Error while importing Tensorflow

Question

I was using this code:

import tensorflow as tf , and it was working fine.

When I now try to import it, I get the following error

In [2]: import tensorflow as tf
---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)
<ipython-input-2-41389fad42b5> in <module>()
----> 1 import tensorflow as tf

/usr/local/lib/python2.7/dist-packages/tensorflow/__init__.py in <module>()
     21 from __future__ import print_function
     22 
---> 23 from tensorflow.python import *

/usr/local/lib/python2.7/dist-packages/tensorflow/python/__init__.py in <module>()
     63 from tensorflow.core.util.event_pb2 import *
     64 # Import things out of contrib
---> 65 import tensorflow.contrib as contrib
     66 
     67 # Framework

/usr/local/lib/python2.7/dist-packages/tensorflow/contrib/__init__.py in <module>()
     28 from tensorflow.contrib import grid_rnn
     29 from tensorflow.contrib import layers
---> 30 from tensorflow.contrib import learn
     31 from tensorflow.contrib import linear_optimizer
     32 from tensorflow.contrib import lookup

/usr/local/lib/python2.7/dist-packages/tensorflow/contrib/learn/__init__.py in <module>()
     70 
     71 # pylint: disable=wildcard-import
---> 72 from tensorflow.contrib.learn.python.learn import *
     73 from tensorflow.python.util.all_util import make_all
     74 

/usr/local/lib/python2.7/dist-packages/tensorflow/contrib/learn/python/__init__.py in <module>()
     21 
     22 # pylint: disable=wildcard-import
---> 23 from tensorflow.contrib.learn.python.learn import *

/usr/local/lib/python2.7/dist-packages/tensorflow/contrib/learn/python/learn/__init__.py in <module>()
     24 # pylint: disable=wildcard-import
     25 from tensorflow.contrib.learn.python.learn import datasets
---> 26 from tensorflow.contrib.learn.python.learn import estimators
     27 from tensorflow.contrib.learn.python.learn import graph_actions
     28 from tensorflow.contrib.learn.python.learn import learn_io as io

/usr/local/lib/python2.7/dist-packages/tensorflow/contrib/learn/python/learn/estimators/__init__.py in <module>()
     21 
     22 from tensorflow.contrib.learn.python.learn.estimators._sklearn import NotFittedError
---> 23 from tensorflow.contrib.learn.python.learn.estimators.autoencoder import TensorFlowDNNAutoencoder
     24 from tensorflow.contrib.learn.python.learn.estimators.base import TensorFlowBaseTransformer
     25 from tensorflow.contrib.learn.python.learn.estimators.base import TensorFlowEstimator

/usr/local/lib/python2.7/dist-packages/tensorflow/contrib/learn/python/learn/estimators/autoencoder.py in <module>()
     23 
     24 from tensorflow.contrib.learn.python.learn import models
---> 25 from tensorflow.contrib.learn.python.learn.estimators.base import TensorFlowBaseTransformer
     26 from tensorflow.python.ops import nn
     27 

/usr/local/lib/python2.7/dist-packages/tensorflow/contrib/learn/python/learn/estimators/base.py in <module>()
     32 from tensorflow.contrib import layers
     33 from tensorflow.contrib.learn.python.learn.estimators import _sklearn
---> 34 from tensorflow.contrib.learn.python.learn.estimators import estimator
     35 from tensorflow.contrib.learn.python.learn.estimators._sklearn import NotFittedError
     36 from tensorflow.contrib.learn.python.learn.learn_io.data_feeder import setup_train_data_feeder

/usr/local/lib/python2.7/dist-packages/tensorflow/contrib/learn/python/learn/estimators/estimator.py in <module>()
     37 from tensorflow.contrib.learn.python.learn.estimators import tensor_signature
     38 from tensorflow.contrib.learn.python.learn.estimators._sklearn import NotFittedError
---> 39 from tensorflow.contrib.learn.python.learn.learn_io import data_feeder
     40 from tensorflow.contrib.learn.python.learn.utils import checkpoints
     41 

/usr/local/lib/python2.7/dist-packages/tensorflow/contrib/learn/python/learn/learn_io/__init__.py in <module>()
     28 from tensorflow.contrib.learn.python.learn.learn_io.graph_io import read_keyed_batch_examples
     29 from tensorflow.contrib.learn.python.learn.learn_io.graph_io import read_keyed_batch_features
---> 30 from tensorflow.contrib.learn.python.learn.learn_io.pandas_io import extract_pandas_data
     31 from tensorflow.contrib.learn.python.learn.learn_io.pandas_io import extract_pandas_labels
     32 from tensorflow.contrib.learn.python.learn.learn_io.pandas_io import extract_pandas_matrix

/usr/local/lib/python2.7/dist-packages/tensorflow/contrib/learn/python/learn/learn_io/pandas_io.py in <module>()
     22 try:
     23   # pylint: disable=g-import-not-at-top
---> 24   import pandas as pd
     25   HAS_PANDAS = True
     26 except ImportError:

/usr/local/lib/python2.7/dist-packages/pandas/__init__.py in <module>()
     23 
     24 try:
---> 25     from pandas import hashtable, tslib, lib
     26 except ImportError as e:  # pragma: no cover
     27     module = str(e).lstrip('cannot import name ')  # hack but overkill to use re

/home/kv/pandas/src/numpy.pxd in init pandas.hashtable (pandas/hashtable.c:38364)()

ValueError: numpy.dtype has the wrong size, try recompiling. Expected 88, got 96

I even tried upgrading numpy, but that didn't solve the issue. Can you help me resolve this error?

EDIT: numpy (1.8.2) tensorflow (0.10.0rc0) python 2.7.6

Answer 1

Try installing numpy with PIP using

sudo apt-get install python-pip
sudo pip install numpy==1.11.1

or pip3 instead of pip for python 3 like

sudo apt-get install python3-pip
sudo pip3 install numpy==1.11.1

this will help as i also had this error because tensorflow is using numpy 1.11.1

Answer 2

I would strongly suggest to use the anaconda distribution and install the latest tensorflow using:

conda install -c conda-forge tensorflow

Read more in the official setup guide

Answer 3

You need to update numpy or recompile pandas.

There is a very good answer and explanation provided here: ValueError: numpy.dtype has the wrong size, try recompiling