I have a project with a good number of runs (~30k). My overall goal is to find the best performing runs across a series of parameters specifying the data configuration (these parameters in my configs). Performance is measured as the maximum of the test accuracy obtained during training. This is relatively burdensome to do computationally if I read out each run individually. However it seems like something that any SQL-like database should be able to handle with ease (group by some parameters, measure maximums and sort, show only the top ones). Is this feasible with the wandb API?
Hey @evanv,
You should be able to sort runs from the API using the order parameter for the api.Runs method. Here are some docs for the api.Runs method.
Please let me know if this resolves your issue.
Thanks,
Ramit
Hi Evan,
We wanted to follow up with you regarding your support request as we have not heard back from you. Please let us know if we can be of further assistance or if your issue has been resolved.
Best,
Weights & Biases
Hi Evan, since we have not heard back from you we are going to close this request. If you would like to re-open the conversation, please let us know!
This topic was automatically closed 60 days after the last reply. New replies are no longer allowed.