使用infer_rec时出现问题 #71

super-tian · 2024-12-27T08:59:25Z

使用infer_rec + svtrv2推理单张图片时，出现问题：

推理代码：

from tools.infer_rec import OpenRecognizer
import yaml
from PIL import Image

config = yaml.load(open('/home/server/python/other_work/OCR/OpenOCR/models/svtrv2_smtr_gtc_ch/config.yml', 'r', encoding='utf-8'), Loader=yaml.FullLoader)
ocr = OpenRecognizer(config=config)
image = Image.open("/home/server/python/other_work/OCR/OpenOCR/test.jpg")
ocr(img_numpy=image)

推测是因为使用了ctc + gtc 解码，但是没进行融合或加权，一个图片推理出现了两个结果；最后输出的部分也出现了问题：

The text was updated successfully, but these errors were encountered:

Topdu · 2024-12-27T13:56:56Z

The following changes to the configuration file are required for svtrv2 to perform inference:

...
  Decoder:
    name: GTCDecoder
    infer_gtc: False # True to False
    detach: False
    gtc_decoder:
...

Loss:
  name: CTCLoss # GTCLoss to CTCLoss
  # ctc_weight: 0.1 # delete
  # gtc_loss: # delete
   #  name: SMTRLoss # delete

PostProcess:
  name: CTCLabelDecode # GTCLabelDecode to CTCLabelDecode
  # gtc_label_decode: # delete
    # name: SMTRLabelDecode # delete
    # next_mode: *next # delete
  character_dict_path: *character_dict_path
  use_space_char: *use_space_char

Metric:
  name: RecMetric # RecGTCMetric to RecMetric
  main_indicator: acc
  # is_filter: True

...

Eval:
  dataset:
    name: RatioDataSetTVResize
    ds_width: True
    padding: False
    data_dir_list: ['../benchmark_bctr/benchmark_bctr_test/scene_test']
    transforms:
      - DecodeImagePIL: # load image
          img_mode: RGB
      - CTCLabelEncode: # GTCLabelEncode to CTCLabelEncode
          # gtc_label_encode: # delete
            # name: ARLabelEncode # delete
          character_dict_path: *character_dict_path
          use_space_char: *use_space_char
          max_text_length: *max_text_length
      - KeepKeys:
          # keep_keys: ['image', 'label', 'length', 'ctc_label', 'ctc_length'] to
          keep_keys: ['image', 'label', 'length']
  sampler:
    name: RatioSampler
    scales: [[128, 32]] # w, h
    # divide_factor: to ensure the width and height dimensions can be devided by downsampling multiple
    first_bs: *bs
    fix_bs: false
    divided_factor: [4, 16] # w, h
    is_training: False
  loader:
    shuffle: False
    drop_last: False
    batch_size_per_card: *bs
    max_ratio: *max_ratio
    num_workers: 4

super-tian · 2024-12-31T08:51:18Z

感谢回复，推理问题已解决；在训练中碰到了新的问题，具体如下：

系统环境：

Ubuntu 20.04.6 LTS

python 3.11.7

CUDA Device Count: 2

Device 0: Tesla V100-SXM2-32GB

Device 1: Tesla V100-SXM2-32GB

torch: 2.1.1

cudnn: 8700

推理代码：
CUDA_VISIBLE_DEVICES=0,1 python -m torch.distributed.launch --nproc_per_node=2 tools/train_rec.py --c runs/train0/svtrv2_smtr_gtc_rctc_ch.yml
训练文件：svtrv2_smtr_gtc_rctc_ch.txt
使用预训练模型为：

