Can we restrict the entity values and also output confidence scores
Thanks for open sourcing the model. I have couple of followup questions.
Is it possible to o/p confidence score for each entity values that the model o/ps.
Is it possible to restrict the output (value of an entity) to predefined set .
In the example,
schema = """{
"Model": {
"Name": "",
"Number of parameters": "",
"Number of max token": "",
"Architecture": []
},
"Usage": {
"Use case": [],
"Licence": ""
}
}"""
Can we restrict values of Model i.e ( entities) to - [ "Llama, "GPT","Phi"] . The model is forced to pick one of these and additionally o/p a confidence score for each of them .
Hi , thanks for trying the model. While we do not have such a feature implemented you should be able to do so by writing your own generation/sampling function which checks the logits for specific vocabulary items.