Tensorflow Opkernel Error

Question

I'm attempting to run a basic tensor flow tutorial. I followed the instructions here for downloading and installing Tensorflow https://software.intel.com/en-us/articles/build-and-install-tensorflow-on-intel-architecture. This is the error I'm getting. Any help would be greatly appreciated!

NotFoundError (see above for traceback): No registered '_MklConv2DWithBias' OpKernel for CPU devices compatible with node conv2d/BiasAdd = _MklConv2DWithBias[T=DT_HALF, _kernel="MklOp", data_format="NHWC", dilations=[1, 1, 1, 1], padding="SAME", strides=[1, 1, 1, 1], use_cudnn_on_gpu=true, _device="/job:localhost/replica:0/task:0/device:CPU:0"](Reshape, conv2d/kernel/read, conv2d/bias/read, DMT/_0, DMT/_1, DMT/_2) (OpKernel was found, but attributes didn't match) . Registered: device='CPU'; label='MklOp'; T in [DT_FLOAT]

[[Node: conv2d/BiasAdd = _MklConv2DWithBias[T=DT_HALF, _kernel="MklOp", data_format="NHWC", dilations=[1, 1, 1, 1], padding="SAME", strides=[1, 1, 1, 1], use_cudnn_on_gpu=true, _device="/job:localhost/replica:0/task:0/device:CPU:0"](Reshape, conv2d/kernel/read, conv2d/bias/read, DMT/_0, DMT/_1, DMT/_2)]]

Answer 1

Tensorflow can be easily installed using the whl file if you have Anaconda.

These are the steps:

1. Create a conda Environment

   conda create -n (env name) -c intel python=3.6 pip numpy

2. Activate the environment

   For linux : source activate (env name)

   For Windows: .activate (env name)

3. Install the whl file.

   pip install https://anaconda.org/intel/tensorflow/1.6.0/download/tensorflow-1.6.0-cp36-cp36m-linux_x86_64.whl

For further details, refer the link below:

https://software.intel.com/en-us/articles/intel-optimized-tensorflow-installation-guide?page=1

Answer 2

This error means that your cpu doesn't support the opkernel which is required by tensorflow-mkl version to accelerate the computation.
Usually this case happens when you install tensorflow of mkl version. Actually tensorflow has three version: tensorflow-gpu, tensorflow-mkl(cpu)， tensorlfow-eigen. Conda or pip will install tensorlfow-mkl by default if you don't specify the version on which circumtances tensorflow-mkl requires that your cpu supoort mkl optimization. if your cpu doesn't have mkl support, it will induce such error.
SOLUTION Uninstall your tensorflow and reinstall tensorflow-eigen as flowed command.

conda uninstall tensorflow
conda install "tensorflow=version_number=eigen*"