config.json

{
  "architectures": [
    "Pix2StructForConditionalGeneration"
  ],
  "decoder_start_token_id": 0,
  "dtype": "float32",
  "eos_token_id": 1,
  "initializer_factor": 1.0,
  "initializer_range": 0.02,
  "is_encoder_decoder": true,
  "is_vqa": false,
  "model_type": "pix2struct",
  "pad_token_id": 0,
  "projection_dim": 64,
  "text_config": {
    "add_cross_attention": false,
    "attention_dropout": 0.1,
    "bos_token_id": 0,
    "d_ff": 2048,
    "d_kv": 3,
    "decoder_start_token_id": 0,
    "dense_act_fn": "gelu_new",
    "dropout": 0.1,
    "dropout_rate": 0.1,
    "eos_token_id": 1,
    "hidden_size": 12,
    "initializer_factor": 1.0,
    "initializer_range": 0.02,
    "intermediate_size": 37,
    "is_decoder": true,
    "is_encoder_decoder": true,
    "layer_norm_epsilon": 1e-06,
    "max_position_embeddings": 512,
    "model_type": "pix2struct_text_model",
    "num_heads": 4,
    "num_layers": 2,
    "pad_token_id": 0,
    "projection_dim": 32,
    "relative_attention_max_distance": 128,
    "relative_attention_num_buckets": 32,
    "tie_word_embeddings": false,
    "use_cache": false,
    "vocab_size": 50344
  },
  "tie_word_embeddings": false,
  "transformers_version": "5.16.0.dev0",
  "vision_config": {
    "attention_dropout": 0.1,
    "d_ff": 2048,
    "d_kv": 64,
    "dense_act_fn": "gelu_new",
    "dropout": 0.1,
    "dropout_rate": 0.0,
    "hidden_size": 12,
    "image_size": 30,
    "initializer_factor": 1.0,
    "initializer_range": 0.02,
    "intermediate_size": 37,
    "layer_norm_eps": 1e-06,
    "model_type": "pix2struct_vision_model",
    "num_attention_heads": 4,
    "num_channels": 3,
    "num_hidden_layers": 2,
    "patch_embed_hidden_size": 12,
    "patch_size": 2,
    "projection_dim": 32,
    "relative_attention_max_distance": 128,
    "relative_attention_num_buckets": 32,
    "seq_len": 4096
  }
}