Running Tensorboard on Windows

In my previous post < link >, I had shown how to install TensorFlow on Windows. However, I found that TensorBoard currently has issues running on Windows. For example, https://github.com/tensorflow/tensorflow/issues/5983 and https://github.com/tensorflow/tensorflow/issues/6005. There are different ways to solve this issue for example using Ubuntu virtual machine or bash on Windows available with anniversary update. The simplest approach I found is to use docker installation.
  1. For docker installation you can navigate to https://github.com/docker/toolbox/releases and download the latest stable release and install. Once installed open Docker Quickstart Terminal.
  2. Enter command docker run --privileged=true -dit -v /c/Users/USERNAME/tensorflow/tensorboard:/media/disk -p 8888:8888 -p 6006:6006 b.gcr.io/tensorflow/tensorflow 
  3. Note the docker ID and run the following command docker exec -it DOCKERID bash
  4. Enter tensorboard --logdir=/media/disk/tflogs assuming your tensorflow logs are in /c/Users/USERNAME/tensorflow/tensorboard/tflogs folder.
  5. Go to browser and enter address for docker machine followed by port 6006, e.g., in the video below it is http://192.168.99.100:6006/

*UPDATE: r0.12rc1 fixed the issue with tensorboard with default compile* Watch https://www.youtube.com/watch?v=cF7tIo6Njo4 to install or follow instructions on http://shan3ahmed.blogspot.co.uk/2016/12/how-to-install-tensorflow-on-windows10.html





No comments:

Post a Comment

How to run R code in Python

R is freely available language and computational tool which is very popular among the statisticians and bioinformatics community. In this p...