search

kjappelbaum / gptchem

https://gptchem.readthedocs.io/en/latest/
MIT License
232 stars 41 forks source link

set logit biases for classification #16

Open kjappelbaum opened 1 year ago

kjappelbaum commented 1 year ago

https://help.openai.com/en/articles/5247780-using-logit-bias-to-define-token-probability

kjappelbaum commented 1 year ago

put all tokens expect the ones needed for the classes and end of sequence to -100.

perhaps also explore tuning this for imbalanced data