Is it possible to finetune with a custom dataset on an STS format? In specific, have "sentence #1", "sentence #2" and their similarity score (0-5). The provided examples use only positive retrievals as an example. On the tasks.py the key mapping is only inputs and targets.
Is it possible to finetune with a custom dataset on an STS format? In specific, have "sentence #1", "sentence #2" and their similarity score (0-5). The provided examples use only positive retrievals as an example. On the
tasks.pythe key mapping is onlyinputsandtargets.Many thanks