Initial training and finetuning #13
Description
Hello, and thank you for the great work!
I have a question about your training stages (initial training and finetuning).
Could you clarify the difference between the initial training stage and the finetuning stage?
And if I want to train on my own custom dataset (video referring), should I run both stages, or is the initial training stage sufficient?
Any recommended workflow would be appreciated (e.g., initialize from your released checkpoints --> finetune on custom data).