I run pytorch model with sweep in slurm machines in China.As the great wall blocks the google api ,I have to use sweep in offline mode. The weird thing is that when I set epoch in small numbers like 10,it works fine.But if set epoch bigger than 30 ,the sweep agent keeps return the same hyperparameters and run id .
It’s not fine in small epoch setting too.It reture run id as a loop. I don’t know where the problem is …
Hi @weicy! Apologies you are running into this! Could you please provide an example code you are running as well as a workspace where you are syncing up your runs to?
Hi there, I wanted to follow up on this request. Please let us know if we can be of further assistance or if your issue has been resolved.
Hi, since we have not heard back from you, we are going to close this request. If you would like to reopen the conversation, please let us know! Unfortunately, at the moment, we do not receive notifications if a thread reopens on Discourse. So, please feel free to create a new ticket regarding your concern if you’d like to continue the conversation.