Native support for TensorFlow on Windows was one of the first requests we received after open-sourcing TensorFlow. Although some Windows users have managed to run TensorFlow in a Docker container, we wanted to provide a more complete experience including GPU support.
With the release of TensorFlow r0.12, we now provide a native TensorFlow package for Windows 7, 10, and Server 2016. This release enables you to speed up your TensorFlow training with any GPU that runs CUDA 8.
We have published the latest release as a pip package in PyPI, so now you can install TensorFlow with a single command:
C:\> pip install tensorflowAnd for GPU support:
C:\> pip install tensorflow-gpu
More details about Windows support and all of the other new features in r0.12 are included in the release notes.
We're excited to offer more people the opportunity to use TF at maximum speed. Follow us on Twitter to be the first to hear about future releases – we're @tensorflow.
Acknowledgements
Many people have contributed to making this release possible. In particular, we'd like to thank Guenther Schmuelling and Vit Stepanovs from Microsoft for their significant contributions to Windows support.