# Copyright (c) 2026 PaddlePaddle Authors. All Rights Reserved. # # Licensed under the Apache License, Version 2.0 (the "License"); # you may not use this file except in compliance with the License. # You may obtain a copy of the License at # # http://www.apache.org/licenses/LICENSE-2.0 # # Unless required by applicable law or agreed to in writing, software # distributed under the License is distributed on an "AS IS" BASIS, # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. # See the License for the specific language governing permissions and # limitations under the License. """Model config.json definitions for pdparams2safetensors conversion. Each entry maps a model_name to the full config dict that will be saved as config.json, matching the official HuggingFace repos exactly. """ # PP-LCNet configs _PPLCNET_BASE = { "model_type": "pp_lcnet", "scale": 1.0, "reduction": 4, "hidden_dropout_prob": 0.2, "class_expand": 1280, "hidden_act": "hardswish", } _TEXTLINE_BLOCK_CONFIGS = [ [[3, 16, 32, 1, False]], [[3, 32, 64, [2, 1], False], [3, 64, 64, 1, False]], [[3, 64, 128, [2, 1], False], [3, 128, 128, 1, False]], [ [3, 128, 256, [2, 1], False], [5, 256, 256, 1, False], [5, 256, 256, 1, False], [5, 256, 256, 1, False], [5, 256, 256, 1, False], [5, 256, 256, 1, False], ], [[5, 256, 512, [2, 1], True], [5, 512, 512, 1, True]], ] # PP-OCRv5 configs _PPOCRV5_MOBILE_DET_CONFIG = { "model_type": "pp_ocrv5_mobile_det", "reduction": 4, "layer_list_out_channels": [12, 18, 42, 360], "backbone_config": { "model_type": "pp_lcnet_v3", "scale": 0.75, "out_features": ["stage2", "stage3", "stage4", "stage5"], "out_indices": [2, 3, 4, 5], "divisor": 16, }, "neck_out_channels": 96, "kernel_list": [3, 2, 2], "interpolate_mode": "nearest", } _PPOCRV5_SERVER_DET_CONFIG = { "model_type": "pp_ocrv5_server_det", "mode": "large", "upsample_mode": "nearest", "upsample_align_mode": 1, "backbone_config": { "model_type": "hgnet_v2", "arch": "L", "return_idx": [0, 1, 2, 3], "freeze_stem_only": True, "freeze_at": 0, "freeze_norm": True, "lr_mult_list": [0, 0.05, 0.05, 0.05, 0.05], "out_features": ["stage1", "stage2", "stage3", "stage4"], }, "use_lab": False, "use_last_conv": True, "class_expand": 2048, "class_num": 1000, "out_indices": [0, 1, 2, 3], "neck_out_channels": 256, "reduce_factor": 2, "intraclass_block_number": 4, "intraclass_block_config": { "reduce_channel": [1, 1, 0], "return_channel": [1, 1, 0], "vertical_long_to_small_conv_longratio": [[7, 1], [1, 1], [3, 0]], "vertical_long_to_small_conv_midratio": [[5, 1], [1, 1], [2, 0]], "vertical_long_to_small_conv_shortratio": [[3, 1], [1, 1], [1, 0]], "horizontal_small_to_long_conv_longratio": [[1, 7], [1, 1], [0, 3]], "horizontal_small_to_long_conv_midratio": [[1, 5], [1, 1], [0, 2]], "horizontal_small_to_long_conv_shortratio": [[1, 3], [1, 1], [0, 1]], "symmetric_conv_long_longratio": [[7, 7], [1, 1], [3, 3]], "symmetric_conv_long_midratio": [[5, 5], [1, 1], [2, 2]], "symmetric_conv_long_shortratio": [[3, 