When you have experiments, you can access the reported metrics per experiment - that should allow you to decide which experiment is the best, and than you can take that experiment's model, WDYT?
Hi RotundPanda18 , how would you choose which is the best model?
maybe select by reported scalar value.
Here's my scenario.
- Proceed the experiment on a scheduled/manual basis.
- The results of the experiment (metric like accuracy) and model are accumulated in ClearML.
- Inferring the results of the experiment using the best (e.g., highest accuracy or lowest loss) model.
I would like to bring the most accurate model of ClearML stacked models using Python sdk for inferring.