UIE训练学生模型报错

[YYGe01]

• 版本、环境信息
  1）PaddleNLP和PaddlePaddle版本：PaddleNLP 2.4.0，PaddlePaddle2.3.1
  2）系统环境：Windows，python3.8
• 复现信息：

python train.py \
    --task_type relation_extraction \
    --train_path student_data/train_data.json \
    --dev_path student_data/dev_data.json \
    --label_maps_path student_data/label_maps.json \
    --num_epochs 200 \
    --encoder ernie-3.0-mini-zh

报错：

Traceback (most recent call last):
  File "C:/Users/49476/glory/work/docparser/DesignDocParser/src/uie/data_distill/train.py", line 194, in <module>
    do_train()
  File "C:/Users/49476/glory/work/docparser/DesignDocParser/src/uie/data_distill/train.py", line 110, in do_train
    loss = sum([criterion(o, l) for o, l in zip(logits, labels)]) / 3
  File "C:/Users/49476/glory/work/docparser/DesignDocParser/src/uie/data_distill/train.py", line 110, in <listcomp>
    loss = sum([criterion(o, l) for o, l in zip(logits, labels)]) / 3
  File "C:\Users\49476\glory\work\docparser\DesignDocParser\src\uie\data_distill\criterion.py", line 59, in __call__
    loss = self._sparse_multilabel_categorical_crossentropy(
  File "C:\Users\49476\glory\work\docparser\DesignDocParser\src\uie\data_distill\criterion.py", line 38, in _sparse_multilabel_categorical_crossentropy
    y_pos_2 = paddle.take_along_axis(y_pred, y_true, axis=-1)
  File "C:\Users\49476\anaconda3\envs\uie\lib\site-packages\paddle\tensor\manipulation.py", line 2854, in take_along_axis
    indices = paddle.broadcast_to(indices, broadcast_shape)
  File "C:\Users\49476\anaconda3\envs\uie\lib\site-packages\paddle\tensor\manipulation.py", line 1917, in broadcast_to
    return _C_ops.expand_v2(x, 'shape', shape)
ValueError: (InvalidArgument) The 2th element of 'shape' for expand_v2 op must be greater than 0, but the value given is 0.
  [Hint: Expected expand_shape[i] > 0, but received expand_shape[i]:0 <= 0:0.] (at ..\paddle\phi\infermeta\unary.cc:511)
  [operator < expand_v2 > error]

经排查发现是labels的第三个维度是零的问题，但某个样本里没有有效信息，是允许的，希望特别处理下这种情况，谢谢。

[dingidng]

参考下面项目，在AIstudio已经跑通
UIE Slim满足工业应用场景，解决推理部署耗时问题，提升效能！
https://aistudio.baidu.com/aistudio/projectdetail/4516470?contributionType=1

[Ecust-ylw]

你好，这个问题后来解决了吗？我也遇到这个问题了