3], [1, 1], [1, 1]], }, "head_in_channels": 1024, "scale_factor": 2, "scale_factor_list": [1, 2, 4, 8], "hidden_act": "relu", "kernel_list": [3, 2, 2], } _PPOCRV5_REC_BASE = { "hidden_act": "silu", "hidden_size": 120, "mlp_ratio": 2.0, "depth": 2, "head_out_channels": 18385, "conv_kernel_size": [1, 3], "qkv_bias": True, "num_attention_heads": 8, "attention_dropout": 0.0, } _PPOCRV5_MOBILE_REC_CONFIG = { "model_type": "pp_ocrv5_mobile_rec", "backbone_config": { "model_type": "pp_lcnet_v3", "scale": 0.95, "out_features": ["stage2", "stage3", "stage4", "stage5"], "out_indices": [2, 3, 4, 5], "divisor": 16, "block_configs": [ [[3, 16, 32, 1, False]], [[3, 32, 64, 1, False], [3, 64, 64, 1, False]], [[3, 64, 128, [2, 1], False], [3, 128, 128, 1, False]], [ [3, 128, 256, [1, 2], False], [5, 256, 256, 1, False], [5, 256, 256, 1, False], [5, 256, 256, 1, False], [5, 256, 256, 1, False], ], [ [5, 256, 512, [2, 1], True], [5, 512, 512, 1, True], [5, 512, 512, [2, 1], False], [5, 512, 512, 1, False], ], ], }, **_PPOCRV5_REC_BASE, } _PPOCRV5_SERVER_REC_CONFIG = { "model_type": "pp_ocrv5_server_rec", "backbone_config": { "model_type": "hgnet_v2", "arch": "L", "return_idx": [0, 1, 2, 3], "freeze_stem_only": True, "freeze_at": 0, "freeze_norm": True, "lr_mult_list": [1.0, 1.0, 1.0, 1.0, 1.0], "out_features": ["stage1", "stage2", "stage3", "stage4"], "stage_downsample": [True, True, True, True], "stem_strides": [2, 1, 1, 1, 1], "stage_downsample_strides": [[2, 1], [1, 2], [2, 1], [2, 1]], }, **_PPOCRV5_REC_BASE, } # RT-DETR / DocLayout configs _RTDETR_BACKBONE_CONFIG = { "arch": "L", "depths": [3, 4, 6, 3], "embedding_size": 64, "hidden_act": "relu", "hidden_sizes": [256, 512, 1024, 2048], "initializer_range": 0.02, "model_type": "hgnet_v2", "num_channels": 3, "out_features": ["stage2", "stage3", "stage4"], "out_indices": [2, 3, 4], "stage_downsample": [False, True, True, True], "stage_downsample_strides": [2, 2, 2, 2], "stage_in_channels": [48, 128, 512, 1024], "stage_kernel_size": [3, 3, 5, 5], "stage_light_block": [False, False, True, True], "stage_mid_channels": [48, 96, 192, 384], "stage_names": ["stem", "stage1", "stage2", "stage3", "stage4"], "stage_num_blocks": [1, 1, 3, 1], "stage_numb_of_layers": [6, 6, 6, 6], "stage_out_channels": [128, 512, 1024, 2048], "stem_channels": [3, 32, 48], "stem_strides": [2, 1, 1, 2, 1], "use_learnable_affine_block": False, "return_idx": [1, 2, 3], "freeze_stem_only": True, "freeze_at": 0, "freeze_norm": False, "lr_mult_list": [0.05, 0.05, 0.1, 0.15, 0.2], } def _rtdetr_config(id2label, label2id): """Build a standard RT-DETR config with given labels.""" return { "activation_dropout": 0.0, "activation_function": "silu", "anchor_image_size": None, "attention_dropout": 0.0, "auxiliary_loss": True, "backbone_config": _RTDETR_BACKBONE_CONFIG, "batch_norm_eps": 1e-05, "box_noise_scale": 1.0, "d_model": 256, "decoder_activation_function": "relu", "decoder_attention_heads": 