If I learn from the log file that my model is doing really well, can I somehow retrieve the model state from my check point **without** waiting for the training to complete (or manually terminate)? Thanks!

Created by Li Shen thefaculty
That would be great! Even if we only were able to get snapshots every 24 hours.
Hi Li, This feature is currently not available but we are thinking about implementing it. It will maybe become available at some point during Round 3. Thanks!
Dear Li, Thanks in advance for your patience. We are looking into this request. Best, Thomas

Can I get the model state during the middle of a training? page is loading…