出现了下方的异常，

(base) server@server:~/python/other_work/OCR/OpenOCR$ CUDA_VISIBLE_DEVICES=0,1 python -m torch.distributed.launch --nproc_per_node=2 tools/train_rec.py --c runs/train0/config.yml
/home/server/anaconda3/lib/python3.11/site-packages/torch/distributed/launch.py:181: FutureWarning: The module torch.distributed.launch is deprecated
and will be removed in future. Use torchrun.
Note that --use-env is set by default in torchrun.
If your script expects `--local-rank` argument to be set, please
change it to read from `os.environ['LOCAL_RANK']` instead. See 
https://pytorch.org/docs/stable/distributed.html#launch-utility for 
further instructions

  warnings.warn(
[2024/12/31 16:45:50] openrec INFO: ----------- Config -----------
[2024/12/31 16:45:50] openrec INFO: Architecture : 
[2024/12/31 16:45:50] openrec INFO:     Decoder : 
[2024/12/31 16:45:50] openrec INFO:         ctc_decoder : 
[2024/12/31 16:45:50] openrec INFO:             name : RCTCDecoder
[2024/12/31 16:45:50] openrec INFO:             only_attn : False
[2024/12/31 16:45:50] openrec INFO:         detach : False
[2024/12/31 16:45:50] openrec INFO:         gtc_decoder : 
[2024/12/31 16:45:50] openrec INFO:             ds : True
[2024/12/31 16:45:50] openrec INFO:             max_len : 25
[2024/12/31 16:45:50] openrec INFO:             name : SMTRDecoder
[2024/12/31 16:45:50] openrec INFO:             next_mode : True
[2024/12/31 16:45:50] openrec INFO:             num_layer : 1
[2024/12/31 16:45:50] openrec INFO:             sub_str_len : 5
[2024/12/31 16:45:50] openrec INFO:         infer_gtc : True
[2024/12/31 16:45:50] openrec INFO:         name : GTCDecoder
[2024/12/31 16:45:50] openrec INFO:     Encoder : 
[2024/12/31 16:45:50] openrec INFO:         depths : [6, 6, 6]
[2024/12/31 16:45:50] openrec INFO:         dims : [128, 256, 384]
[2024/12/31 16:45:50] openrec INFO:         feat2d : True
[2024/12/31 16:45:50] openrec INFO:         last_stage : False
[2024/12/31 16:45:50] openrec INFO:         local_k : [[5, 5], [5, 5], [-1, -1]]
[2024/12/31 16:45:50] openrec INFO:         mixer : [['Conv', 'Conv', 'Conv', 'Conv', 'Conv', 'Conv'], ['Conv', 'Conv', 'FGlobal', 'Global', 'Global', 'Global'], ['Global', 'Global', 'Global', 'Global', 'Global', 'Global']]
[2024/12/31 16:45:50] openrec INFO:         name : SVTRv2LNConvTwo33
[2024/12/31 16:45:50] openrec INFO:         num_convs : [[2, 2, 2, 2, 2, 2], [2, 2, 2, 2, 2, 2], [3, 3, 3, 3, 3, 3]]
[2024/12/31 16:45:50] openrec INFO:         num_heads : [4, 8, 12]
[2024/12/31 16:45:50] openrec INFO:         out_channels : 256
[2024/12/31 16:45:50] openrec INFO:         sub_k : [[1, 1], [2, 1], [-1, -1]]
[2024/12/31 16:45:50] openrec INFO:         use_pos_embed : False
[2024/12/31 16:45:50] openrec INFO:     Transform : None
[2024/12/31 16:45:50] openrec INFO:     algorithm : BGPD
[2024/12/31 16:45:50] openrec INFO:     in_channels : 3
[2024/12/31 16:45:50] openrec INFO:     model_type : rec
[2024/12/31 16:45:50] openrec INFO: Eval : 
[2024/12/31 16:45:50] openrec INFO:     dataset : 
[2024/12/31 16:45:50] openrec INFO:         data_dir_list : ['datas/handwriting_ic13_test_images', 'datas/HW_Chinese_test', 'datas/HWDB2.0Test_label', 'datas/HWDB2.1Test_label']
[2024/12/31 16:45:50] openrec INFO:         ds_width : True
[2024/12/31 16:45:50] openrec INFO:         name : RatioDataSetTVResize
[2024/12/31 16:45:50] openrec INFO:         padding : False
[2024/12/31 16:45:50] openrec INFO:         transforms : 
[2024/12/31 16:45:50] openrec INFO:             DecodeImagePIL : 
[2024/12/31 16:45:50] openrec INFO:                 img_mode : RGB
[2024/12/31 16:45:50] openrec INFO:             GTCLabelEncode : 
[2024/12/31 16:45:50] openrec INFO:                 character_dict_path : ./tools/utils/ppocr_keys_v1.txt
[2024/12/31 16:45:50] openrec INFO:                 gtc_label_encode : 
[2024/12/31 16:45:50] openrec INFO:                     name : ARLabelEncode
[2024/12/31 16:45:50] openrec INFO:                 max_text_length : 25
[2024/12/31 16:45:50] openrec INFO:                 use_space_char : False
[2024/12/31 16:45:50] openrec INFO:             KeepKeys : 