8, "decoder_ffn_dim": 1024, "decoder_in_channels": [256, 256, 256], "decoder_layers": 6, "decoder_n_points": 4, "disable_custom_kernels": True, "dropout": 0.0, "encode_proj_layers": [2], "encoder_activation_function": "gelu", "encoder_attention_heads": 8, "encoder_ffn_dim": 1024, "encoder_hidden_dim": 256, "encoder_in_channels": [512, 1024, 2048], "encoder_layers": 1, "eos_coefficient": 0.0001, "eval_size": None, "feat_strides": [8, 16, 32], "focal_loss_alpha": 0.75, "focal_loss_gamma": 2.0, "freeze_backbone_batch_norms": True, "hidden_expansion": 1.0, "id2label": id2label, "initializer_bias_prior_prob": None, "initializer_range": 0.01, "is_encoder_decoder": True, "label2id": label2id, "label_noise_ratio": 0.5, "layer_norm_eps": 1e-05, "learn_initial_query": False, "matcher_alpha": 0.25, "matcher_bbox_cost": 5.0, "matcher_class_cost": 2.0, "matcher_gamma": 2.0, "matcher_giou_cost": 2.0, "model_type": "rt_detr", "normalize_before": False, "num_denoising": 100, "num_feature_levels": 3, "num_queries": 300, "positional_encoding_temperature": 10000, "transformers_version": "5.3.0.dev0", "use_focal_loss": True, "weight_loss_bbox": 5.0, "weight_loss_giou": 2.0, "weight_loss_vfl": 1.0, "with_box_refine": True, } _DOCLAYOUTV2_CONFIG = { "activation_dropout": 0.0, "activation_function": "silu", "anchor_image_size": None, "architectures": ["PPDocLayoutV2ForObjectDetection"], "attention_dropout": 0.0, "backbone": None, "backbone_config": { "model_type": "hgnet_v2", "arch": "L", "return_idx": [1, 2, 3], "freeze_stem_only": True, "freeze_at": 0, "freeze_norm": True, "lr_mult_list": [0, 0.05, 0.05, 0.05, 0.05], "out_features": ["stage2", "stage3", "stage4"], }, "backbone_kwargs": None, "batch_norm_eps": 1e-05, "box_noise_scale": 1.0, "d_model": 256, "decoder_activation_function": "relu", "decoder_attention_heads": 8, "decoder_ffn_dim": 1024, "decoder_in_channels": [256, 256, 256], "decoder_layers": 6, "decoder_n_points": 4, "disable_custom_kernels": True, "dropout": 0.0, "encode_proj_layers": [2], "encoder_activation_function": "gelu", "encoder_attention_heads": 8, "encoder_ffn_dim": 1024, "encoder_hidden_dim": 256, "encoder_in_channels": [512, 1024, 2048], "encoder_layers": 1, "eos_coefficient": 0.0001, "eval_size": None, "feat_strides": [8, 16, 32], "hidden_expansion": 1.0, "id2label": { "0": "abstract", "1": "algorithm", "2": "aside_text", "3": "chart", "4": "content", "5": "formula", "6": "doc_title", "7": "figure_title", "8": "footer", "9": "footer", "10": "footnote", "11": "formula_number", "12": "header", "13": "header", "14": "image", "15": "formula", "16": "number", "17": "paragraph_title", "18": "reference", "19": "reference_content", "20": "seal", "21": "table", "22": "text", "23": "text", "24": "vision_footnote", }, "initializer_range": 0.01, "is_encoder_decoder": True, "label2id": {}, "label_noise_ratio": 0.5, "layer_norm_eps": 1e-05, "learn_initial_query": False, "matcher_alpha": 0.25, "matcher_bbox_cost": 5.0, "matcher_class_cost": 