PP-OCRv5_server_det模型训练速度过于慢

[Maxx2208]

🐛 Bug (问题描述)

由于Paddle目前不支持50系的显卡只能先用CPU训练，目前准备了两台电脑进行训练
CPU：AMD Ryzen 7 7840HS w，batch_size_per_card：1，每一步花费的时间是15s左右，如果调高batch_size花费的平均时间会更长；
CPU： i5-14600KF，数据集和配置文件同上，Paddle、PaddleOCR版本也保持一致，每一步花费的时间是27s左右；
因为训练速度过慢，暂时使用100张左右的图片进行训练，可能是数据量过少导致训练结束时hmean为0，但是之前使用PaddleOCR-release-2.6版本训练v2模型时用100张图片得到了想要的结果，更新版本后反而训练速度变慢相同的数据也得不到同样的结果了。

🏃‍♂️ Environment (运行环境)

python 3.9
paddlepaddle 3.1.0
PaddleOCR-release-3.1

配置文件如下：

Global:
  model_name: PP-OCRv5_server_det # To use static model for inference.
  debug: false
  use_gpu: false
  epoch_num: &epoch_num 100
  log_smooth_window: 20
  print_batch_step: 1
  save_model_dir: ./output/PP-OCRv5_server_det
  save_epoch_step: 100
  eval_batch_step:
  - 0
  - 1500
  cal_metric_during_train: false
  checkpoints:
  pretrained_model: ./pretrain_models/PP-OCRv5_server_det_pretrained
  save_inference_dir: null
  use_visualdl: false
  infer_img: doc/imgs_en/img_10.jpg
  save_res_path: ./checkpoints/det_db/predicts_db.txt
  distributed: true

Architecture:
  model_type: det
  algorithm: DB
  Transform: null
  Backbone:
    name: PPHGNetV2_B4
    det: True
  Neck:
    name: LKPAN
    out_channels: 256
    intracl: true
  Head:
    name: PFHeadLocal
    k: 50
    mode: "large"
    

Loss:
  name: DBLoss
  balance_loss: true
  main_loss_type: DiceLoss
  alpha: 5
  beta: 10
  ohem_ratio: 3

Optimizer:
  name: Adam
  beta1: 0.9
  beta2: 0.999
  lr:
    name: Cosine
    learning_rate: 0.001 #(8*8c)
    warmup_epoch: 2
  regularizer:
    name: L2
    factor: 1e-6

PostProcess:
  name: DBPostProcess
  thresh: 0.3
  box_thresh: 0.6
  max_candidates: 1000
  unclip_ratio: 1.5

Metric:
  name: DetMetric
  main_indicator: hmean

Train:
  dataset:
    name: SimpleDataSet
    data_dir: ./train_data/
    label_file_list:
      - ./train_data/det/train.txt
    ratio_list: [1.0]
    transforms:
    - DecodeImage:
        img_mode: BGR
        channel_first: false
    - DetLabelEncode: null
    - CopyPaste: null
    - IaaAugment:
        augmenter_args:
        - type: Fliplr
          args:
            p: 0.5
        - type: Affine
          args:
            rotate:
            - -10
            - 10
        - type: Resize
          args:
            size:
            - 0.5
            - 3
    - EastRandomCropData:
        size:
        - 640
        - 640
        max_tries: 50
        keep_ratio: true
    - MakeBorderMap:
        shrink_ratio: 0.4
        thresh_min: 0.3
        thresh_max: 0.7
        total_epoch: *epoch_num
    - MakeShrinkMap:
        shrink_ratio: 0.4
        min_text_size: 8
        total_epoch: *epoch_num
    - NormalizeImage:
        scale: 1./255.
        mean:
        - 0.485
        - 0.456
        - 0.406
        std:
        - 0.229
        - 0.224
        - 0.225
        order: hwc
    - ToCHWImage: null
    - KeepKeys:
        keep_keys:
        - image
        - threshold_map
        - threshold_mask
        - shrink_map
        - shrink_mask
  loader:
    shuffle: true
    drop_last: false
    batch_size_per_card: 1
    num_workers: 1

Eval:
  dataset:
    name: SimpleDataSet
    data_dir: ./train_data/
    label_file_list:
      - ./train_data/det/val.txt
    transforms:
    - DecodeImage:
        img_mode: BGR
        channel_first: false
    - DetLabelEncode: null
    - DetResizeForTest:
    - NormalizeImage:
        scale: 1./255.
        mean:
        - 0.485
        - 0.456
        - 0.406
        std:
        - 0.229
        - 0.224
        - 0.225
        order: hwc
    - ToCHWImage: null
    - KeepKeys:
        keep_keys:
        - image
        - shape
        - polys
        - ignore_tags
  loader:
    shuffle: false
    drop_last: false
    batch_size_per_card: 1
    num_workers: 2
profiler_options: null

上图中训练刚开始学习率(lr)一直在上升，在第40步左右开始一直下降，到训练结束时学习率已经几乎为0，我刚接触AI训练，找不到原因，希望可以指点我一下，可以提供训练用的数据集

[liuhongen1234567]

您好，PP-OCRv5_server_det 的字典远大于v2模型 ，确实CPU训练会比较慢，在有限步数下也不容易收敛，可以试下更轻量的 PP-OCRv5_mobile_det 模型。