[2024/12/31 16:45:50] openrec INFO:                 keep_keys : ['image', 'label', 'length', 'ctc_label', 'ctc_length']
[2024/12/31 16:45:50] openrec INFO:     loader : 
[2024/12/31 16:45:50] openrec INFO:         batch_size_per_card : 256
[2024/12/31 16:45:50] openrec INFO:         drop_last : False
[2024/12/31 16:45:50] openrec INFO:         max_ratio : 8
[2024/12/31 16:45:50] openrec INFO:         num_workers : 4
[2024/12/31 16:45:50] openrec INFO:         shuffle : False
[2024/12/31 16:45:50] openrec INFO:     sampler : 
[2024/12/31 16:45:50] openrec INFO:         divided_factor : [4, 16]
[2024/12/31 16:45:50] openrec INFO:         first_bs : 256
[2024/12/31 16:45:50] openrec INFO:         fix_bs : False
[2024/12/31 16:45:50] openrec INFO:         is_training : False
[2024/12/31 16:45:50] openrec INFO:         name : RatioSampler
[2024/12/31 16:45:50] openrec INFO:         scales : [[128, 32]]
[2024/12/31 16:45:50] openrec INFO: Global : 
[2024/12/31 16:45:50] openrec INFO:     cal_metric_during_train : False
[2024/12/31 16:45:50] openrec INFO:     character_dict_path : ./tools/utils/ppocr_keys_v1.txt
[2024/12/31 16:45:50] openrec INFO:     checkpoints : None
[2024/12/31 16:45:50] openrec INFO:     device : gpu
[2024/12/31 16:45:50] openrec INFO:     distributed : True
[2024/12/31 16:45:50] openrec INFO:     epoch_num : 100
[2024/12/31 16:45:50] openrec INFO:     eval_batch_step : [0, 2000]
[2024/12/31 16:45:50] openrec INFO:     eval_epoch_step : [0, 1]
[2024/12/31 16:45:50] openrec INFO:     grad_clip_val : 20
[2024/12/31 16:45:50] openrec INFO:     infer_img : None
[2024/12/31 16:45:50] openrec INFO:     log_smooth_window : 20
[2024/12/31 16:45:50] openrec INFO:     max_text_length : 25
[2024/12/31 16:45:50] openrec INFO:     output_dir : ./train0/rec/ch/svtrv2_smtr_gtc_rctc_ch
[2024/12/31 16:45:50] openrec INFO:     pretrained_model : models/svtrv2_smtr_gtc_ch/best.pth
[2024/12/31 16:45:50] openrec INFO:     print_batch_step : 10
[2024/12/31 16:45:50] openrec INFO:     save_epoch_step : 1
[2024/12/31 16:45:50] openrec INFO:     save_res_path : ./output/rec/predicts_smtr.txt
[2024/12/31 16:45:50] openrec INFO:     use_amp : True
[2024/12/31 16:45:50] openrec INFO:     use_space_char : False
[2024/12/31 16:45:50] openrec INFO:     use_tensorboard : False
[2024/12/31 16:45:50] openrec INFO: LRScheduler : 
[2024/12/31 16:45:50] openrec INFO:     cycle_momentum : False
[2024/12/31 16:45:50] openrec INFO:     name : OneCycleLR
[2024/12/31 16:45:50] openrec INFO:     warmup_epoch : 5
[2024/12/31 16:45:50] openrec INFO: Loss : 
[2024/12/31 16:45:50] openrec INFO:     ctc_weight : 0.1
[2024/12/31 16:45:50] openrec INFO:     gtc_loss : 
[2024/12/31 16:45:50] openrec INFO:         name : SMTRLoss
[2024/12/31 16:45:50] openrec INFO:     name : GTCLoss
[2024/12/31 16:45:50] openrec INFO: Metric : 
[2024/12/31 16:45:50] openrec INFO:     main_indicator : acc
[2024/12/31 16:45:50] openrec INFO:     name : RecGTCMetric
[2024/12/31 16:45:50] openrec INFO: Optimizer : 
[2024/12/31 16:45:50] openrec INFO:     filter_bias_and_bn : True
[2024/12/31 16:45:50] openrec INFO:     lr : 0.00065
[2024/12/31 16:45:50] openrec INFO:     name : AdamW
[2024/12/31 16:45:50] openrec INFO:     weight_decay : 0.05
[2024/12/31 16:45:50] openrec INFO: PostProcess : 
[2024/12/31 16:45:50] openrec INFO:     character_dict_path : ./tools/utils/ppocr_keys_v1.txt
[2024/12/31 16:45:50] openrec INFO:     gtc_label_decode : 
[2024/12/31 16:45:50] openrec INFO:         name : SMTRLabelDecode
[2024/12/31 16:45:50] openrec INFO:         next_mode : True
[2024/12/31 16:45:50] openrec INFO:     name : GTCLabelDecode
[2024/12/31 16:45:50] openrec INFO:     use_space_char : False
[2024/12/31 16:45:50] openrec INFO: Train : 
[2024/12/31 16:45:50] openrec INFO:     dataset : 
[2024/12/31 16:45:50] openrec INFO:         data_dir_list : ['datas/12', 'datas/handwriting_hwdb_train_images', 'datas/HW_Chinese_train', 'datas/HWDB2.0Train_label', 'datas/HWDB2.1Train_label', 'datas/HWDB2.2Train_label', 'datas/local_lmdb_split_note', 'datas/tal_ocr_eng']
[2024/12/31 16:45:50] openrec INFO:         ds_width : True
[2024/12/31 16:45:50] openrec INFO:         name : RatioDataSetTVResize
