I’d like to pre-train a model under several different conditions, then finetune each of those resulting models. The simple way to do this would be two separate sweeps. But then I need to manually start the second one. Is there a way to combine these into one single sweep?
With a bash script, I can simply call pre-train and finetune in sequence, passing the respective arguments via a config file. Is it somehow possible to tell the respective scripts that they are part of the same sweep and should therefore use certain parameters?
I don’t believe so. A sweep is inherently tied to its config, so there is no way to have 2 different configs - one for pre-training and one for fine-tuning.
The only good way to do this currently is to have 2 sweeps to do this.
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.
Hi Sebastian, 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!