Progress bar for training about clearml

[bmartinn]

@israelwei thanks for the suggestion, I guess the problem is how will trains know what the target epochs/batch/steps are?

That said, I can offer here a quick hack:

Assuming you are training for args.epoch epochs and each one is of args.epoch_size mini-batches.
Now, consider a for-loop with i as the increasing counter of batches (iterations/steps), we could do:

from trains import Logger

Logger.current_logger().report_scalar(title="progress", series="progress", iteration=i, value=round(100.0*i/(args.epoch_size*args.epoch), 1))

Now we have a graph of our progress (i.e. a number between 0 to 100, that represents our progress).

In order to view it in the experiment table we can add a custom column based on the metric we just added 'progress.progress'

Details and screenshots on custom columns can be found here