-
Notifications
You must be signed in to change notification settings - Fork 79
Open
Description
Based on the code in pretrain.py and the configuration in pretrain_3g.yaml, the pretraining process generates 10 checkpoints (one after each epoch).
I'm wondering which checkpoint is the released ultra_3g.pth. Since there are multiple checkpoints generated during training.
Additionally, could you please clarify if there was a specific selection strategy(e.g., best performance on a validation set, last epoch, etc.) used to determine which checkpoint to release?
Thank you for your time and clarification.
Metadata
Metadata
Assignees
Labels
No labels