Examples: query, "exact match", wildcard*, wild?ard, wild*rd
Fuzzy search: cake~ (finds cakes, bake)
Term boost: "red velvet"^4, chocolate^2
Field grouping: tags:(+work -"fun-stuff")
Escaping: Escape characters +-&|!(){}[]^"~*?:\ with \, e.g. \+
Range search: properties.timestamp:[1587729413488 TO *] (inclusive), properties.title:{A TO Z}(excluding A and Z)
Combinations: chocolate AND vanilla, chocolate OR vanilla, (chocolate OR vanilla) NOT "vanilla pudding"
Field search: properties.title:"The Title" AND text
Hi, We Use Clearml To Track All Our Experiments. For Each Experiment The Accuracy The Logged For Both The Training And The Test Set:

Hi, we use clearml to track all our experiments. For each experiment the accuracy the logged for both the training and the test set:

self.logger.report_scalar('accuracy', 'train', acc_train, iteration=i) self.logger.report_scalar('accuracy', 'val', acc_val, iteration=i)
However, all the images also have labels like 'day' or 'night'. Now I would also like to see accuracy for the part of the train set with label 'day' or the accuracy for the part of the validation set with label 'night' for example.

possibilities are:
creating extra metrics: 'accuracy-day' and 'accuracy-night', so then there are 3 metrics 'accuracy', 'accuracy-day', 'accuracy-night' creating extra variants: 'train-day', 'train-night', 'val-train', 'val-night'
A third options, which I think would be very nice:
having multiple levels of hierarchy possible: accuracy > day > train or accuracy > night > val
As far as I know this is currently not possible, is this worth considering as a feature for the near future?

Posted 2 years ago
Votes Newest

Answers 4

Hi GreasyPenguin14
Quick question, any reason not to use a 2D scatter ? or a histogram (or any other non time-series plot)?

Posted 2 years ago

Are you also adding those metrics to the experiment table as extra columns ?

Posted 2 years ago

GreasyPenguin14 makes total sense.
In that case I would say variants to the accuracy make sense to me, I would suggest:
title='trains', series='accuracy/day' and title='trains', series='accuracy/night'
Regrading hierarchy, from the implementation perspective a unique identifier is always the combination of title/series (or in other words metric/variant), introducing another level is a system wide change.
This means it might be more challenging than expected ...

Posted 2 years ago

Yes, I add these metrics as extra columns and then I sort them. I want to know which experiments performs best in daylight for example or which during night. Therefore I think a is not the right choice

Posted 2 years ago