Global:
  use_gpu: True
  epoch_num: 10
  log_smooth_window: 20
  print_batch_step: 10
  save_model_dir: ./output/rec/r45_abinet/
  save_epoch_step: 1
  # evaluation is run every 2000 iterations
  eval_batch_step: [0, 2000]
  cal_metric_during_train: True
  pretrained_model:
  checkpoints:
  save_inference_dir:
  use_visualdl: False
  infer_img: doc/imgs_words_en/word_10.png
  # for data or label process
  character_dict_path:
  character_type: en
  max_text_length: 25
  infer_mode: False
  use_space_char: False
  save_res_path: ./output/rec/predicts_abinet.txt

Optimizer:
  name: Adam
  beta1: 0.9
  beta2: 0.99
  clip_norm: 20.0
  lr:
    name: Piecewise
    decay_epochs: [6]
    values: [0.0001, 0.00001] 
  regularizer:
    name: 'L2'
    factor: 0.

Architecture:
  model_type: rec
  algorithm: ABINet
  in_channels: 3
  Transform:
  Backbone:
    name: ResNet45
  Head:
    name: ABINetHead
    use_lang: True
    iter_size: 3
    

Loss:
  name: CELoss
  ignore_index: &ignore_index 100 # Must be greater than the number of character classes

PostProcess:
  name: ABINetLabelDecode

Metric:
  name: RecMetric
  main_indicator: acc

Train:
  dataset:
    name: LMDBDataSet
    data_dir: ./train_data/data_lmdb_release/training/
    transforms:
      - DecodeImage: # load image
          img_mode: RGB
          channel_first: False
      - ABINetRecAug:
      - ABINetLabelEncode: # Class handling label
          ignore_index: *ignore_index
      - ABINetRecResizeImg:
          image_shape: [3, 32, 128]
      - KeepKeys:
          keep_keys: ['image', 'label', 'length'] # dataloader will return list in this order
  loader:
    shuffle: True
    batch_size_per_card: 96
    drop_last: True
    num_workers: 4

Eval:
  dataset:
    name: LMDBDataSet
    data_dir: ./train_data/data_lmdb_release/evaluation/
    transforms:
      - DecodeImage: # load image
          img_mode: RGB
          channel_first: False
      - ABINetLabelEncode: # Class handling label
          ignore_index: *ignore_index
      - ABINetRecResizeImg:
          image_shape: [3, 32, 128]
      - KeepKeys:
          keep_keys: ['image', 'label', 'length'] # dataloader will return list in this order
  loader:
    shuffle: False
    drop_last: False
    batch_size_per_card: 256
    num_workers: 4
    use_shared_memory: False