2.0, "matcher_gamma": 2.0, "matcher_giou_cost": 2.0, "model_type": "pp_doclayout_v2", "normalize_before": False, "num_denoising": 100, "num_feature_levels": 3, "num_queries": 300, "positional_encoding_temperature": 10000, "torch_dtype": "float32", "use_pretrained_backbone": False, "use_timm_backbone": False, "reading_order_config": { "hidden_size": 512, "num_attention_heads": 8, "attention_probs_dropout_prob": 0.1, "has_relative_attention_bias": False, "has_spatial_attention_bias": True, "layer_norm_eps": 1e-05, "hidden_dropout_prob": 0.1, "intermediate_size": 2048, "hidden_act": "gelu", "num_hidden_layers": 6, "rel_pos_bins": 32, "max_rel_pos": 128, "rel_2d_pos_bins": 64, "max_rel_2d_pos": 256, "max_position_embeddings": 514, "max_2d_position_embeddings": 1024, "type_vocab_size": 1, "vocab_size": 4, "start_token_id": 0, "pad_token_id": 1, "end_token_id": 2, "pred_token_id": 3, "coordinate_size": 171, "shape_size": 170, "num_classes": 20, "relation_bias_embed_dim": 16, "relation_bias_theta": 10000, "relation_bias_scale": 100, "global_pointer_head_size": 64, "tril_mask": True, }, "class_thresholds": [ 0.5, 0.5, 0.5, 0.5, 0.5, 0.4, 0.4, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.4, 0.5, 0.4, 0.5, 0.5, 0.45, 0.5, 0.4, 0.4, 0.5, ], "class_order": [ 4, 2, 14, 1, 5, 7, 8, 6, 11, 11, 9, 13, 10, 10, 1, 2, 3, 0, 2, 2, 12, 1, 2, 15, 6, ], } _DOCLAYOUTV3_CONFIG = { "activation_dropout": 0.0, "activation_function": "silu", "anchor_image_size": None, "architectures": ["PPDocLayoutV3ForObjectDetection"], "attention_dropout": 0.0, "backbone": None, "backbone_config": { "model_type": "hgnet_v2", "arch": "L", "return_idx": [0, 1, 2, 3], "freeze_stem_only": True, "freeze_at": 0, "freeze_norm": True, "lr_mult_list": [0, 0.05, 0.05, 0.05, 0.05], "out_features": ["stage1", "stage2", "stage3", "stage4"], }, "backbone_kwargs": None, "batch_norm_eps": 1e-05, "box_noise_scale": 1.0, "d_model": 256, "decoder_activation_function": "relu", "decoder_attention_heads": 8, "decoder_ffn_dim": 1024, "decoder_in_channels": [256, 256, 256], "decoder_layers": 6, "decoder_n_points": 4, "disable_custom_kernels": True, "dropout": 0.0, "encode_proj_layers": [2], "encoder_activation_function": "gelu", "encoder_attention_heads": 8, "encoder_ffn_dim": 1024, "encoder_hidden_dim": 256, "encoder_in_channels": [512, 1024, 2048], "encoder_layers": 1, "eos_coefficient": 0.0001, "eval_size": None, "feature_strides": [8, 16, 32], "hidden_expansion": 1.0, "id2label": { "0": "abstract", "1": "algorithm", "2": "aside_text", "3": "chart", "4": "content", "5": "formula", "6": "doc_title", "7": "figure_title", "8": "footer", "9": "footer", "10": "footnote", "11": "formula_number", "12": "header", "13": "header", "14": "image", "15": "formula", "16": "number", "17": "paragraph_title", "18": "reference", "19": "reference_content", "20": "seal", "21": "table", "22": "text", "23": "text", "24": "vision_footnote", }, "initializer_range": 