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
Hello, I Have A Question Regarding Creating A Clearml Pipeline Using Pytorch Lightning. I Am Not Really Sure Where To Begin. Should I Create A Task For Each Pytorch Lightning Class In My Pipeline? Is There A Demo Or Clearml Project That Specifically Uses

Hello,

I have a question regarding creating a clearml pipeline using pytorch lightning. I am not really sure where to begin. Should I create a task for each Pytorch lightning class in my pipeline? Is there a demo or clearml project that specifically uses pytorch ligthning in it's pipeline? Anything would help :D

  
  
Posted one year ago
Votes Newest

Answers 5


Hi @<1547028031053238272:profile|MassiveGoldfish6>
What is the use case? the gist is you want each component to be running on a different machine. and you want to have clearml do the routing of data and logic between.
How would that work in your use case?

  
  
Posted one year ago

I would like to implement MLOPS best practices to my project.
So in my Datamodule class, i would load the clearml data and prep it into train and test. In the lightning module class, i would create my model, and finally use trainer class to train.
How do I best utilize clearml in this scenario such that any coworker of mine is able to reproduce my work with the same pipeline?

  
  
Posted one year ago

in other words, how do you combine a pytorchlightning Module with a ClearML task?

  
  
Posted one year ago

How do I best utilize clearml in this scenario such that any coworker of mine is able to reproduce my work with the same pipeline?

Basically this sounds to me like proper software developemnt design (i.e. the class vs stages).
In order to make sure Anyone can reproduce it, you mean anyone can rerun the "pipeline" ? If this is the case just add Task.init (maybe use a specific Task type) and the agents will make sure this is Fully reproducible.
If you mean the data itself is stored, then you have to store the Datamodule as dataset, and maybe add an argument to your code weather to pull the latest datd from the datasource (i.e. DB?) or use a stored dataset, and in that case pass the dataset UID,
wdyt ?

a pytorchlightning Module with a ClearML task

No need to "specially" combine it. The moment you store the Module in pytorch lighting it is stored in the ClearML model repository, with a pointer to the generating Task (see above, by definition fully repdocubible)

Am I missing something ?

  
  
Posted one year ago

No, this is very useful, thank you

  
  
Posted one year ago
1K Views
5 Answers
one year ago
one year ago
Tags
Similar posts