微调PaddleOCR识别模型时遭遇预训练参数不匹配警告及递归错误的求助

我最近一直在尝试微调PaddleOCR的识别模型，用来处理OCR-A字体的字符——默认模型对这类字体识别不准，比如会把1认成L，6认成b。我查了好多文章和官方文档，但还是遇到一堆问题，而且感觉文档写得有点模糊。

我的操作流程 #

1. 用PPOCRLabel标注图片，生成了Label.txt标注文件；
2. 基于rec_mv3_none_bilstm_ctc配置文件修改适配我的数据集；
3. 执行的训练命令：

!python3 /content/PaddleOCR/tools/train.py -c /content/config.yml -o Global.pretrained_model=/content/pretrain_models_rec/en_PP-OCRv3_rec_train/best_accuracy

遇到的问题 #

1. 预训练参数不匹配警告

命令执行后先弹出了一串参数不匹配的警告，虽然最后提示预训练模型加载成功，但还是担心有问题：

[2024/12/09 03:43:11] ppocr WARNING: The pretrained params head.sar_head.decoder.rnn_decoder.1.cell.weight_ih not in model
[2024/12/09 03:43:11] ppocr WARNING: The pretrained params head.sar_head.decoder.rnn_decoder.1.cell.weight_hh not in model
[2024/12/09 03:43:11] ppocr WARNING: The pretrained params head.sar_head.decoder.rnn_decoder.1.cell.bias_ih not in model
[2024/12/09 03:43:11] ppocr WARNING: The pretrained params head.sar_head.decoder.rnn_decoder.1.cell.bias_hh not in model
[2024/12/09 03:43:11] ppocr WARNING: The pretrained params head.sar_head.decoder.embedding.weight not in model
[2024/12/09 03:43:11] ppocr WARNING: The pretrained params head.sar_head.decoder.prediction.weight not in model
[2024/12/09 03:43:11] ppocr WARNING: The pretrained params head.sar_head.decoder.prediction.bias not in model
[2024/12/09 03:43:11] ppocr INFO: load pretrain successful from /content/pretrain_models_rec/en_PP-OCRv3_rec_train/best_accuracy

2. 致命的递归错误

警告之后，程序直接抛出了递归错误，训练中断，完整报错栈如下：

Exception in thread Thread-1 (_thread_loop):
Traceback (most recent call last):
  File "/usr/lib/python3.10/threading.py", line 1016, in _bootstrap_inner
    self.run()
  File "/usr/lib/python3.10/threading.py", line 953, in run
    self._target(*self._args, **self._kwargs)
  File "/usr/local/lib/python3.10/dist-packages/paddle/io/dataloader/dataloader_iter.py", line 603, in _thread_loop
    batch = self._get_data()
  File "/usr/local/lib/python3.10/dist-packages/paddle/io/dataloader/dataloader_iter.py", line 752, in _get_data
    batch.reraise()
  File "/usr/local/lib/python3.10/dist-packages/paddle/io/dataloader/worker.py", line 187, in reraise
    raise self.exc_type(msg)
RecursionError: DataLoader worker(0) caught RecursionError with message:
Traceback (most recent call last):
  File "/usr/local/lib/python3.10/dist-packages/paddle/io/dataloader/worker.py", line 372, in _worker_loop
    batch = fetcher.fetch(indices)
  File "/usr/local/lib/python3.10/dist-packages/paddle/io/dataloader/fetcher.py", line 77, in fetch
    data.append(self.dataset[idx])
  File "/content/PaddleOCR/ppocr/data/simple_dataset.py", line 163, in __getitem__
    return self.__getitem__(rnd_idx)
  File "/content/PaddleOCR/ppocr/data/simple_dataset.py", line 163, in __getitem__
    return self.__getitem__(rnd_idx)
  File "/content/PaddleOCR/ppocr/data/simple_dataset.py", line 163, in __getitem__
    return self.__getitem__(rnd_idx)
  [Previous line repeated 7 more times]
  File "/content/PaddleOCR/ppocr/data/simple_dataset.py", line 161, in __getitem__
    raise RecursionError("Maximum recursion depth exceeded in __getitem__")
RecursionError: Maximum recursion depth exceeded in __getitem__

Traceback (most recent call last):
  File "/content/PaddleOCR/tools/train.py", line 269, in <module>
    main(config, device, logger, vdl_writer, seed)
  File "/content/PaddleOCR/tools/train.py", line 222, in main
    program.train(
  File "/content/PaddleOCR/tools/program.py", line 312, in train
    for idx, batch in enumerate(train_dataloader):
  File "/usr/local/lib/python3.10/dist-packages/paddle/io/dataloader/dataloader_iter.py", line 826, in __next__
    self._reader.read_next_list()[0]
SystemError: (Fatal) Blocking queue is killed because the data reader raises an exception.
  [Hint: Expected killed_ != true, but received killed_:1 == true:1.] (at /paddle/paddle/fluid/operators/reader/blocking_queue.h:175)

已尝试的排查动作 #

我已经反复检查过配置文件里的图片路径、标注文件路径，确认路径是正确的，但问题依然存在。

想请教的问题 #

1. 这些预训练参数不匹配的警告是什么原因导致的？会不会影响后续的训练效果？
2. 这个递归错误到底怎么解决？看起来是Dataset的__getitem__方法无限递归了，和我的数据集配置或者标注文件有关吗？

备注：内容来源于stack exchange，提问作者l3126r