Train again with the same settings
Load the 11-step settings of a previous training run as they were and run it once more
Required permission: DevelopDEVELOPER · OPERATOR · ADMIN
Use this when you have uploaded more data, or want to try changing just one hyperparameter.
- Open a previous training run from the Training (학습) list and press Train again with these settings (이 설정으로 다시 학습) ①.
Training run detail — ① Train again with these settings - With a "Settings loaded (설정을 불러왔습니다)" notification, you land directly on 11. Review and start (검토·시작) with steps 1–10 all filled in. If there is a step to change, click it in the list on the left and edit it, then press Start training (학습 시작).
The settings are loaded and you land on Review and start — only the experiment name is empty
What carries over and what does not
| Item | Behavior |
|---|---|
| Task, framework, model, dataset, classes, split, hyperparameters, evaluation, Export, GPU, registry registration | Carried over as they were |
| Experiment name | Cleared. To pile runs into the same experiment, enter the previous name again |
| Training runtime variant | Kept if it can still be chosen; if it is gone, it switches to the default variant and the notification says so |
| Starting weights (ones you uploaded) | Kept if still Ready (사용 가능); if deleted, it switches to the catalog weights and the notification says so |
| Dataset contents | The dataset is referenced by ID, so training uses its current contents, including files added in the meantime |