config.json

{
  "action_dim": 7,
  "architectures": [
    "BCTransformerPolicy"
  ],
  "chunk_size": 64,
  "d_model": 256,
  "dim_feedforward": 1024,
  "dropout": 0.1,
  "feature_aggregation": "spatial_softmax",
  "image_size": [
    256,
    256
  ],
  "imagenet_norm": true,
  "loss_type": "mse",
  "model_type": "bc_transformer",
  "nhead": 8,
  "num_cameras": 2,
  "num_kp": 32,
  "num_layers": 4,
  "predict_relative_actions": false,
  "share_rgb_model": false,
  "state_dim": 7,
  "torch_dtype": "float32",
  "transformers_version": "4.45.2",
  "vision_feature_dim": 128,
  "vision_frozen": false,
  "vision_model_name": "resnet18",
  "vision_pretrained": true
}