从Tensorflow运行RNN教程的TypeError

Question

Simply trying to run the unmodified RNN Tensorflow tutorial on a Centos 7 machine. I have yum update d the machine. Installed python3.6, pulled the tutorial from github, downloaded and untarred the sample data. Other tutorials run. Then I run:

python3.6 ptb_word_lm.py --data_path=simple-examples/data/

Which returns:

Traceback (most recent call last):
  File "ptb_word_lm.py", line 374, in <module>
    tf.app.run()
  File "/usr/lib/python3.6/site-packages/tensorflow/python/platform/app.py", line 43, in run
    sys.exit(main(sys.argv[:1] + flags_passthrough))
  File "ptb_word_lm.py", line 334, in main
    train_input = PTBInput(config=config, data=train_data, name="TrainInput")
  File "ptb_word_lm.py", line 94, in __init__
    data, batch_size, num_steps, name=name)
  File "/root/tensorflow/models/tutorials/rnn/ptb/reader.py", line 117, in ptb_producer
    [batch_size, (i + 1) * num_steps])
TypeError: strided_slice() missing 1 required positional argument: 'strides'

As a total NOOB, I am completely stuck! Can anyone help?

Answer 1

Per the comment on the question above, I installed TensorFlow 1.0.0rc1 . I used pip3.6 , but as I was on Centos, the OS Setup guide doesn't address it, so I had to try each download:

(tensorflow) [tensorflow]# export TF_BINARY_URL=https://storage.googleapis.com/tensorflow/linux/cpu/tensorflow-1.0.0rc1-cp35-cp35m-linux_x86_64.whl
(tensorflow) [tensorflow]# pip3.6 install --upgrade $TF_BINARY_URL
tensorflow-1.0.0rc1-cp35-cp35m-linux_x86_64.whl is not a supported wheel on this platform.
(tensorflow) [tensorflow]# export TF_BINARY_URL=https://storage.googleapis.com/tensorflow/linux/cpu/tensorflow-1.0.0rc1-cp34-cp34m-linux_x86_64.whl
(tensorflow) [tensorflow]# pip3.6 install --upgrade $TF_BINARY_URL
tensorflow-1.0.0rc1-cp34-cp34m-linux_x86_64.whl is not a supported wheel on this platform.
(tensorflow) [tensorflow]# export TF_BINARY_URL=https://storage.googleapis.com/tensorflow/linux/cpu/tensorflow-1.0.0rc1-cp27-none-linux_x86_64.whl
(tensorflow) [tensorflow]# pip3.6 install --upgrade $TF_BINARY_URL
tensorflow-1.0.0rc1-cp27-none-linux_x86_64.whl is not a supported wheel on this platform.
(tensorflow) [tensorflow]# export TF_BINARY_URL=https://storage.googleapis.com/tensorflow/linux/cpu/tensorflow-1.0.0rc1-cp33-cp33m-linux_x86_64.whl
(tensorflow) [tensorflow]# pip3.6 install --upgrade $TF_BINARY_URL
tensorflow-1.0.0rc1-cp33-cp33m-linux_x86_64.whl is not a supported wheel on this platform.
(tensorflow) [tensorflow]# export TF_BINARY_URL=https://storage.googleapis.com/tensorflow/linux/cpu/tensorflow-1.0.0rc1-cp36-cp36m-linux_x86_64.whl
(tensorflow) [tensorflow]# pip3.6 install --upgrade $TF_BINARY_URL
    Collecting tensorflow==1.0.0rc1 from https://storage.googleapis.com/tensorflow/linux/cpu/tensorflow-1.0.0rc1-cp36-cp36m-linux_x86_64.whl
  Downloading https://storage.googleapis.com/tensorflow/linux/cpu/tensorflow-1.0.0rc1-cp36-cp36m-linux_x86_64.whl (43.5MB)
    100% |████████████████████████████████| 43.5MB 28kB/s 
Requirement already up-to-date: six>=1.10.0 in /usr/lib/python3.6/site-packages (from tensorflow==1.0.0rc1)
Requirement already up-to-date: wheel>=0.26 in /usr/lib/python3.6/site-packages (from tensorflow==1.0.0rc1)
Requirement already up-to-date: protobuf>=3.1.0 in /usr/lib/python3.6/site-packages (from tensorflow==1.0.0rc1)
Requirement already up-to-date: numpy>=1.11.0 in /usr/lib64/python3.6/site-packages (from tensorflow==1.0.0rc1)
Requirement already up-to-date: setuptools in /usr/lib/python3.6/site-packages (from protobuf>=3.1.0->tensorflow==1.0.0rc1)
Requirement already up-to-date: packaging>=16.8 in /usr/lib/python3.6/site-packages (from setuptools->protobuf>=3.1.0->tensorflow==1.0.0rc1)
Requirement already up-to-date: appdirs>=1.4.0 in /usr/lib/python3.6/site-packages (from setuptools->protobuf>=3.1.0->tensorflow==1.0.0rc1)
Requirement already up-to-date: pyparsing in /usr/lib/python3.6/site-packages (from packaging>=16.8->setuptools->protobuf>=3.1.0->tensorflow==1.0.0rc1)
Installing collected packages: tensorflow
  Found existing installation: tensorflow 0.12.1
    Uninstalling tensorflow-0.12.1:
      Successfully uninstalled tensorflow-0.12.1
Successfully installed tensorflow-1.0.0rc1

So the correct URL for RHEL/Centos is: tensorflow-1.0.0rc1-cp36-cp36m-linux_x86_64.whl . Verify your upgrade by running:

(tensorflow) [tensorflow]# python3.6 -c 'import tensorflow as tf; print(tf.__version__)'
1.0.0-rc1

Now run the tutorial again, the TypeError is gone, but is replaced by an AttributeError :

(tensorflow) [tensorflow]# python3.6 ptb_word_lm.py --data_path=simple-examples/data/
Traceback (most recent call last):
  ...
  File "ptb_word_lm.py", line 149, in __init__
     output = tf.reshape(tf.concat_v2(outputs, 1), [-1, size])
AttributeError: module 'tensorflow' has no attribute 'concat_v2'

Turns out that TWO tensorflow APIs changed:

1. concat_v2 goes away and is replaced simply by: concat
2. Likewise, scalar_summary has been renamed to: summary.scalar . There are 3 such references in ptb_word_lm.py

Edit ptb_word_lm.py to make both of the changes above AND the tutorial should now run correctly for you!