[2024/12/31 16:45:50] openrec INFO:         padding : False
[2024/12/31 16:45:50] openrec INFO:         transforms : 
[2024/12/31 16:45:50] openrec INFO:             DecodeImagePIL : 
[2024/12/31 16:45:50] openrec INFO:                 img_mode : RGB
[2024/12/31 16:45:50] openrec INFO:             PARSeqAugPIL : None
[2024/12/31 16:45:50] openrec INFO:             GTCLabelEncode : 
[2024/12/31 16:45:50] openrec INFO:                 character_dict_path : ./tools/utils/ppocr_keys_v1.txt
[2024/12/31 16:45:50] openrec INFO:                 gtc_label_encode : 
[2024/12/31 16:45:50] openrec INFO:                     name : SMTRLabelEncode
[2024/12/31 16:45:50] openrec INFO:                     sub_str_len : 5
[2024/12/31 16:45:50] openrec INFO:                 max_text_length : 25
[2024/12/31 16:45:50] openrec INFO:                 use_space_char : False
[2024/12/31 16:45:50] openrec INFO:             KeepKeys : 
[2024/12/31 16:45:50] openrec INFO:                 keep_keys : ['image', 'label', 'label_subs', 'label_next', 'length_subs', 'label_subs_pre', 'label_next_pre', 'length_subs_pre', 'length', 'ctc_label', 'ctc_length']
[2024/12/31 16:45:50] openrec INFO:     loader : 
[2024/12/31 16:45:50] openrec INFO:         batch_size_per_card : 256
[2024/12/31 16:45:50] openrec INFO:         drop_last : True
[2024/12/31 16:45:50] openrec INFO:         max_ratio : 8
[2024/12/31 16:45:50] openrec INFO:         num_workers : 4
[2024/12/31 16:45:50] openrec INFO:         shuffle : True
[2024/12/31 16:45:50] openrec INFO:     sampler : 
[2024/12/31 16:45:50] openrec INFO:         divided_factor : [4, 16]
[2024/12/31 16:45:50] openrec INFO:         first_bs : 256
[2024/12/31 16:45:50] openrec INFO:         fix_bs : False
[2024/12/31 16:45:50] openrec INFO:         is_training : True
[2024/12/31 16:45:50] openrec INFO:         name : RatioSampler
[2024/12/31 16:45:50] openrec INFO:         scales : [[128, 32]]
[2024/12/31 16:45:50] openrec INFO: config : runs/train0/config.yml
[2024/12/31 16:45:50] openrec INFO: eval : True
[2024/12/31 16:45:50] openrec INFO: filename : config
[2024/12/31 16:45:50] openrec INFO: local_rank : 0
[2024/12/31 16:45:50] openrec INFO: ---------------------------------------------
[2024-12-31 16:45:52,585] torch.distributed.elastic.multiprocessing.api: [ERROR] failed (exitcode: -7) local_rank: 0 (pid: 467397) of binary: /home/server/anaconda3/bin/python
Traceback (most recent call last):
  File "<frozen runpy>", line 198, in _run_module_as_main
  File "<frozen runpy>", line 88, in _run_code
  File "/home/server/anaconda3/lib/python3.11/site-packages/torch/distributed/launch.py", line 196, in <module>
    main()
  File "/home/server/anaconda3/lib/python3.11/site-packages/torch/distributed/launch.py", line 192, in main
    launch(args)
  File "/home/server/anaconda3/lib/python3.11/site-packages/torch/distributed/launch.py", line 177, in launch
    run(args)
  File "/home/server/anaconda3/lib/python3.11/site-packages/torch/distributed/run.py", line 797, in run
    elastic_launch(
  File "/home/server/anaconda3/lib/python3.11/site-packages/torch/distributed/launcher/api.py", line 134, in __call__
    return launch_agent(self._config, self._entrypoint, list(args))
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/server/anaconda3/lib/python3.11/site-packages/torch/distributed/launcher/api.py", line 264, in launch_agent
    raise ChildFailedError(
torch.distributed.elastic.multiprocessing.errors.ChildFailedError: 
======================================================
tools/train_rec.py FAILED
------------------------------------------------------
Failures:
[1]:
  time      : 2024-12-31_16:45:52
  host      : server
  rank      : 1 (local_rank: 1)
  exitcode  : -7 (pid: 467398)
  error_file: <N/A>
  traceback : Signal 7 (SIGBUS) received by PID 467398
------------------------------------------------------
Root Cause (first observed failure):
[0]:
  time      : 2024-12-31_16:45:52
  host      : server
  rank      : 0 (local_rank: 0)
  exitcode  : -7 (pid: 467397)
  error_file: <N/A>
  traceback : Signal 7 (SIGBUS) received by PID 467397
======================================================

