transformer/config.json

{
  "_class_name": "MaskGiTUViT",
  "_version": "0.0.1",
  "add_cond_embeds": true,
  "add_cross_attention": true,
  "add_micro_cond_embeds": true,
  "attention_dropout": 0.0,
  "block_has_attention": [
    true
  ],
  "block_num_heads": [
    12
  ],
  "block_out_channels": [
    768
  ],
  "codebook_size": 8192,
  "cond_embed_dim": 768,
  "encoder_hidden_size": 768,
  "ffn_type": "glu",
  "hidden_dropout": 0.0,
  "hidden_size": 1024,
  "in_channels": 768,
  "initializer_range": 0.02,
  "intermediate_size": 2816,
  "layer_norm_before_mlm": false,
  "layer_norm_embedddings": false,
  "layer_norm_eps": 0.000001,
  "learn_uncond_embeds": false,
  "ln_elementwise_affine": true,
  "mask_token_id": 8255,
  "max_position_embeddings": 256,
  "micro_cond_embed_dim": 1280,
  "micro_cond_encode_dim": 256,
  "norm_type": "rmsnorm",
  "num_attention_heads": 16,
  "num_classes": null,
  "num_hidden_layers": 22,
  "num_res_blocks": 3,
  "num_vq_tokens": 256,
  "patch_size": 1,
  "project_encoder_hidden_states": true,
  "res_ffn_factor": 4,
  "use_bias": false,
  "use_codebook_size_for_output": true,
  "use_empty_embeds_for_uncond": true,
  "use_encoder_layernorm": false,
  "use_normformer": false,
  "use_position_embeddings": false,
  "use_vannilla_resblock": false,
  "vocab_size": 8256,
  "xavier_init_embed": true
}