Reputation: 1
From this link, we can see
The Gen AI evaluation service uses gemini-1.5-pro as a judge model
But I want to customize judge model by myself, for example, using Claude. I didn't find some instruction about it.
And then I found an official document in Github: Customize Model-based Metrics to Evaluate a Gen AI model . And there's a instruction named Create a fully-customized PointwiseMetric
. I'm not sure if I can use model = GenerativeModel("model_version", generation_config=generation_config)
to generate a model and then assign it as my judge model by free_form_pointwise_result = free_form_pointwise_eval_task.evaluate(model=model)
.
Dose anybody know about this?
Upvotes: 0
Views: 29