{ "architectures": [ "GotOcr2Model" ], "dtype": "float32", "image_seq_length": 576, "image_token_index": 1, "model_type": "got_ocr2", "text_config": { "attention_dropout": 0.0, "bos_token_id": 2, "eos_token_id": 3, "hidden_act": "silu", "hidden_size": 128, "initializer_range": 0.02, "intermediate_size": 37, "layer_types": [ "full_attention", "full_attention" ], "max_position_embeddings": 512, "max_window_layers": 28, "mlp_ratio": 4, "model_type": "qwen2", "num_attention_heads": 4, "num_hidden_layers": 2, "num_key_value_heads": 2, "output_channels": 64, "pad_token_id": 4, "rms_norm_eps": 1e-06, "rope_parameters": { "rope_theta": 10000, "rope_type": "default" }, "sliding_window": null, "tie_word_embeddings": true, "use_cache": true, "use_sliding_window": false, "vocab_size": 99 }, "tie_word_embeddings": true, "transformers_version": "5.16.0.dev0", "vision_config": { "attention_dropout": 0.0, "global_attn_indexes": [ 2, 5, 8, 11 ], "hidden_act": "quick_gelu", "hidden_size": 32, "image_size": 64, "initializer_range": 1e-10, "layer_norm_eps": 1e-06, "mlp_dim": 128, "model_type": "", "num_attention_heads": 4, "num_channels": 3, "num_hidden_layers": 2, "output_channels": 64, "patch_size": 2, "qkv_bias": true, "use_abs_pos": true, "use_rel_pos": true, "window_size": 14 } }