KIE: re train error

[hbo-lambda]

SER模型训练时，尝试在class_list_xfun.txt中添加了4个label，如下

OTHER
QUESTION
ANSWER
HEADER
NAME_KEY
NAME_VAL
SEX_KEY
SEX_VAL

SER的模型已训练完毕。想训练RE模型时，沿用SER的class_list_xfun.txt标签，却出现如下报错

  File "/home/aistudio/PaddleOCR/ppocr/data/simple_dataset.py", line 137, in __getitem__
    outs = transform(data, self.ops)
  File "/home/aistudio/PaddleOCR/ppocr/data/imaug/__init__.py", line 57, in transform
    data = op(data)
  File "/home/aistudio/PaddleOCR/ppocr/data/imaug/vqa/token/vqa_token_chunk.py", line 109, in __call__
    item['entities']['label'] = [
  File "/home/aistudio/PaddleOCR/ppocr/data/imaug/vqa/token/vqa_token_chunk.py", line 110, in <listcomp>
    self.entities_labels[x] for x in item['entities']['label']
KeyError: 'NAME_KEY'

RE配置文件 configs/kie/vi_layoutxlm/re_vi_layoutxlm_xfund_zh.yml 中的class_path指定路径无误，配置文件还有其他地方涉及到label的地方吗

[tink2123]

看报错是这个位置：

PaddleOCR/ppocr/data/imaug/vqa/token/vqa_token_chunk.py

Line 56 in 18ddb6d

self.entities_labels = {

可以修改配置文件：

      - VQAReTokenChunk:
          max_seq_len: *max_seq_len
          entities_labels: { 'HEADER': 0, 'QUESTION': 1, 'ANSWER': 2, 'NAME_KEY':3, 'SEX_KEY':4, 'SEX_VAL':5}

[hbo-lambda]

ok，谢谢

[harshsummit]

After this, it will give float point exception error

[harshsummit]

Hey @hbo-lambda if you are able to successfully train this RE model, please let me know, I need help for the same

[hbo-lambda]

Hey @hbo-lambda if you are able to successfully train this RE model, please let me know, I need help for the same
fail. !!!!!

[ariefwijaya]

any update on this?

[github-actions]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed in 7 days if no further activity occurs. Thank you for your contributions.

[bagaabu]

same issue, any update? or anyone run the traing can help?

[c-avan]

我也出现这个问题了，配置文件添加entities_labels:{}后出现如下错误：
File "/usr/local/lib/python3.10/dist-packages/paddle/tensor/manipulation.py", line 1842, in stack
return _C_ops.stack(x, axis)
ValueError: (InvalidArgument) x dim number should greater than 0, but received value is: 0
[Hint: Expected x_dim > 0, but received x_dim:0 <= 0:0.] (at ../paddle/phi/backends/gpu/gpu_launch_config.h:175)

[woguwo]

我也出现这个问题了，配置文件添加entities_labels:{}后出现如下错误： File "/usr/local/lib/python3.10/dist-packages/paddle/tensor/manipulation.py", line 1842, in stack return _C_ops.stack(x, axis) ValueError: (InvalidArgument) x dim number should greater than 0, but received value is: 0 [Hint: Expected x_dim > 0, but received x_dim:0 <= 0:0.] (at ../paddle/phi/backends/gpu/gpu_launch_config.h:175)

您好，我在训练kie的时候也出现一样问题了，您解决这个问题了吗？

[XiYuan68]

@c-avan @woguwo 看这里
#9184 (comment)

[UserWangZz]

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

[freezehe]

@tink2123 这个bug修复了吗？我现在也遇到了这样的问题？是否有解决方案

[freezehe]

@tink2123 这个bug修复了吗？我现在也遇到了这样的问题？是否有解决方案

@XiYuan68 你好，这个bug修复了吗？

[chairc]

我测了一下出一个一般的解决方案
1、KeyError参考 #9184 (comment) 修改entities_labels为自己的类别
2、可能会报[Hint: Expected x_dim > 0, but received x_dim:0 <= 0:0.]错误，参考#11261 (comment)，这是因为数据集的label的顺序问题导致的，可以修改部分代码跳过报错问题解决
3、潜在版本问题会报AttributeError: 'bool' object has no attribute 'sum'错误，这是paddlenlp新版本修改了部分代码导致的，参考#13368 (comment)降低版本paddlepaddle==2.4.1 paddlenlp==2.5.2