PaddleOcr安装和训练备忘录

目录

如果你要安装PaddleHub，不要使用python 3.12

设置系统变量：HUB_HOME，修改PaddleHub模型目录

Could not locate zlibwapi.dll问题

模型训练主要参数说明：

加入自动混合精度训练提高训练速度

---

如果你要安装PaddleHub，不要使用python 3.12

用到了sentencepiece 0.1.99版本，3.12版本py无法编译成功
可以使用 pip install sentencepiece==0.1.99 测试是否可以安装

设置系统变量：HUB_HOME，修改PaddleHub模型目录

Could not locate zlibwapi.dll问题

缺少zlibwapi.dll，网上都说去nvidia下载，里面内容已经更新了，找不到下载的地方了

模型训练主要参数说明：

epoch_num: 总的训练轮数

print_batch_step: 每隔多少个batch打印一回信息

save_model_dir: 模型保存路径

save_epoch_step: 模型保存时间间隔，以epoch计

eval_batch_step: 在第0个batch后每隔500个batch评估一次

batch_size_per_card: 每个batch运行的图片数，过多会爆显存

pretrained_model: 预训练模型(F:\paddle_train\output\verify\tran\en_PP-OCRv4_rec_train\best_accuracy)

character_dict_path:字典(F:\paddle_train\output\verify\data\new_dict.txt)

加入自动混合精度训练提高训练速度

  use_amp: True 
  scale_loss: 1024.0
  use_dynamic_loss_scaling: True
  use_gpu: true

Global:
  debug: false
  use_amp: True 
  scale_loss: 1024.0
  use_dynamic_loss_scaling: True
  use_gpu: true
  epoch_num: 50
  log_smooth_window: 20
  print_batch_step: 10
  save_model_dir: F:\paddle_train\output\verify\output\rec_verify_v4
  save_epoch_step: 3
  eval_batch_step:
  - 0
  - 500
  cal_metric_during_train: true
  pretrained_model: F:\paddle_train\output\verify\tran\en_PP-OCRv4_rec_train\best_accuracy
  checkpoints: null
  save_inference_dir: F:\paddle_train\output\verify\output\inference
  use_visualdl: false
  infer_img: doc\imgs_words\ch\word_1.jpg
  character_dict_path: F:\paddle_train\output\verify\data\new_dict.txt
  max_text_length: 6
  infer_mode: false
  use_space_char: false
  distributed: true
  save_res_path: F:\paddle_train\output\verify\output\rec\predicts_ppocrv3.txt
Optimizer:
  name: Adam
  beta1: 0.9
  beta2: 0.999
  lr:
    name: Cosine
    learning_rate: 0.0005
    warmup_epoch: 5
  regularizer:
    name: L2
    factor: 3.0e-05
Architecture:
  model_type: rec
  algorithm: SVTR_LCNet
  Transform: null
  Backbone:
    name: PPLCNetV3
    scale: 0.95
  Head:
    name: MultiHead
    head_list:
    - CTCHead:
        Neck:
          name: svtr
          dims: 120
          depth: 2
          hidden_dims: 120
          kernel_size:
          - 1
          - 3
          use_guide: true
        Head:
          fc_decay: 1.0e-05
    - NRTRHead:
        nrtr_dim: 384
        max_text_length: 6
Loss:
  name: MultiLoss
  loss_config_list:
  - CTCLoss: null
  - NRTRLoss: null
PostProcess:
  name: CTCLabelDecode
Metric:
  name: RecMetric
  main_indicator: acc
  ignore_space: false
Train:
  dataset:
    name: MultiScaleDataSet
    ds_width: false
    data_dir: F:\paddle_train\output\verify\data\verification_code
    ext_op_transform_idx: 1
    label_file_list:
    - F:\paddle_train\output\verify\data\train_list.txt
    transforms:
    - DecodeImage:
        img_mode: BGR
        channel_first: false
    - RecConAug:
        prob: 0.5
        ext_data_num: 2
        image_shape:
        - 48
        - 320
        - 3
        max_text_length: 6
    - RecAug: null
    - MultiLabelEncode:
        gtc_encode: NRTRLabelEncode
    - KeepKeys:
        keep_keys:
        - image
        - label_ctc
        - label_gtc
        - length
        - valid_ratio
  sampler:
    name: MultiScaleSampler
    scales:
    - - 320
      - 32
    - - 320
      - 48
    - - 320
      - 64
    first_bs: 96
    fix_bs: false
    divided_factor:
    - 8
    - 16
    is_training: true
  loader:
    shuffle: true
    batch_size_per_card: 65
    drop_last: true
    num_workers: 4
Eval:
  dataset:
    name: SimpleDataSet
    data_dir: F:\paddle_train\output\verify\data\verification_code
    label_file_list:
    - F:\paddle_train\output\verify\data\test_list.txt
    transforms:
    - DecodeImage:
        img_mode: BGR
        channel_first: false
    - MultiLabelEncode:
        gtc_encode: NRTRLabelEncode
    - RecResizeImg:
        image_shape:
        - 3
        - 48
        - 320
    - KeepKeys:
        keep_keys:
        - image
        - label_ctc
        - label_gtc
        - length
        - valid_ratio
  loader:
    shuffle: false
    drop_last: false
    batch_size_per_card: 32
    num_workers: 4
profiler_options: null