{ "architectures": [ "PaddleOCRVLForConditionalGeneration" ], "dtype": "float32", "image_token_id": 100295, "model_type": "paddleocr_vl", "text_config": { "bos_token_id": 1, "eos_token_id": 2, "head_dim": 128, "hidden_act": "silu", "hidden_dropout_prob": 0.0, "hidden_size": 32, "ignored_index": -100, "image_token_id": 100295, "initializer_range": 0.02, "intermediate_size": 32, "max_position_embeddings": 512, "model_type": "paddleocr_vl_text", "num_attention_heads": 4, "num_hidden_layers": 2, "num_key_value_heads": 2, "pad_token_id": 0, "rms_norm_eps": 1e-05, "rope_parameters": { "mrope_section": [ 16, 24, 24 ], "rope_theta": 500000, "rope_type": "default", "type": "default" }, "tie_word_embeddings": false, "use_bias": false, "use_cache": true, "vocab_size": 101314 }, "tie_word_embeddings": true, "transformers_version": "5.16.0.dev0", "video_token_id": 100296, "vision_config": { "attention_dropout": 0.0, "hidden_act": "gelu_pytorch_tanh", "hidden_size": 144, "image_size": 384, "intermediate_size": 32, "layer_norm_eps": 1e-06, "model_type": "paddleocr_vl_vision", "num_attention_heads": 4, "num_channels": 3, "num_hidden_layers": 2, "pad_token_id": 0, "patch_size": 14, "spatial_merge_size": 2 }, "vision_end_token_id": 101306, "vision_start_token_id": 101305 }