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
Hey Guys!

Hey guys! 🙂 I just found out about Trains - what a great piece of work! I have on-premise system I want to use as an agent and server at the same time. I have set up both, web UI is visible, worker is started and visible in the web UI. I am curious how can I create an experiment? Documentation only talks about cloning, and as far as I can see the only way to create a new experiment is to have at least one manual run of your script with Train code snippets in it, indicating which previously defined experiment this run belongs to. Am I right?

Posted 3 years ago
Votes Newest

Answers 4

HI QuizzicalDove0
I guess the reason is that the idea is integration is literally 2 lines, and it will take less time to execute the code on a system with working env (we assume there is one) then to configure all the git , python packages, arguments etc...
All that said you can create an experiment from code , using Task.import_task https://allegro.ai/docs/task.html#trains.task.Task.import_task

Posted 3 years ago

You can always clone a “template” task and change everything (it will be on draft mode), what is you use case? maybe we already have a solution for it

Posted 3 years ago

Hi QuizzicalDove0 ,

Yes, you need to create the definition of the experiment, but you can add to your code the following line
task.execute_remotely(queue_name=<queue name>)This will stop running you code and enqueue the task to a specific queue (which the agent will run).

Posted 3 years ago

Great, thanks for confirming! Is there any reason why it's not possible to define the experiment in the UI directly?

Posted 3 years ago
4 Answers
3 years ago
one year ago
Similar posts