Topdu · 2025-01-02T03:21:12Z

Are there more error logs When running training commands, the information provided so far is not able to determine the error.

super-tian · 2025-01-02T05:47:31Z

Are there more error logs When running training commands, the information provided so far is not able to determine the error.

The above are all the running logs, but I checked the issue records of Torch and found that it may be a version issue. I solved the problem by replacing the Torch version，now，my torch version is 2.0.1

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

使用infer_rec时出现问题 #71

使用infer_rec时出现问题 #71

super-tian commented Dec 27, 2024

Topdu commented Dec 27, 2024

super-tian commented Dec 31, 2024

Topdu commented Jan 2, 2025

super-tian commented Jan 2, 2025

使用infer_rec时出现问题 #71

使用infer_rec时出现问题 #71

Comments

super-tian commented Dec 27, 2024

Topdu commented Dec 27, 2024

super-tian commented Dec 31, 2024

感谢回复，推理问题已解决；在训练中碰到了新的问题，具体如下：

Ubuntu 20.04.6 LTS

python 3.11.7

CUDA Device Count: 2

Device 0: Tesla V100-SXM2-32GB

Device 1: Tesla V100-SXM2-32GB

torch: 2.1.1

cudnn: 8700

Topdu commented Jan 2, 2025

super-tian commented Jan 2, 2025