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 Everyone,

Hi Everyone,

I am trying to make the Hyperparameter tuning part work. When I change the hyper parameters from the allegro web app, I enqueue it and run the train agent. The new cloned script goes in running state and then I get an error โ€œgit diff failedโ€

  
  
Posted 4 years ago
Votes Newest

Answers 30


the agent uses the same configuration file

  
  
Posted 4 years ago

Anyways from what I see in the logs it shows agent.default_python = 3.7, cuda = 100, cudnn=75

  
  
Posted 4 years ago

It fails only when I make any changes

  
  
Posted 4 years ago

It worked finally ๐Ÿ™‚

  
  
Posted 4 years ago

No

  
  
Posted 4 years ago

It would be great if these issues are well elaborated in the documentation. Though the documentation is pretty good.

  
  
Posted 4 years ago

Yes. You can see the agent's configuration in the experiment's log - all values are printed there

  
  
Posted 4 years ago

Did you make sure the agent's default python version and cuda / cudnn are configured correctly?

  
  
Posted 4 years ago

I only see trains conf file

  
  
Posted 4 years ago

In the conf file?

  
  
Posted 4 years ago

And how do I change agent configuration file?

  
  
Posted 4 years ago

But then how the normal one I.e without cloning worked well

  
  
Posted 4 years ago

However all packages are cached so it won't download again

  
  
Posted 4 years ago

Yes very much

  
  
Posted 4 years ago

Yes ๐Ÿ˜‚

  
  
Posted 4 years ago

Are you running the agent on the same machine?

  
  
Posted 4 years ago

Wherever running an experiment, it will install all required packages in a virtual environment to make sure the experiment is executed exactly as expected

  
  
Posted 4 years ago

Good ๐Ÿ™‚ I see you still have issues with your CUDA installation

  
  
Posted 4 years ago

I ran it outside the examples folder and it works

  
  
Posted 4 years ago

I can see you failed experiments in the demo server, but I can't see any completed experiment from which they were cloned...

  
  
Posted 4 years ago

However will it always install all the packages again and again? Is there any workaround for that?

  
  
Posted 4 years ago

I just checked there are some uncommitted changes that I see in execution

  
  
Posted 4 years ago

I need to showcase this to my senior tomorrow

  
  
Posted 4 years ago

You can first try to run your experiment again (not by cloning and running in the agent but by executing it again locally). If you like, you can copy the example and run it from another folder which is not located inside a git repo

  
  
Posted 4 years ago

So now what do I do?

  
  
Posted 4 years ago

Is there a difference in the uncommitted changes section before your changes and after?

  
  
Posted 4 years ago

I know, they represent the changes you made to the example script

  
  
Posted 4 years ago

and are these the correct values?

  
  
Posted 4 years ago

I did that with the agent without making changes it works

  
  
Posted 4 years ago
767 Views
30 Answers
4 years ago
one year ago
Tags