{
  "_class_name": "MaskGiTUViT_v2",
  "_version": "0.0.1",
  "adapter_proj_dim": 32,
  "add_cond_embeds": true,
  "add_micro_cond_embeds": true,
  "attention_dropout": 0.0,
  "block_num_heads": 12,
  "block_out_channels": [
    768
  ],
  "codebook_size": 8192,
  "cond_embed_dim": 768,
  "decay": 0.9999,
  "encoder_hidden_size": 768,
  "force_down_up_sample": true,
  "hidden_dropout": 0.0,
  "hidden_size": 1024,
  "in_channels": 768,
  "intermediate_size": 2816,
  "inv_gamma": 1.0,
  "is_adapter_shared": false,
  "layer_norm_eps": 1e-06,
  "ln_elementwise_affine": true,
  "mask_token_id": 8255,
  "micro_cond_embed_dim": 1280,
  "micro_cond_encode_dim": 256,
  "min_decay": 0.0,
  "norm_type": "rmsnorm",
  "num_attention_heads": 16,
  "num_hidden_layers": 22,
  "num_res_blocks": 3,
  "optimization_step": 600,
  "power": 0.6666666666666666,
  "update_after_step": 0,
  "use_adapter": true,
  "use_bias": false,
  "use_ema_warmup": false,
  "use_fused_mlp": false,
  "use_fused_residual_norm": true,
  "vocab_size": 8256
}