Annotation of logs

[dreamgonfly]

I'd like to annotate on logs, especially what epoch a log belongs to.

Iteration-level (or each nth-iteration) logging makes it hard to know what epoch a specific data point is. If I do epoch-level logging, it leads to a low number of observations and I have to wait quite long for information.

An obvious solution would be to provide an additional parameter or add tuple support for global_step. Is there a reason why this is not possible yet?

This issue is related to https://github.com/lanpa/tensorboardX/issues/269 , but I don't think the questions was properly answered.

[lanpa]

To support this, you have to open a feature request at tensorflow/tensorboard. So that they can design how to visualize the epoch on the webpage and decide a protocol on how to log the information. A quick workaround would be logging epoch number along with the current global_step. In this way, you can know the epoch number by comparing the epoch graph with your original graph side by side.