【bert4keras】：AttributeError: ‘PaddedBatchDataset‘ object has no attribute ‘ndim‘

问题

在使用bert4keras的时候遇到一个问题：

AttributeError: 'PaddedBatchDataset' object has no attribute 'ndim'

解决

花了好长时间定位问题，原本以为是keras版本问题，后来才从bert4keras源码看到，作者其实是设置了在import keras之前会判断环境变量中的TF_KERAS，但是我在写代码之前会将这个设置放在main中，导致其实使用的还是keras而不是tf.keras

import os

import json
import math
import random
import numpy as np
import tensorflow as tf
from bert4keras.backend import keras, K # 这里先引用了keras
from bert4keras.layers import Loss
from bert4keras.models import build_transformer_model
from bert4keras.tokenizers import Tokenizer
from bert4keras.optimizers import Adam
from bert4keras.optimizers import extend_with_weight_decay
from bert4keras.optimizers import extend_with_piecewise_linear_lr
from bert4keras.optimizers import extend_with_gradient_accumulation
from bert4keras.snippets import sequence_padding, open
from bert4keras.snippets import DataGenerator
from bert4keras.snippets import text_segmentate
import jieba_fast as jieba
...

if __name__ == '__main__':
	os.environ['TF_KERAS'] = '1'  # 必须使用tf.keras,注意，这里要放在引用keras之前

改为：

import os

import json
import math
import random
import numpy as np
import tensorflow as tf
os.environ['TF_KERAS'] = '1'  # 必须使用tf.keras,注意，这里要放在引用keras之前
from bert4keras.backend import keras, K # 这里先引用了keras
from bert4keras.layers import Loss
from bert4keras.models import build_transformer_model
from bert4keras.tokenizers import Tokenizer
from bert4keras.optimizers import Adam
from bert4keras.optimizers import extend_with_weight_decay
from bert4keras.optimizers import extend_with_piecewise_linear_lr
from bert4keras.optimizers import extend_with_gradient_accumulation
from bert4keras.snippets import sequence_padding, open
from bert4keras.snippets import DataGenerator
from bert4keras.snippets import text_segmentate
import jieba_fast as jieba
...

if __name__ == '__main__':
	...