_artifact: true
  project: real_occluded_insertion
  notes: ''
fps: 33
env:
  name: real_world
  task: null
  state_dim: 21
  action_dim: 21
  fps: ${fps}
override_dataset_stats:
  observation.images.left_eye_cam:
    mean:
    - - - 0.485
    - - - 0.456
    - - - 0.406
    std:
    - - - 0.229
    - - - 0.224
    - - - 0.225
  observation.images.right_eye_cam:
    mean:
    - - - 0.485
    - - - 0.456
    - - - 0.406
    std:
    - - - 0.229
    - - - 0.224
    - - - 0.225
policy:
  name: act
  n_obs_steps: 1
  chunk_size: 50
  n_action_steps: 50
  input_shapes:
    observation.images.left_eye_cam:
    - 3
    - 480
    - 640
    observation.images.right_eye_cam:
    - 3
    - 480
    - 640
    observation.state:
    - ${env.state_dim}
  output_shapes:
    action:
    - ${env.action_dim}
  input_normalization_modes:
    observation.images.left_eye_cam: mean_std
    observation.images.right_eye_cam: mean_std
    observation.state: mean_std
  output_normalization_modes:
    action: mean_std
  image_size:
  - 336
  - 448
  vision_backbone: dinov2
  pretrained_backbone_weights: dinov2_vits14_reg
  freeze_backbone: true
  pre_norm: false
  dim_model: 512
  n_heads: 8
  dim_feedforward: 3200
  feedforward_activation: relu
  n_encoder_layers: 4
  n_decoder_layers: 1
  use_vae: true
  latent_dim: 32
  n_vae_encoder_layers: 4
  temporal_ensemble_coeff: null
  dropout: 0.1
  kl_weight: 10.0