{
  "dim_output": 512,
  "fps": 1,
  "input_normalization_modes": {
    "observation.images.wrist": "mean_std",
    "observation.state": "mean_std"
  },
  "input_shapes": {
    "observation.images.wrist": [
      3,
      120,
      160
    ],
    "observation.state": [
      22
    ]
  },
  "n_obs_steps": 3,
  "predictor_activation": "ReLU",
  "predictor_cnn_kernel_size": [
    10,
    6,
    4
  ],
  "predictor_cnn_out_channels": [
    16,
    32,
    64
  ],
  "predictor_cnn_padding": [
    0,
    0,
    0
  ],
  "predictor_cnn_spatial_softmax_num_keypoints": 32,
  "predictor_cnn_stride": [
    5,
    3,
    2
  ],
  "predictor_cnn_use_batchnorm": false,
  "predictor_cnn_use_maxpool": false,
  "predictor_cnn_use_spatial_softmax": true,
  "predictor_mlp_hidden_sizes": [
    512,
    512
  ],
  "target_activation": "ReLU",
  "target_cnn_kernel_size": [
    10,
    6,
    4
  ],
  "target_cnn_out_channels": [
    16,
    32,
    64
  ],
  "target_cnn_padding": [
    0,
    0,
    0
  ],
  "target_cnn_spatial_softmax_num_keypoints": 32,
  "target_cnn_stride": [
    5,
    3,
    2
  ],
  "target_cnn_use_batchnorm": false,
  "target_cnn_use_maxpool": false,
  "target_cnn_use_spatial_softmax": true,
  "target_mlp_hidden_sizes": [
    512,
    512
  ]
}