SVTRv2 recognition accuracy drops to 0 after certain epochs

[amandhapola]

Hi,
I have been trying to finetune SVTRv2 link on custom dataset . the model link is link . The training starts fine and the accuracy increases till 89% for starting epochs and then suddenly drops to 0 and remains there . My config
file is as follows :
`Global:
debug: false
use_gpu: true
epoch_num: 100
log_smooth_window: 20
print_batch_step: 10
save_model_dir: /app/PaddleOCR/train_exps/exp4_add_5_datasets_currency_supersubscripts2_svtrv2
save_epoch_step: 1
eval_batch_step:

• 0
• 4000
  cal_metric_during_train: true
  pretrained_model: /app/PaddleOCR/best_accuracy_rec_svtrv2_ch_train/best_accuracy
  checkpoints: null
  save_inference_dir: null
  use_visualdl: false
  infer_img: /app/PaddleOCR/train_data/original_format/test_data/dataset11_15_86661426_89878311_2.jpg
  character_dict_path: /app/PaddleOCR/train_dict/en_dict.txt
  max_text_length: 25
  infer_mode: false
  use_space_char: true
  distributed: true
  save_res_path: /app/PaddleOCR/train_exps/exp4_add_5_datasets_currency_supersubscripts2_svtrv2
  d2s_train_image_shape:
• 3
• 48
• 320
  Optimizer:
  name: AdamW
  beta1: 0.9
  beta2: 0.999
  epsilon: 1.0e-08
  weight_decay: 0.05
  no_weight_decay_name: norm
  one_dim_param_no_weight_decay: true
  lr:
  name: Cosine
  learning_rate: 0.001
  warmup_epoch: 5
  Architecture:
  model_type: rec
  algorithm: SVTR_HGNet
  Transform: null
  Backbone:
  name: SVTRv2
  use_pos_embed: false
  dims:
  • 128
  • 256
  • 384
    depths:
  • 6
  • 6
  • 6
    num_heads:
  • 4
  • 8
  • 12
    mixer:
  • • Conv
    • Conv
    • Conv
    • Conv
    • Conv
    • Conv
  • • Conv
    • Conv
    • Global
    • Global
    • Global
    • Global
  • • Global
    • Global
    • Global
    • Global
    • Global
    • Global
      local_k:
  • • 5
    • 5
  • • 5
    • 5
  • • -1
    • -1
      sub_k:
  • • 2
    • 1
  • • 2
    • 1
  • • -1
    • -1
      last_stage: false
      use_pool: true
      Head:
      name: MultiHead
      head_list:
  • CTCHead:
    Neck:
    name: svtr
    dims: 256
    depth: 2
    hidden_dims: 256
    kernel_size:
    - 1
    - 3
    use_guide: true
    Head:
    fc_decay: 1.0e-05
  • NRTRHead:
    nrtr_dim: 384
    max_text_length: 25
    num_decoder_layers: 2
    Loss:
    name: MultiLoss
    loss_config_list:
• CTCLoss: null
• NRTRLoss: null
  PostProcess:
  name: CTCLabelDecode
  Metric:
  name: RecMetric
  main_indicator: acc
  Train:
  dataset:
  name: MultiScaleDataSet
  ds_width: false
  data_dir: /app/PaddleOCR/train_data/original_format
  label_file_list: /app/PaddleOCR/train_data/original_format/new_train_labels_1.txt
  ext_op_transform_idx: 1
  transforms:
  • DecodeImage:
    img_mode: BGR
    channel_first: false
  • 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: 8
      fix_bs: false
      divided_factor:
  • 8
  • 16
    is_training: true
    loader:
    shuffle: true
    batch_size_per_card: 1536
    drop_last: true
    num_workers: 4
    use_shared_memory: true
    Eval:
    dataset:
    name: SimpleDataSet
    data_dir: /app/PaddleOCR/train_data/original_format
    label_file_list: /app/PaddleOCR/train_data/original_format/new_test_labels_1.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: 512
      num_workers: 6
      profiler_options: null
      `
      I am training on NVIDIA RTX 4000 20gb single GPU . RAM : 64 GB . OS : CentOS Stream 9