ocr识别，单一的ctc-head输出，使用MultiScaleSampler，精度一直为0，loss为nanxxx，使用SimpleDataSet精度正常

[zhangxiaopang88]

Architecture: model_type: rec algorithm: SVTR Transform: Backbone: name: PPLCNetV3 scale: 0.95 Neck: name: SequenceEncoder encoder_type: reshape Head: name: CTCHead

fc_decay: 0.00001

Loss: name: CTCLoss

PostProcess:
name: CTCLabelDecode

Metric: name: RecMetric main_indicator: acc

Train: dataset: name: MultiScaleDataSet ds_width: false data_dir: ./ ext_op_transform_idx: 1 label_file_list:

• data/train.txt transforms:
• DecodeImage: img_mode: BGR channel_first: false
• RecConAug: prob: 0.5 ext_data_num: 2 image_shape: [48, 128, 3] max_text_length: *max_text_length
• RecAug:
• MultiLabelEncode: gtc_encode: NRTRLabelEncode
• KeepKeys: keep_keys:
  • image
  • label_ctc
  • label_gtc
  • length
  • valid_ratio sampler: name: MultiScaleSampler scales: [[128, 32], [128, 48], [128, 64]] first_bs: &bs 192 fix_bs: false divided_factor: [8, 16] # w, h is_training: True loader: shuffle: true batch_size_per_card: *bs drop_last: true num_workers: 16 Eval: dataset: name: SimpleDataSet data_dir: ./ label_file_list:
• data/eval.txt transforms:
• DecodeImage: img_mode: BGR channel_first: false
• MultiLabelEncode: gtc_encode: NRTRLabelEncode
• RecResizeImg: image_shape: [3, 48, 128]
• KeepKeys: keep_keys:
  • image
  • label_ctc
  • label_gtc
  • length
  • valid_ratio loader: shuffle: false drop_last: false batch_size_per_card: 128 num_workers: 16

log一直是这样 ppocr INFO: epoch: [1/200], global_step: 1580, lr: 0.000004, acc: 0.000000, norm_edit_dis: 0.000000, loss: nanxxx, avg_reader_cost: 0.00013 s, avg_batch_cost: 0.32655 s, avg_samples: 147.2, ips: 450.77714 samples/s, eta: 6 days, 4:14:47

[IbrarBabar009]

is your problem solved?