0.01, "is_encoder_decoder": True, "label2id": {}, "label_noise_ratio": 0.5, "layer_norm_eps": 1e-05, "learn_initial_query": False, "matcher_alpha": 0.25, "matcher_bbox_cost": 5.0, "matcher_class_cost": 2.0, "matcher_gamma": 2.0, "matcher_giou_cost": 2.0, "model_type": "pp_doclayout_v3", "normalize_before": False, "num_denoising": 100, "num_feature_levels": 3, "num_queries": 300, "positional_encoding_temperature": 10000, "torch_dtype": "float32", "use_pretrained_backbone": False, "use_timm_backbone": False, "global_pointer_head_size": 64, "mask_feature_channels": [64, 64], "x4_feat_dim": 128, } # UVDoc config _UVDOC_CONFIG = { "model_type": "uvdoc", "kernel_size": 5, "backbone_config": { "model_type": "uvdoc_backbone", "resnet_head": [[3, 32], [32, 32]], "resnet_configs": [ [[32, 32, 1, False], [32, 32, 3, False], [32, 32, 3, False]], [ [32, 64, 1, True], [64, 64, 3, False], [64, 64, 3, False], [64, 64, 3, False], ], [ [64, 128, 1, True], [128, 128, 3, False], [128, 128, 3, False], [128, 128, 3, False], [128, 128, 3, False], [128, 128, 3, False], ], ], "stage_configs": [ [[128, 1]], [[128, 2]], [[128, 5]], [[128, 8], [128, 3], [128, 2]], [[128, 12], [128, 7], [128, 4]], [[128, 18], [128, 12], [128, 6]], ], "out_features": [ "stage1", "stage2", "stage3", "stage4", "stage5", "stage6", ], "out_indices": [1, 2, 3, 4, 5, 6], }, "bridge_connector": [128, 128], "out_point_positions2D": [[128, 32], [32, 2]], "dilation_values": [[1], [2], [5], [8, 3, 2], [12, 7, 4], [18, 12, 6]], "padding_mode": "reflect", "hidden_act": "prelu", } # Label dicts for RT-DETR models _TABLE_CELL_LABELS = {"0": "cell"} _TABLE_CELL_LABEL2ID = {"cell": 0} _DOC_LAYOUT_PLUS_ID2LABEL = { "0": "paragraph_title", "1": "image", "2": "text", "3": "number", "4": "abstract", "5": "content", "6": "figure_title", "7": "formula", "8": "table", "9": "reference", "10": "doc_title", "11": "footnote", "12": "header", "13": "algorithm", "14": "footer", "15": "seal", "16": "chart", "17": "formula_number", "18": "aside_text", "19": "reference_content", } _DOC_LAYOUT_PLUS_LABEL2ID = {v: int(k) for k, v in _DOC_LAYOUT_PLUS_ID2LABEL.items()} _DOC_BLOCK_LABELS = {"0": "Region"} _DOC_BLOCK_LABEL2ID = {"Region": 0} # PP-FormulaNet configs (L and _plus-L share the vision encoder + decoder # architecture; only max_position_embeddings differs). Field order matches # the published config.json on PaddlePaddle/PP-FormulaNet-L_safetensors and # PaddlePaddle/PP-FormulaNet_plus-L_safetensors. _PP_FORMULANET_VISION = { "image_size": 768, "output_channels": 256, "num_channels": 3, "patch_size": 16, "hidden_act": "gelu", "layer_norm_eps": 1e-6, "attention_dropout": 0.0, "qkv_bias": True, "use_abs_pos": True, "use_rel_pos": True, "window_size": 14, "hidden_size": 768, "num_hidden_layers": 12, "num_attention_heads": 12, "global_attn_indexes": [2, 5, 8, 11], "mlp_dim": 3072, "post_conv_in_channels": 256, "post_conv_mid_channels": 512, "post_conv_out_channels": 1024, "decoder_hidden_size": 512, } def _pp_formulanet_text(max_position_embeddings): return { "activation_dropout": 0.0, "activation_function": "gelu", "attention_dropout": 0.0, "bos_token_id": 0, "d_model": 512, "decoder_attention_heads": 16, "decoder_ffn_dim": 2048, "decoder_layerdrop": 0.0, "decoder_layers": 8, "dropout": 0.1, "encoder_attention_heads": 16, "encoder_layers": 12, "eos_token_id": 2, "forced_eos_token_id": 2, "init_std": 0.02, "max_position_embeddings": max_position_embeddings, "num_hidden_layers": 12, "pad_token_id": 1, "scale_embedding": True, "tie_word_embeddings": False, "vocab_size": 50000, } _PP_FORMULANET_CONFIG_L = { "model_type": "pp_formulanet", "text_config": _pp_formulanet_text(1024), "vision_config": dict(_PP_FORMULANET_VISION), } # Published _plus-L/config.json omits vision_config.decoder_hidden_size and # relies on the HF default (512). Match that exactly. _pp_formulanet_plus_vision = dict(_PP_FORMULANET_VISION) _pp_formulanet_plus_vision.pop("decoder_hidden_size") _PP_FORMULANET_CONFIG_PLUS_L = { "model_type": "pp_formulanet", "text_config": _pp_formulanet_text(2560), "vision_config": _pp_formulanet_plus_vision, } # Model config registry MODEL_CONFIGS = { "PP-LCNet_x1_0_doc_ori": { **_PPLCNET_BASE, "id2label": {"0": "0", "1": "90", "2": "180", "3": "270"}, }, "PP-LCNet_x1_0_table_cls": { **_PPLCNET_BASE, "id2label": {"0": "wired_table", "1": "wireless_table"}, }, "PP-LCNet_x0_25_textline_ori": { **_PPLCNET_BASE, "scale": 0.25, "block_configs": _TEXTLINE_BLOCK_CONFIGS, "id2label": {"0": "0_degree", "1": "180_degree"}, }, "PP-LCNet_x1_0_textline_ori": { **_PPLCNET_BASE, "block_configs": _TEXTLINE_BLOCK_CONFIGS, "id2label": {"0": "0_degree", "1": "180_degree"}, }, "PP-OCRv5_mobile_det": _PPOCRV5_MOBILE_DET_CONFIG, "PP-OCRv5_server_det": _PPOCRV5_SERVER_DET_CONFIG, "PP-OCRv5_mobile_rec": _PPOCRV5_MOBILE_REC_CONFIG, "PP-OCRv5_server_rec": _PPOCRV5_SERVER_REC_CONFIG, "SLANet": { "model_type": "slanet", "backbone_config": { "model_type": "pp_lcnet", "scale": 1, "out_features": ["stage2", "stage3", "stage4", "stage5"], "out_indices": [2, 3, 4, 5], }, "post_conv_out_channels": 96, "out_channels": 50, "hidden_size": 256, "max_text_length": 500, }, "SLANet_plus": { "model_type": "slanet", "backbone_config": { "model_type": "pp_lcnet", "scale": 1, "out_features": ["stage2", "stage3", "stage4", "stage5"], "out_indices": [2, 3, 4, 5], }, "post_conv_out_channels": 96, "out_channels": 50, "hidden_size": 256, "max_text_length": 500, }, "SLANeXt_wired": { "model_type": "slanext", "vision_config": { "image_size": 512, "output_channels": 256, "num_channels": 3, "patch_size": 16, "hidden_act": "gelu", "layer_norm_eps": 1e-6, "attention_dropout": 0.0, "qkv_bias": True, "use_abs_pos": True, "use_rel_pos": True, "window_size": 14, "hidden_size": 768, "num_hidden_layers": 12, "num_attention_heads": 12, "global_attn_indexes": [2, 5, 8, 11], "mlp_dim": 3072, }, "post_conv_in_channels": 256, "post_conv_out_channels": 512, "out_channels": 50, "hidden_size": 