bert 无法运行 Colab

5uzkadbs  于 5个月前  发布在  其他
关注(0)|答案(3)|浏览(61)

在尝试运行 colab 时,多个地方出现问题...
那一行

from bert import run_classifier

失败了(因为 TensorFlow 不再有一个名为 contrib 的模块)。堆栈跟踪如下:

---------------------------------------------------------------------------
ModuleNotFoundError                       Traceback (most recent call last)
<ipython-input-4-1ccb11d8dffa> in <module>()
      1 import bert
----> 2 from bert import run_classifier
      3 from bert import optimization
      4 from bert import tokenization

1 frames
/usr/local/lib/python3.6/dist-packages/bert/run_classifier.py in <module>()
     22 import csv
     23 import os
---> 24 from bert import modeling
     25 from bert import optimization
     26 from bert import tokenization

/usr/local/lib/python3.6/dist-packages/bert/modeling.py in <module>()
     27 import six
     28 import tensorflow.compat.v1 as tf
---> 29 from tensorflow.contrib import layers as contrib_layers
     30 
     31 

ModuleNotFoundError: No module named 'tensorflow.contrib'

---------------------------------------------------------------------------
NOTE: If your import is failing due to a missing package, you can
manually install dependencies using either !pip or !apt.

To view examples of installing some common dependencies, click the
"Open Examples" button below.
---------------------------------------------------------------------------

然后在下面的单元格中,它允许你为模型选择输出目录:

---------------------------------------------------------------------------
AttributeError                            Traceback (most recent call last)
<ipython-input-5-ba2fc219037b> in <module>()
     20     # Doesn't matter if the directory didn't exist
     21     pass
---> 22 tf.gfile.MakeDirs(OUTPUT_DIR)
     23 print('***** Model output directory: {} *****'.format(OUTPUT_DIR))

AttributeError: module 'tensorflow' has no attribute 'gfile'

新的 TensorFlow 版本需要在这里进行更新。

d7v8vwbk

d7v8vwbk1#

将所有tf.gfile的调用替换为tf.io.gfile(然后也修复所有其他调用,因为tf.gile.MakeDirs现在是tf.io.gfile.makedirs等,所以不仅仅是sed替换)

x0fgdtte

x0fgdtte2#

我应该使用tf版本<=1.5吗?

omjgkv6w

omjgkv6w3#

switch to tf 1.x solved all my import problems. Here's the code that I use (make sure to restart runtime before running it)

%tensorflow_version 1.x
import tensorflow
print(tensorflow.__version__)

相关问题