I want a measure of perplexity for each token in a string like in the research doc “Perplexity of Fixed Length Models” by @joeddav.
What is the quickest way to accomplish for testing an idea? Is there some way to do it via the Inference API?
I want a measure of perplexity for each token in a string like in the research doc “Perplexity of Fixed Length Models” by @joeddav.
What is the quickest way to accomplish for testing an idea? Is there some way to do it via the Inference API?