config.json

{
  "model_type": "vision-encoder-decoder",
  "encoder": {
    "model_type": "paligemma",
    "architecture": "paligemma-3b",
    "image_size": [224, 224],
    "input_size": [224, 224],
    "num_hidden_layers": 12,
    "num_attention_heads": 16,
    "hidden_size": 1024,
    "patch_size": 16,
    "dropout": 0.1
  },
  "decoder": {
    "model_type": "bert",
    "vocab_size": 30522,
    "max_position_embeddings": 512,
    "num_hidden_layers": 12,
    "num_attention_heads": 12,
    "hidden_size": 768,
    "initializer_range": 0.02,
    "layer_norm_eps": 1e-12,
    "hidden_act": "gelu",
    "attention_probs_dropout_prob": 0.1,
    "hidden_dropout_prob": 0.1
  },
  "max_length": 64,
  "decoder_start_token_id": 101,
  "pad_token_id": 0,
  "eos_token_id": 102,
  "bos_token_id": 101
}