config.json

{
  "action_space": null,
  "freeze_mineclip": true,
  "mineclip_kwargs": {
    "arch": "vit_base_p16_fz.v2.t2",
    "hidden_dim": 512,
    "image_feature_dim": 512,
    "mlp_adapter_spec": "v0-2.t0",
    "pool_type": "attn.d2.nh8.glusw",
    "resolution": [
      160,
      256
    ]
  },
  "policy_kwargs": {
    "attention_heads": 16,
    "attention_mask_style": "clipped_causal",
    "attention_memory_size": 256,
    "diff_mlp_embedding": false,
    "hidsize": 2048,
    "img_shape": [
      128,
      128,
      3
    ],
    "impala_chans": [
      16,
      32,
      32
    ],
    "impala_kwargs": {
      "post_pool_groups": 1
    },
    "impala_width": 8,
    "init_norm_kwargs": {
      "batch_norm": false,
      "group_norm_groups": 1
    },
    "n_recurrence_layers": 4,
    "only_img_input": true,
    "pointwise_ratio": 4,
    "pointwise_use_activation": false,
    "recurrence_is_residual": true,
    "recurrence_type": "transformer",
    "timesteps": 128,
    "use_pointwise_layer": true,
    "use_pre_lstm_ln": false
  },
  "prior_kwargs": {
    "hidden_dim": 512,
    "input_dim": 512,
    "latent_dim": 512
  }
}