How to reset callbacks when calling "load_from_checkpoint"

[jannisborn]

Say I'm loading pretrained model using load_from_checkpoint(ckpt_path). If I used callbacks during the pretraining, the callback values will be restored (e.g., the minimal loss).
This is very useful if one continues to train on the same data but it is prohibitive if the model is finetuned on another dataset.
I'm looking for an option to reset all callback values. Is this supported?

I know that I can overwrite hyperparameters (the docs are pretty verbose about it) but that's not what I want.
Thanks for any advice!

[rohitgr7]

load_from_checkpoint just loads the model weights. It doesn't load the callback states.