config.json

{
  "architectures": [
    "GotOcr2ForConditionalGeneration"
  ],
  "ignore_index": -100,
  "image_seq_length": 576,
  "image_token_index": 151859,
  "model_type": "got_ocr2",
  "text_config": {
    "attention_dropout": 0.0,
    "hidden_act": "silu",
    "hidden_size": 32,
    "initializer_range": 0.02,
    "intermediate_size": 32,
    "max_position_embeddings": 32768,
    "max_window_layers": 1,
    "model_type": "qwen2",
    "num_attention_heads": 4,
    "num_hidden_layers": 2,
    "num_key_value_heads": 4,
    "rms_norm_eps": 1e-06,
    "rope_scaling": null,
    "rope_theta": 1000000.0,
    "sliding_window": 4096,
    "tie_word_embeddings": true,
    "torch_dtype": "float32",
    "use_cache": true,
    "use_sliding_window": false,
    "vocab_size": 151860
  },
  "torch_dtype": "float32",
  "transformers_version": "4.50.0.dev0",
  "vision_config": {
    "attention_dropout": 0.0,
    "global_attn_indexes": [
      2,
      5,
      8,
      11
    ],
    "hidden_act": "quick_gelu",
    "hidden_size": 32,
    "image_size": 128,
    "initializer_range": 1e-10,
    "layer_norm_eps": 1e-06,
    "mlp_dim": 16,
    "model_type": "",
    "num_attention_heads": 4,
    "num_channels": 3,
    "num_hidden_layers": 2,
    "output_channels": 8,
    "patch_size": 2,
    "qkv_bias": true,
    "torch_dtype": "float32",
    "use_abs_pos": true,
    "use_rel_pos": true,
    "window_size": 14
  }
}