Global:
  model_name: SLANeXt_wired # To use static model for inference.
  use_gpu: true
  epoch_num: 400
  log_smooth_window: 20
  print_batch_step: 20
  save_model_dir: ./output/SLANeXt_wired
  save_epoch_step: 400
  eval_batch_step:
  - 0
  - 331
  cal_metric_during_train: true
  pretrained_model: "https://paddle-model-ecology.bj.bcebos.com/paddlex/official_pretrained_model/SLANeXt_wired_pretrained.pdparams"
  checkpoints: null
  save_inference_dir: ./output/SLANeXt_wired/infer
  use_visualdl: false
  infer_img: ppstructure/docs/table/table.jpg
  character_dict_path: ppocr/utils/dict/table_structure_dict_ch.txt
  character_type: en
  max_text_length: 500
  box_format: xyxyxyxy
  infer_mode: false
  use_sync_bn: true
  save_res_path: output/infer
  d2s_train_image_shape: [3, 512, 512]

Optimizer:
  name: AdamW
  beta1: 0.9
  beta2: 0.999
  clip_norm: 5.0
  lr:
    name: Cosine
    learning_rate: 0.0001
    warmup_epoch: 1
  regularizer:
    name: L2
    factor: 0.0

Architecture:
  model_type: table
  algorithm: SLANeXt
  Backbone:
    name: Vary_VIT_B
    image_size: 512 
    encoder_embed_dim: 768
    encoder_depth: 12
    encoder_num_heads: 12
    encoder_global_attn_indexes: [2, 5, 8, 11]
  Head:
    name: SLAHead
    hidden_size: 512
    max_text_length: 500
    loc_reg_num: 8

Loss:
  name: SLALoss
  structure_weight: 1.0
  # SLANeXt does not train the cell location task by default, set the loc_weight if needed.
  loc_weight: 0.0
  loc_loss: smooth_l1

PostProcess:
  name: TableLabelDecode
  merge_no_span_structure: true

Metric:
  name: TableMetric
  main_indicator: acc
  compute_bbox_metric: false
  loc_reg_num: 8
  box_format: xyxyxyxy
  del_thead_tbody: true

Train:
  dataset:
    name: PubTabDataSet
    data_dir: train_data/table/train/
    label_file_list:
    - train_data/table/train.txt
    ratio_list:
    - 1
    transforms:
    - DecodeImage:
        img_mode: BGR
        channel_first: false
    - TableLabelEncode:
        learn_empty_box: false
        merge_no_span_structure: true
        replace_empty_cell_token: false
        loc_reg_num: 8
        max_text_length: 500
    - TableBoxEncode:
        in_box_format: xyxyxyxy
        out_box_format: xyxyxyxy
    - ResizeTableImage:
        max_len: 512
        resize_bboxes: true
    - NormalizeImage:
        scale: 1./255.
        mean:
        - 0.485
        - 0.456
        - 0.406
        std:
        - 0.229
        - 0.224
        - 0.225
        order: hwc
    - PaddingTableImage:
        size:
        - 512
        - 512
    - ToCHWImage: null
    - KeepKeys:
        keep_keys:
        - image
        - structure
        - bboxes
        - bbox_masks
        - length
        - shape
  loader:
    shuffle: true
    batch_size_per_card: 48
    drop_last: true
    num_workers: 1

Eval:
  dataset:
    name: PubTabDataSet
    data_dir: train_data/table/val/
    label_file_list:
    - train_data/table/val.txt
    transforms:
    - DecodeImage:
        img_mode: BGR
        channel_first: false
    - TableLabelEncode:
        learn_empty_box: false
        merge_no_span_structure: true
        replace_empty_cell_token: false
        loc_reg_num: 8
        max_text_length: 500
    - TableBoxEncode:
        in_box_format: xyxyxyxy
        out_box_format: xyxyxyxy
    - ResizeTableImage:
        max_len: 512
        resize_bboxes: true
    - NormalizeImage:
        scale: 1./255.
        mean:
        - 0.485
        - 0.456
        - 0.406
        std:
        - 0.229
        - 0.224
        - 0.225
        order: hwc
    - PaddingTableImage:
        size:
        - 512
        - 512
    - ToCHWImage: null
    - KeepKeys:
        keep_keys:
        - image
        - structure
        - bboxes
        - bbox_masks
        - length
        - shape
  loader:
    shuffle: false
    drop_last: false
    batch_size_per_card: 48
    num_workers: 1

profiler_options: null