This is not the first time I have encountered this problem, but when continuing training and resuming state from last-state, and saving new checkpoints, the same progress is retained. for example, new checkpoints have exactly the same hash. I use notebook for finetuning.
This is not the first time I have encountered this problem, but when continuing training and resuming state from last-state, and saving new checkpoints, the same progress is retained. for example, new checkpoints have exactly the same hash. I use notebook for finetuning.