使用 TF2.2 执行 TF1.X 代码会导致错误

Question

如何修复 Tensorflow 2.2 中的以下错误？如果可能的话，我更喜欢将代码转换为兼容 TF2.2 的代码，而不是使用紧凑版本。

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

---

[3306:3298 0:1022] 01:57:24 Tue Dec 29 [mona@goku:pts/0 +1] ~/research/code/DJ-RN/pointnet
$ python train.py 
Traceback (most recent call last):
  File "train.py", line 260, in <module>
    train()
  File "train.py", line 96, in train
    pointclouds_pl, labels_pl = MODEL.placeholder_inputs(BATCH_SIZE, NUM_POINT)
  File "/home/mona/research/code/DJ-RN/pointnet/models/pointnet_cls.py", line 13, in placeholder_inputs
    pointclouds_pl = tf.placeholder(tf.float32, shape=(batch_size, num_point, 3))
AttributeError: module 'tensorflow' has no attribute 'placeholder'

还有

[3306:3298 0:1023] 01:57:31 Tue Dec 29 [mona@goku:pts/0 +1] ~/research/code/DJ-RN/pointnet
$ python
Python 3.8.5 (default, Sep  4 2020, 07:30:14) 
[GCC 7.3.0] :: Anaconda, Inc. on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> import tensorflow as tf
>>> tf.__version__
'2.2.0'
>>> quit()
12149/31772MB
[3306:3298 0:1024] 01:59:05 Tue Dec 29 [mona@goku:pts/0 +1] ~/research/code/DJ-RN/pointnet
$ nvcc --version
nvcc: NVIDIA (R) Cuda compiler driver
Copyright (c) 2005-2019 NVIDIA Corporation
Built on Sun_Jul_28_19:07:16_PDT_2019
Cuda compilation tools, release 10.1, V10.1.243
12149/31772MB

$ lsb_release -a
LSB Version:    core-11.1.0ubuntu2-noarch:security-11.1.0ubuntu2-noarch
Distributor ID: Ubuntu
Description:    Ubuntu 20.04.1 LTS
Release:    20.04
Codename:   focal

如下所示，占位符不是方法：

>>> tf.compat.v1.summary.
tf.compat.v1.summary.Event(                    tf.compat.v1.summary.get_summary_description(
tf.compat.v1.summary.FileWriter(               tf.compat.v1.summary.histogram(
tf.compat.v1.summary.FileWriterCache(          tf.compat.v1.summary.image(
tf.compat.v1.summary.SessionLog(               tf.compat.v1.summary.initialize(
tf.compat.v1.summary.Summary(                  tf.compat.v1.summary.merge(
tf.compat.v1.summary.SummaryDescription(       tf.compat.v1.summary.merge_all(
tf.compat.v1.summary.TaggedRunMetadata(        tf.compat.v1.summary.scalar(
tf.compat.v1.summary.all_v2_summary_ops(       tf.compat.v1.summary.tensor_summary(
tf.compat.v1.summary.audio(                    tf.compat.v1.summary.text(

我也尝试了以下在论坛和 git 问题中提到的导入，但是它不起作用（它也在用于代码迁移的官方 tensorflow 文档中：https://www.tensorflow.org/guide/migrate）：

import tensorflow.compat.v1 as tf
tf.disable_v2_behavior()

这是我得到的错误：

$ python train.py 
WARNING:tensorflow:From /home/mona/anaconda3/lib/python3.8/site-packages/tensorflow/python/compat/v2_compat.py:96: disable_resource_variables (from tensorflow.python.ops.variable_scope) is deprecated and will be removed in a future version.
Instructions for updating:
non-resource variables are not supported in the long term
Traceback (most recent call last):
  File "train.py", line 260, in <module>
    train()
  File "train.py", line 96, in train
    pointclouds_pl, labels_pl = MODEL.placeholder_inputs(BATCH_SIZE, NUM_POINT)
  File "/home/mona/research/code/DJ-RN/pointnet/models/pointnet_cls.py", line 15, in placeholder_inputs
    pointclouds_pl = tf.compact.v1.placeholder(tf.float32, shape=(batch_size, num_point, 3))
AttributeError: module 'tensorflow.compat.v1' has no attribute 'compact'

代码位于此 repo 中：https://github.com/charlesq34/pointnet/issues/265

Answer 1

import tensorflow.compat.v1 as tf
tf.disable_v2_behavior()

而不是

import tensorflow as tf

然后像这样使用它：

tf.placeholder()

例如，

pointclouds_ph = tf.placeholder(tf.float32, shape=(batch_size, point_num, 3))