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
Answered
Hi There, Is There A Maximum Number Of Plots (I.E. Losses/Scalars Being Logged During Training)? I Have An Application With Many Different Losses (About 200) And I Can See That Some Are Missing From The Dashboard. When This Number Is Smaller Everything Ru

Hi there, is there a maximum number of plots (i.e. losses/scalars being logged during training)? I have an application with many different losses (about 200) and I can see that some are missing from the dashboard. When this number is smaller everything runs smoothly

  
  
Posted one year ago
Votes Newest

Answers 15


I am using v.1.3.2

The SDK I assume.

Is this a self hosted version? What is the server version?

  
  
Posted one year ago

When the number of logged losses/metrics is around 80, everything works smoothly

  
  
Posted one year ago

I recall a big fix to plots in server version 1.6.0, can you try upgrading to see if it fixes the issue?

  
  
Posted one year ago

Thanks will try that!

  
  
Posted one year ago

Hello CostlyOstrich36 ! I tried to use the max_samples argument but it didn’t help since all scalars are returned in its default value (0).

  
  
Posted one year ago

StormyLeopard75 has helped me reproduce this unexpected behaviour

  
  
Posted one year ago

CostlyOstrich36 TimelyPenguin76 any updates in here?

  
  
Posted one year ago

Thanks for the swift reply btw 🙂

  
  
Posted one year ago

Just a quick update on this. I have tried simulating my issue using the API and it seems that get_reported_scalars has an upper-limit of 100, whereas get_last_scalar_metrics returns all the logged metrics (200 in the example below)

  
  
Posted one year ago

I'm not sure. Maybe TimelyPenguin76 can elaborate.
Looking at the docs for:
https://clear.ml/docs/latest/docs/references/sdk/task#get_reported_scalars

I can see it has a max_samples argument

  
  
Posted one year ago

Yes, self-hosted version
WebApp: 1.3.0-165 • Server: 1.3.0-165 • API: 2.17

  
  
Posted one year ago

Could this be an issue/bug CostlyOstrich36 ?

  
  
Posted one year ago

Hi IdealGorilla64 , can you please provide a screenshot? How much smaller than 200 have you tried. And finally which version of ClearML are you using? 🙂

  
  
Posted one year ago

TimelyPenguin76 can you assist?

  
  
Posted one year ago

Hi there, I am using v.1.3.2

Here is a screenshot, it seems that the metrics are not logged as they don’t appear in the “Toggle” window

  
  
Posted one year ago
717 Views
15 Answers
one year ago
one year ago
Tags