
Reputation
Badges 1
20 × Eureka!What optimization method did you use?
AFAIK, it's the default one - "Bandit-based Bayesian ..."
Also, what are the values that the cloned experiments get?
As far as I see, there are a few experiments with the same value, for instance 0.3.
I'll try to add a log, and to play with number's format
Can you add a log of the HPO app? It's in hidden projects (Go to settings and enable viewing hidden projects).
Here is the log of 'hpo-app' app. And it's really seems to be related to float type...
CostlyOstrich36 - thank you!
I guess the right word is 'results' - it's list of KPI metrics, that we use for compare our experiments.
Hmm, I don't know. I think, I saw this feature in 'manual' HPO code, but probably I'm wrong.
There is no code snippet, I created HPO app via UI.
Thanks, I'll try this workaround.
In this case though, there are only 11 possible parameter "combinations". But by default, ClearML sets the maximum amount of jobs much higher than that (check advanced settings in the wizard).
I could understand if it will use the minimum amount of jobs, not maximum 🙂
BTW, I created another HPO app with two parameters. And instead of 11*6=66 jobs I saw 92.
I'll open a bug.
@<1523701118159294464:profile|ExasperatedCrab78> - many thanks!
Thank you again! I read about task.connect in https://clear.ml/docs/latest/docs/clearml_sdk/task_sdk and now I see how I can add that.
Many thanks - using Logger it seems much better!
Now I see my metrics under 'Plots'. But when I compare two experiments, ClearML just shows me two tables. I hoped to see the diff.
Thank you again - found it. I haven't used them until now, I'll learn about them.
Thank you @<1523701070390366208:profile|CostlyOstrich36> , but I don't see a report into HPO app experiment, just a few graphs and a Summary table.
@<1523701070390366208:profile|CostlyOstrich36>
Sorry for delay, please see below:
CostlyOstrich36 - thank you, got it!
AgitatedDove14 thank you!
Do you mean in my train.py, after Task.init(..)?