motion_tokenizer/config.json

{
  "_class_name": "VQVAEVermo2DTK",
  "_name_or_path": "checkpoints/vermo_vqvae2d_16k_rescale_iter47k",
  "attn_scales": [],
  "base_dim": 96,
  "decoder_base_dim": null,
  "dim_mult": [
    1,
    2,
    4,
    4
  ],
  "dropout": 0.0,
  "in_channels": 6,
  "is_residual": false,
  "num_res_blocks": 2,
  "out_channels": 6,
  "quantizer_cfg": {
    "dim": 16,
    "levels": [
      8,
      8,
      8,
      6,
      5
    ],
    "type": "FSQuantizer"
  },
  "scale_factor_temporal": 4,
  "temporal_downsample": [
    false,
    true,
    true
  ],
  "token_rescale": true,
  "use_rollout_trans": true,
  "use_static": false,
  "z_dim": 16
}