decoder_config.json

{
  "base_model_name": "Qwen/Qwen3.5-4B",
  "vlm_dim": 2560,
  "decoder_dim": 1024,
  "num_steps": 6,
  "action_dim_per_step": 20,
  "num_layers": 2,
  "nhead": 8,
  "ff_mult": 4,
  "dropout": 0.0,
  "task_start_id": 248077,
  "task_end_id": 248078,
  "special_tokens": [
    "<|task_start|>",
    "<|task_end|>"
  ],
  "num_buttons": 16,
  "button_names": [
    "south",
    "east",
    "west",
    "north",
    "left_shoulder",
    "right_shoulder",
    "left_trigger",
    "right_trigger",
    "left_thumb",
    "right_thumb",
    "dpad_up",
    "dpad_down",
    "dpad_left",
    "dpad_right",
    "start",
    "back"
  ]
}