{
  "_auto_class": null,
  "_name_or_path": "/scratch/giuliano_albanese/spear-hf",
  "architectures": [
    "SPEAR1"
  ],
  "attribute_map": {},
  "auto_map": {
    "AutoConfig": "configuration_spear.SPEAR1Config",
    "AutoModel": "modeling_spear.SPEAR1"
  },
  "autoclass": "barrel.pipes.vlams.models.vlams.vlam.VLAM",
  "base_config_key": "",
  "control_module_config": {
    "control_decoder_config": {
      "block_config": {
        "activation": "GELU",
        "attn_implementation": "sdpa",
        "dropout": 0.0,
        "feature_size": 1024,
        "head_dim": 256,
        "hidden_size": 4096,
        "norm": "RMSNorm",
        "num_heads": 8,
        "num_kv_heads": 1,
        "position_embed_config": {
          "base": 10000,
          "cached": true,
          "embedding_dim": 256,
          "num_embeddings": 512
        }
      },
      "num_blocks": 18
    },
    "noised_control_proj_config": {
      "activation": "SiLU",
      "layers": [
        8,
        2048,
        1024,
        1024
      ],
      "norm": null,
      "time_embed": {
        "activation": "SiLU",
        "layers": [],
        "learnable_features": false,
        "max_period": 10000.0,
        "norm": null,
        "num_features": 1024
      }
    },
    "robot_state_proj_config": {
      "activation": "SiLU",
      "fourier": false,
      "layers": [
        8,
        1024
      ],
      "mode": "ee_pose_gripper"
    },
    "rotation_components": 4,
    "token_size": 1024
  },
  "is_composition": false,
  "model_type": "spear1",
  "processor_config": {
    "control_io_config": {
      "future_control_offset_sec": 0.0,
      "future_controls_sequence_length": 5,
      "future_controls_sequence_stride_sec": 0.2,
      "future_frames_sequence_length": 1,
      "future_frames_sequence_stride_sec": null,
      "past_frames_sequence_length": 1,
      "past_frames_stride_sec": null,
      "past_scalars_sequence_length": 1,
      "past_scalars_stride_sec": null,
      "sequence_frames": 1,
      "sequence_frames_stride_sec": null
    },
    "control_stats_path": "barrel/pipes/vlams/types/control_stats.yaml",
    "control_tokenizer_config": {},
    "delta_controls": true,
    "distribution_hyperparams": {
      "alpha": 1.5,
      "beta": 1.0
    },
    "eef_control_frame": false,
    "image_resize": "smart",
    "joints_norm": {
      "high": [
        3.141592653589793,
        3.141592653589793,
        3.141592653589793,
        3.141592653589793,
        3.141592653589793,
        3.141592653589793,
        3.141592653589793
      ],
      "low": [
        -3.141592653589793,
        -3.141592653589793,
        -3.141592653589793,
        -3.141592653589793,
        -3.141592653589793,
        -3.141592653589793,
        -3.141592653589793
      ]
    },
    "num_inference_steps": 10,
    "obs_rotation_norm": "none",
    "obs_translation_norm": "bounds_q99",
    "observation_stats_path": "barrel/pipes/vlams/types/observation_stats.yaml",
    "r0_distribution": "uniform",
    "rotation_format": "quaternion",
    "rotation_norm": "none",
    "sig_min": 0.001,
    "timestep_distribution": "beta",
    "translation_norm": {
      "high": [
        0.04,
        0.04,
        0.04
      ],
      "low": [
        -0.04,
        -0.04,
        -0.04
      ]
    }
  },
  "sub_configs": {},
  "torch_dtype": "float32",
  "transformers_version": "4.47.0",
  "vlm_config": {
    "attn_implementation": "flash_attention_2",
    "depth_tokens": 1024,
    "lm_head": false,
    "mean_resizing": false,
    "model_id": "google/paligemma-3b-mix-224",
    "paligemma_3d_config": {
      "depth_config": {
        "hf_filename": "moge/moge-vit-large-patch-14-backbone.pt",
        "hf_hub_repo": "nikonikolov/vlams"
      },
      "depth_layers": 4,
      "depth_only": false,
      "mask_prob": 0.0,
      "projection": "features_add"
    },
    "processor_config": {
      "image_sizes": {
        "main": {
          "height": 210,
          "width": 280
        },
        "wrist": {
          "height": 112,
          "width": 112
        }
      },
      "image_token": "<image>",
      "max_language_tokens": 75
    },
    "train_only_depth_tokens": false
  }
}