Best way to re-run a run

What is the best/most efficient way to re-run a run from a sweep. I have tied downloading the config YAML from the files section of my run, removing the ‘wand’ key and renaming the YAML to config-defaults.yaml. This works but I wonder is there a more streamlined way to do this?

Hi @alison-peard, are you ok with removing the run? If so, you can just delete the run from the UI and then assign an agent to the sweep again. It will pick up any parameter combinations that haven’t been tried yet automatically.

Let me know if that helps!

Hi @alison-peard, did this workout as a solution?

Hi Nathan,

No I actually wanted to re-run the run exactly, as in it converged well but I forgot to save the weights and now all the rest of the models I am training are not converging so well. I would have liked to recover it

Hi @alison-peard, what other factors besides the hyperparameters from the run are you looking to replicate? Is there anything else in config…yaml that you are looking for? If so, I think your solution might be the best option.

Hi @alison-peard, I wanted to follow up and see if I could still help with this?

This topic was automatically closed 60 days after the last reply. New replies are no longer allowed.