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 All, I'M Using Clearml 1.0.3 With Clearml-Server <1 (How Do I Get The Current Running Version?) In Pytorch-Lightning I Use Ddp And I See Multiple Tasks (As The Number Of Gpus) Being Created And Remaining In Draft Mode. Is It A Problem Running Clearml

Hi all,
I'm using clearml 1.0.3 with clearml-server <1 (how do I get the current running version?)
In Pytorch-Lightning I use DDP and I see multiple tasks (as the number of gpus) being created and remaining in draft mode.

Is it a problem running clearml > 1 with server < 1?
If not, what could be the problem?

  
  
Posted 3 years ago
Votes Newest

Answers 18


server

  
  
Posted 3 years ago

Maybe we should rename it?! it actually creates a Task but will not auto connect it...

  
  
Posted 3 years ago

An easier fix for now will probably be some kind of warning to the user that a task is created but not connected

  
  
Posted 3 years ago

got it

  
  
Posted 3 years ago

pip install clearml==1.0.4rc1

  
  
Posted 3 years ago

Is this you case?

  
  
Posted 3 years ago

still the same 😞

  
  
Posted 3 years ago

ohh that's why 🙂

  
  
Posted 3 years ago

I can't find it in PyPi

  
  
Posted 3 years ago

oops. I used create instead of init 😳

  
  
Posted 3 years ago

Hi ExcitedFish86

In Pytorch-Lightning I use DDP

I think a fix for pytorch multi-node / process distribution was commited to 1.0.4rc1, could you verify it solves the issue ? (rc1 should fix this specific issue)
BTW: no problem working with cleaml-server < 1

  
  
Posted 3 years ago

sounds great.
BTW the code is working now out-of-the-box. Just 2 magic line - import + Task.init

  
  
Posted 3 years ago

ExcitedFish86 You came to ClearML because it's free, you stayed because of the magic 🎊 🎉

  
  
Posted 3 years ago

LOL 😊

  
  
Posted 3 years ago

ExcitedFish86 regarding the <1 version - are you talking about ClearML Server or ClearML SDK?

  
  
Posted 3 years ago

I think so. IMHO all API calls should maybe reside in a different module since they usually happen inside some control code

  
  
Posted 3 years ago

Task.init should be called before pytorch distribution is called, then on each instance you need to call Task.current_task() to get the instance (and make sure the logs are tracked).

  
  
Posted 3 years ago

An easier fix for now will probably be some kind of warning to the user that a task is created but not connected

That is a good point, maybe if you do not have a "main" Task, then we print the warning (with some flag to disable the warning) ?

  
  
Posted 3 years ago
1K Views
18 Answers
3 years ago
one year ago
Tags