512, "max_text_length": 500, "loc_reg_num": 8, }, "SLANeXt_wireless": { "model_type": "slanext", "vision_config": { "image_size": 512, "output_channels": 256, "num_channels": 3, "patch_size": 16, "hidden_act": "gelu", "layer_norm_eps": 1e-6, "attention_dropout": 0.0, "qkv_bias": True, "use_abs_pos": True, "use_rel_pos": True, "window_size": 14, "hidden_size": 768, "num_hidden_layers": 12, "num_attention_heads": 12, "global_attn_indexes": [2, 5, 8, 11], "mlp_dim": 3072, }, "post_conv_in_channels": 256, "post_conv_out_channels": 512, "out_channels": 50, "hidden_size": 512, "max_text_length": 500, "loc_reg_num": 8, }, "PP-DocLayoutV2": _DOCLAYOUTV2_CONFIG, "PP-DocLayoutV3": _DOCLAYOUTV3_CONFIG, "RT-DETR-L_wired_table_cell_det": _rtdetr_config( _TABLE_CELL_LABELS, _TABLE_CELL_LABEL2ID, ), "RT-DETR-L_wireless_table_cell_det": _rtdetr_config( _TABLE_CELL_LABELS, _TABLE_CELL_LABEL2ID, ), "PP-DocLayout_plus-L": _rtdetr_config( _DOC_LAYOUT_PLUS_ID2LABEL, _DOC_LAYOUT_PLUS_LABEL2ID, ), "PP-DocBlockLayout": _rtdetr_config( _DOC_BLOCK_LABELS, _DOC_BLOCK_LABEL2ID, ), "UVDoc": _UVDOC_CONFIG, "PP-FormulaNet-L": _PP_FORMULANET_CONFIG_L, "PP-FormulaNet_plus-L": _PP_FORMULANET_CONFIG_PLUS_L, "PP-Chart2Table": { "model_type": "pp_chart2table", "architectures": ["GotOcr2ForConditionalGeneration"], "auto_map": { "AutoConfig": "configuration_pp_chart2table.PPChart2TableConfig", "AutoModel": "modeling_pp_chart2table.PPChart2TableForConditionalGeneration", }, "freeze_vision_tower": False, "image_seq_length": 256, "use_im_start_end": True, "bos_token_id": 151643, "eos_token_id": 151643, "im_start_token": 151857, "im_end_token": 151858, "image_token_index": 151859, "vocab_size": 151860, "initializer_range": 0.02, "max_window_layers": 21, "tie_word_embeddings": True, "dtype": "bfloat16", "use_cache": True, "use_sliding_window": False, "vision_config": { "architectures": ["PPChart2TableVisionModel"], "auto_map": { "AutoConfig": "configuration_pp_chart2table.PPChart2TableConfig", "AutoModel": "modeling_pp_chart2table.PPChart2TableVisionModel", }, "num_hidden_layers": 12, "hidden_size": 768, "output_channels": 256, "hidden_act": "gelu", "num_channels": 3, "image_size": 1024, "mlp_ratio": 4.0, "mlp_dim": 3072, "layer_norm_eps": 1e-05, "num_attention_heads": 12, "patch_size": 16, "qkv_bias": True, "use_rel_pos": True, "use_abs_pos": True, "global_attn_indexes": [2, 5, 8, 11], "window_size": 14, "attention_dropout": 0.0, }, "text_config": { "architectures": ["PPChart2TableTextModel"], "auto_map": { "AutoConfig": "configuration_qwen2.Qwen2Config", "AutoModel": "modeling_qwen2.Qwen2Model", }, "hidden_size": 1024, "hidden_act": "silu", "num_hidden_layers": 24, "num_attention_heads": 16, "num_key_value_heads": 16, "intermediate_size": 2816, "attention_dropout": 0.0, "sliding_window": 32768, "rms_norm_eps": 1e-06, "vocab_size": 151860, "max_position_embeddings": 32768, "rope_parameters": { "rope_theta": 1000000.0, "rope_type": "default", }, }, }, }