search

tensorflow / tensorboard-plugin-example

Apache License 2.0
135 stars 59 forks source link

Distributing and using a plugin #41

Closed ted-kapustin closed 5 years ago

ted-kapustin commented 5 years ago

Do I understand it correctly that I have to build a custom TensorBoard just to make it possible to use my plugin? It sounds very unusual and surprising. Wouldn't it be better just to register a plugin somehow or pass its name/path as an argument to TensorBoard? Also, how can I start my custom TensorBoard without Bazel? The way the demo Greeter tensorboard is started is this:

bazel run //greeter_tensorboard -- --logdir=/tmp/greeter_demo

I am new to both Python and Tensorboard, so I could not figure out a different way to start it. If this is the only way to start a custom tensorboard, this means every user of my custom tensorboard would have to install Bazel. But I could not even get Bazel to build this Greeter sample on Windows because VC15 compiler (the one that comes with Visual Studio 2017) cannot compile C++ used in TensorBoard (I guess because VC does not fully support modern C++ standard). I did some googling, but I could not find answers to these questions. Could you write a paragraph explaining how exactly custom tensorboards should be distributed and started? Sorry if this is the wrong place to ask these questions.

ted-kapustin commented 5 years ago

I have figured out how to do it. Here is the answer:

https://stackoverflow.com/questions/53637674/deploying-a-tensorboard-plugin