发票关键信息抽取.md FileNotFoundError: configuration file<config.json> or <model_config.json> not found

[greatliu]

请提供下述完整信息以便快速定位问题/Please provide the following information to quickly locate the problem

• 系统环境/System Environment：win11 python3.9 cuda11.8
• 版本号/Version：
  paddlenlp 2.6.1
  paddleocr 2.7.3
  paddlepaddle-gpu 2.6.1
• 问题相关组件/Related components：不会确定
• 运行指令/Command Code：
  python tools/infer_kie_token_ser.py -c ./configs/kie/vi_layoutxlm/ser_vi_layoutxlm_xfund_zh_udml.yml -o Architecture.Models.Student.Backbone.checkpoints="./fapiao/ser_vi_layoutxlm_fapiao_trained/best_accuracy/" Global.infer_img=./train_data/zzsfp/val.json Global.infer_mode=False
• 完整报错/Complete Error Message：
  [2024-05-10 16:38:43,480] [ WARNING] - Some weights of LayoutXLMForTokenClassification were not initialized from the model checkpoint at vi-layoutxlm-base-uncased and are newly initialized: ['classifier.bias', 'classifier.weight']
  You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference.
  Traceback (most recent call last):
  File "E:\Workspace\GitHub\PaddleOCR\tools\infer_kie_token_ser.py", line 119, in
  ser_engine = SerPredictor(config)
  File "E:\Workspace\GitHub\PaddleOCR\tools\infer_kie_token_ser.py", line 68, in init
  self.model = build_model(config['Architecture'])
  File "E:\Workspace\GitHub\PaddleOCR\ppocr\modeling\architectures_init_.py", line 34, in build_model
  arch = getattr(mod, name)(config)
  File "E:\Workspace\GitHub\PaddleOCR\ppocr\modeling\architectures\distillation_model.py", line 47, in init
  model = BaseModel(model_config)
  File "E:\Workspace\GitHub\PaddleOCR\ppocr\modeling\architectures\base_model.py", line 55, in init
  self.backbone = build_backbone(config["Backbone"], model_type)
  File "E:\Workspace\GitHub\PaddleOCR\ppocr\modeling\backbones_init_.py", line 82, in build_backbone
  module_class = eval(module_name)(**config)
  File "E:\Workspace\GitHub\PaddleOCR\ppocr\modeling\backbones\vqa_layoutlm.py", line 142, in init
  super(LayoutXLMForSer, self).init(
  File "E:\Workspace\GitHub\PaddleOCR\ppocr\modeling\backbones\vqa_layoutlm.py", line 55, in init
  self.model = model_class.from_pretrained(checkpoints)
  File "C:\Users\great\miniconda3\envs\Paddle\lib\site-packages\paddlenlp\transformers\model_utils.py", line 1894, in from_pretrained
  config, model_kwargs = cls.config_class.from_pretrained(
  File "C:\Users\great\miniconda3\envs\Paddle\lib\site-packages\paddlenlp\transformers\configuration_utils.py", line 749, in from_pretrained
  config_dict, kwargs = cls.get_config_dict(pretrained_model_name_or_path, **kwargs)
  File "C:\Users\great\miniconda3\envs\Paddle\lib\site-packages\paddlenlp\transformers\configuration_utils.py", line 775, in get_config_dict
  config_dict, kwargs = cls._get_config_dict(
  File "C:\Users\great\miniconda3\envs\Paddle\lib\site-packages\paddlenlp\transformers\configuration_utils.py", line 850, in _get_config_dict
  raise FileNotFoundError(f"configuration file<{CONFIG_NAME}> or <{LEGACY_CONFIG_NAME}> not found")
  FileNotFoundError: configuration file<config.json> or <model_config.json> not found

使用了百度提供的发票预训练模型，想实现发票关键信息抽取的推理。yml用的是https://github.com/PaddlePaddle/PaddleOCR/blob/main/applications/%E5%8F%91%E7%A5%A8%E5%85%B3%E9%94%AE%E4%BF%A1%E6%81%AF%E6%8A%BD%E5%8F%96.md
中“4.3.2 开始训练”中修改的训练yml
有模型下载进度，然后报上述错误。

[zhangyubo0722]

Architecture.Models.Student.Backbone.checkpoints="./fapiao/ser_vi_layoutxlm_fapiao_trained/best_accuracy/ 断点训练权重传入错误，需要填写完整路径

