Training from checkpoint fails

[lbugnon]

EDIT: Trying to use checkpoint as starting point, realized that provided checkpoint is not meant for resuming training.
To work around this, i added a "pretrained" to the structure training config to load the checkpoint. Maybe a cleaner finetuning script will be required

[gcorso]

Hi @lbugnon, yes for finetuning the config to use should follow this one https://github.com/jwohlwend/boltz/blob/main/scripts/train/configs/full.yaml