data/mar_ckpt/langtable/config.json

{
  "Diffusion": true,
  "S": 1024,
  "T": 12,
  "action_contrastive_loss": false,
  "action_domains": [
    "language_table"
  ],
  "action_loss_weight": 0.5,
  "action_network": "conconcat+modulatecat",
  "action_stats": [
    [
      [
        0.00014842326345387846,
        -0.0005635050474666059
      ],
      [
        0.030163198709487915,
        0.042305462062358856
      ]
    ]
  ],
  "action_token_size": 64,
  "arch": "STTransformerDecoder",
  "attn_drop": 0.1,
  "attn_dropout": 0.1,
  "buffer_size": 64,
  "d_action": 28,
  "d_actions": [
    2
  ],
  "d_model": 256,
  "dataloader_apply_corruption": false,
  "dataloader_apply_mask": true,
  "dataloader_mask_ratio_min": 0.1,
  "diffloss_d": 4,
  "diffloss_w": 1024,
  "diffusion_batch_mul": 1,
  "dim": 512,
  "drop_action_ratio": 0.0,
  "factored_vocab_size": 512,
  "grad_checkpointing": false,
  "image_vocab_size": null,
  "init_actions": true,
  "jointly_predict_actions": false,
  "jointly_predict_states": true,
  "label_drop_prob": 0.5,
  "mask_ratio_min": 0.7,
  "maskgit_steps": 16,
  "max_corrupt_rate": 0.2,
  "mlp_bias": false,
  "mlp_drop": 0.05,
  "mlp_ratio": 4.0,
  "non_mlm_ratio": 0.2,
  "num_factored_vocabs": 2,
  "num_heads": 8,
  "num_layers": 32,
  "num_prompt_frames": 4,
  "num_sampling_steps": "100",
  "patch_size": 2,
  "predict_unmask": false,
  "proj_bias": true,
  "proj_dropout": 0.1,
  "qk_norm": false,
  "qkv_bias": true,
  "random_dummy_action": true,
  "shared_action_mlps": true,
  "use_actions": true,
  "use_mup": false,
  "vae_embed_dim": 4,
  "vae_stride": 1
}