action_tokenizer_hf/config.json

{
  "architectures": [
    "ActionCodecV2Model"
  ],
  "block_dct_block_size": 8,
  "c_mults": [
    1,
    2,
    2
  ],
  "codebook_dim": 8,
  "codebook_size": 4096,
  "commitment_loss_weight": 0.25,
  "consistency_delta_max_begin": 1,
  "consistency_delta_max_end": 3,
  "consistency_eps_begin": 0.02,
  "consistency_eps_end": 0.1,
  "consistency_layer_weights_begin": [
    0.0,
    0.0,
    0.0,
    0.0
  ],
  "consistency_layer_weights_end": [
    1.0,
    0.7,
    0.3,
    0.1
  ],
  "consistency_loss_weight": 0.0,
  "consistency_schedule_begin_fraction": 0.2,
  "consistency_schedule_end_fraction": 0.8,
  "conv_in_action_kernel": 2,
  "dim_heads": 64,
  "dtype": "float32",
  "ema_decay": 0.95,
  "encoder_channels": 256,
  "ffn_mult": 4.0,
  "fsq_levels": [
    4,
    4,
    4,
    4,
    4,
    4
  ],
  "horizon": 32,
  "horizon_patch_size": 8,
  "latent_dim": 128,
  "layer_scale_init": 0.01,
  "max_component_dim": 9,
  "model_type": "actioncodecv2",
  "n_codebooks": 4,
  "num_heads": 8,
  "quantizer_dropout": 0.5,
  "reconstruction_loss_weight": 1.0,
  "rope_base": 10000,
  "strides": [
    [
      1,
      1
    ],
    [
      2,
      1
    ],
    [
      2,
      1
    ]
  ],
  "threshold_ema_dead": 2,
  "transformer_depths": [
    2,
    2,
    2
  ],
  "transformers_version": "4.57.1",
  "use_block_dct": true,
  "use_fsq": false,
  "use_layer_scale": true,
  "use_lipschitz_encoder": false,
  "use_qk_norm": true,
  "use_rotation_trick": false
}