tensorflow: Can't access gs:// logfiles using tensorboard

I am also facing the same issue:

System information:

Have I written custom code (as opposed to using a stock example script provided in TensorFlow): No

OS Platform and Distribution (e.g., Linux Ubuntu 16.04): Windows 10

TensorFlow installed from (source or binary): pip3 install tensorflow

TensorFlow version (use command below): 1.3.0

Python version: 3.5.2

Exact command to reproduce: tensorboard --logdir=gs://mybucket

Facing the same issue, on MacOS

This can happen if you build TensorFlow from source and answer “No” to the question “Do you wish to build TensorFlow with Google Cloud Platform support?”. However you indicated that you installed a pre-built TensorFlow via pip install which should presumably have this option enabled. Adding @jhseu to help.

Unassigning because TensorBoard probably does not need this.

Running TensorBoard on a Windows PC to ingest gigabytes of event logs from Google Cloud Storage is almost certainly a mistake. It can result in bandwidth charges. Both from your ISP and GCP. If you have a Windows PC and use GCS, please run TensorBoard on GCE in the same zone. For example:

# Create VM
gcloud config set compute/zone us-central1-a
gcloud compute instances create tensorboard
gcloud compute ssh tensorboard -- -T sudo /bin/sh <<EOF
apt-get update
apt-get install python-setuptools
easy_install -U pip
pip install -U tensorflow
EOF

# Run TensorBoard remotely on VM and ask SSH to port tunnel (don't edit GCP firewall).
gcloud compute ssh tensorboard -- -L 6006:localhost:6006 tensorboard --logdir=gs://bucket/experiment

# Show TensorBoard
google-chrome http://localhost:6006/

Windows has GCS disabled by default. Is this something that should be changed?

https://github.com/tensorflow/tensorflow/blob/master/configure.py#L1466

Did #14856 fully add in GCS support in Windows?

Both 1.5.0rc1 and nightly (1.6.0.dev20180117) still complain that gs scheme is not implemented.

/CC @jart for package collisions.

It seems that the problem comes from installing the package called tensorboard. If I just go pip install tensorflow, it installs tensorflow and tensorflow-tensorboard which works fine.

So: Avoiding the package tensorboard and only installing tensorflow solved this for me.