在 Anaconda 中安装 tensorflow-gpu 时出现无法满足的错误

Question

I have installed Anaconda Python 3.7 in Ubuntu 18.04 and then executed the commands:

conda update --all
conda install cudnn

Now when I try to install tensorflow-gpu using the command conda install tensorflow-gpu , I get Unsatisfiable error like this:

UnsatisfiableError: The following specifications were found to be incompatible with each other:

• pkgs/main/linux-64::_ipyw_jlab_nb_ext_conf==0.1.0=py37_0 -> ipywidgets -> widgetsnbextension[version='>=3.4.0,<3.5.0'] -> notebook[version='>=4.4.1'] -> nbconvert -> bleach
• pkgs/main/linux-64::bleach==3.1.0=py37_0
• pkgs/main/linux-64::ipywidgets==7.4.2=py37_0 -> widgetsnbextension[version='>=3.4.0,<3.5.0'] -> notebook[version='>=4.4.1'] -> nbconvert -> bleach
• pkgs/main/linux-64::jupyterlab==0.35.5=py37hf63ae98_0 -> jupyterlab_server[version='>=0.2.0,<0.3.0'] -> notebook -> nbconvert -> bleach
• pkgs/main/linux-64::jupyterlab_server==0.2.0=py37_0 -> notebook -> nbconvert -> bleach
• pkgs/main/linux-64::notebook==5.7.8=py37_0 -> nbconvert -> bleach
• pkgs/main/linux-64::widgetsnbextension==3.4.2=py37_0 -> notebook[version='>=4.4.1'] -> nbconvert -> bleach
• pkgs/main/noarch::nbconvert==5.5.0=py_0 -> bleach

As multiple important packages are concerned here, I am getting confused what to do. I have NVIDIA GTX 1070 Max-Q in my PC, so tensorflow-gpu should work perfectly.

Answer 1

You can install Tensorflow GPU using the below code:

1. Create a New Virtual Environment

   conda create -n tensorflow_gpu pip python=3.6

2. Activate the Virtual Environment

   activate tensorflow_gpu

3. Install CUDA Toolkit using

   conda install -c anaconda cudatoolkit

4. Install Tensorflow GPU

   pip install --ignore-installed --upgrade tensorflow-gpu==1.15

For more information, please refer this Tensorflow Blog .