[greatliu]

Architecture.Models.Student.Backbone.checkpoints="./fapiao/ser_vi_layoutxlm_fapiao_trained/best_accuracy/ 断点训练权重传入错误，需要填写完整路径

不好意思，刚上手这个，不太懂。
我看了下我下载的模型下的确没有/best_accuracy/ 这个文件夹。
然后这里应该是需要一个json file吧？例子应该怎样写？

[zhangyubo0722]

先传入正确的路径试一下呢

[greatliu]

先传入正确的路径试一下呢

我的fapiao文件夹下，只有ser_vi_layoutxlm_fapiao_trained和re_vi_layoutxlm_fapiao_trained两个模型文件，再往下就没有文件夹了。
我指向fapiao文件夹运行，
(Paddle) E:\Workspace\GitHub\PaddleOCR>python tools/infer_kie_token_ser.py -c ./configs/kie/vi_layoutxlm/ser_vi_layoutxlm_xfund_zh_udml.yml -o Architecture.Models.Student.Backbone.checkpoints="./fapiao/" Global.infer_img=./train_data/zzsfp/
val.json Global.infer_mode=False

报错为
[2024-05-10 19:30:51,225] [ WARNING] - Some weights of LayoutXLMForTokenClassification were not initialized from the model checkpoint at vi-layoutxlm-base-uncased and are newly initialized: ['classifier.bias', 'classifier.weight']
You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference.
Traceback (most recent call last):
File "E:\Workspace\GitHub\PaddleOCR\tools\infer_kie_token_ser.py", line 119, in
ser_engine = SerPredictor(config)
File "E:\Workspace\GitHub\PaddleOCR\tools\infer_kie_token_ser.py", line 68, in init
self.model = build_model(config['Architecture'])
File "E:\Workspace\GitHub\PaddleOCR\ppocr\modeling\architectures_init_.py", line 34, in build_model
arch = getattr(mod, name)(config)
File "E:\Workspace\GitHub\PaddleOCR\ppocr\modeling\architectures\distillation_model.py", line 47, in init
model = BaseModel(model_config)
File "E:\Workspace\GitHub\PaddleOCR\ppocr\modeling\architectures\base_model.py", line 55, in init
self.backbone = build_backbone(config["Backbone"], model_type)
File "E:\Workspace\GitHub\PaddleOCR\ppocr\modeling\backbones_init_.py", line 82, in build_backbone
module_class = eval(module_name)(**config)
File "E:\Workspace\GitHub\PaddleOCR\ppocr\modeling\backbones\vqa_layoutlm.py", line 142, in init
super(LayoutXLMForSer, self).init(
File "E:\Workspace\GitHub\PaddleOCR\ppocr\modeling\backbones\vqa_layoutlm.py", line 55, in init
self.model = model_class.from_pretrained(checkpoints)
File "C:\Users\great\miniconda3\envs\Paddle\lib\site-packages\paddlenlp\transformers\model_utils.py", line 1894, in from_pretrained
config, model_kwargs = cls.config_class.from_pretrained(
File "C:\Users\great\miniconda3\envs\Paddle\lib\site-packages\paddlenlp\transformers\configuration_utils.py", line 749, in from_pretrained
config_dict, kwargs = cls.get_config_dict(pretrained_model_name_or_path, **kwargs)
File "C:\Users\great\miniconda3\envs\Paddle\lib\site-packages\paddlenlp\transformers\configuration_utils.py", line 775, in get_config_dict
config_dict, kwargs = cls._get_config_dict(
File "C:\Users\great\miniconda3\envs\Paddle\lib\site-packages\paddlenlp\transformers\configuration_utils.py", line 831, in _get_config_dict
raise FileNotFoundError(
FileNotFoundError: please make sure there is model_config.json under the dir, or you can pass the _configuration_file param into from_pretarined method to specific the configuration file name

这好像又回到我提问的第一层的问题了。我猜不是我传参数不对，就是哪里有bug，找不到model_config.json这个文件。

[zhangyubo0722]

上面的报错根本上还是断点训练权重传入错误了，Architecture.Models.Student.Backbone.checkpoints="./fapiao/" 传入断点训练权重时需要具体到哪个模型，比如***.pdparams。

[UserWangZz]

This issue has not been updated for a long time. This issue is temporarily closed and can be reopened if necessary.