search

GoogleCloudPlatform / tensorflow-without-a-phd

A crash course in six episodes for software developers who want to become machine learning practitioners.
Apache License 2.0
2.79k stars 911 forks source link

Not able to monitor the progress of training, using TensorBoard on GCP #13

Closed ClaudeCoulombe closed 6 years ago

ClaudeCoulombe commented 6 years ago

I'm trying to monitor the progress of training, using TensorBoard on GCP... (It's my first time with GCP, but I've used AppEngine long time ago...)

1) I go into the ML Engine Jobs view and activate a Google Cloud Shell 2) In the Cloud Shell I type the command: tensorboard --logdir gs://exercices-tensorflow/pong_20180708_033245 --port 8080 3) I open a Web Preview (preview on port 8080)

I've got a TensorBoard page but it's inactive... with the message «No dashboards are active for the current data set.» See below screenshot...

capture d ecran 2018-07-08 a 15 48 57

The link with the training data seems not to be done ???? What is the events file? And how to set it?

ClaudeCoulombe commented 6 years ago

OK, again I've found the solution, since in the TensorBoard command I've put the job path not the data bucket path. When I put the data bucket path it works of course! How silly of me! ;(

Summaries: capture d ecran 2018-07-08 a 16 24 11

Images (hidden layer weights) capture d ecran 2018-07-08 a 16 19 51

The compute network: capture d ecran 2018-07-08 a 16 27 17

Histograms: capture d ecran 2018-07-08 a 16 29 23

Etc.

Keep it if it can help someone else...