Is there a way to support multithreading with sweeps for hyper parameter tuning. (like sklearn GridSearchCV ‘n_jobs’ parameter).
can you just launch multiple wandb.agents with the same sweep id (passed from outside) in different threads? the sweep controller, i guess, will orchestrate that they get different parameter sets
Hello @vikrant17 ,
We’ll look into this if W&B can support this one, let us check things on our end first and will get back to you.
Hello @vikrant17 ,
We would like to know your specific use case, what exactly you are trying to do and what is the expected result so we can understand it better.
Hi @vikrant17 ,
Good Day!
I would like to follow up on our inquiry for this reported concern please.
Thanks 
Hi @joana-marie , I wanted to do a random search for the hyperparameters quickly. I wanted to know if there is way to use all the cores of the system to do the hyperparameter sweeping (like the sklearn CV methods allow).
The only solution I have come across is to run the sweep via multiple processes using the same sweep id.
Any solution to this?
Hi @vikrant17 ,
You will need to use parallelized agents Parallelize agents and you can’t run multithreading to parse through the different hyperparameter combinations.
Hi @vikrant17 , 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!
Hi @joana-marie
I was able to run the sweep in parallel using the joblib package as multiple process. Although this could be done using the sweep id as mentioned by you, I wanted something that could automatically use separate cores automatically to perform each run for the sweep in parallel. That way I don’t have to open new terminals and manually run the agent.
Anyways, thank you.
1 Like
This topic was automatically closed 60 days after the last reply. New replies are no longer allowed.