0

I tried to install autokeras (pip install autokeras) on my windows10 computer but got the following errors:

tensorflow-gpu 1.9.0 has requirement tensorboard<1.10.0,>=1.9.0, but you'll have tensorboard 1.10.0 which is incompatible.
tensorflow 1.10.0 has requirement numpy<=1.14.5,>=1.13.3, but you'll have numpy 1.15.4 which is incompatible.
Installing collected packages: scikit-learn, unidecode, pandas, inflect, tqdm, imageio, tensorboard, tensorflow, opencv-python, lightgbm, nltk, lws, joblib, librosa, networkx, cloudpickle, toolz, dask, PyWavelets, scikit-image, chardet, idna, urllib3, requests, autokeras
Found existing installation: scikit-learn 0.18.1
Cannot uninstall 'scikit-learn'. It is a distutils installed project and thus we cannot accurately determine which files belong to it which would lead to only a partial uninstall.

My python is 3.6.5, numpy 1.12.0 & 1.15.4, tensorboard 1.9.0, tensorflow-gpu 1.9.0.
Is the gpu version of tensorflow a problem for autokeras?

I would appreciate it a lot if someone can help me to solver this problem.

Xinyu
  • 63
  • 11

1 Answers1

0

I had a similar problem when installing tensorflow-gpu on linux. Your tensorboard and numpy are too recent, meaning they are incompatible.

Your best bet is to set up a virtual environment (use virtualenv), install tensorboard 1.9.0 and numpy 1.13.3 and specify those for the virtual environment. Run the virtual environment and install tf-gpu in there and it should work.

ZWang
  • 832
  • 5
  • 14
  • Thank you. I'll try it on monday. Actually I didn't have numpy of version 1.15.4. But after trying to install autokeras, it is installed automatically! Why? – Xinyu Jan 19 '19 at 07:29
  • Not sure on windows but maybe it tries to install the dependencies and NumPy 1.15.4 was the most recent one? And then it finds that it’s incompatible. – ZWang Jan 19 '19 at 09:12
  • I tried it today. The error about numpy is still there, but autokeras is successfully installed! – Xinyu Jan 19 '19 at 10:35