I was wondering whether we can use OpenAIGPTDoubleHeadsModel to get the most likely sentence given the previous sentences.
Since it is mentioned in the documentation that it can be used for RocStories, this seems to be the case indeed. But how can we do that? Is there a way to adapt the example (shown in the documentation: https://huggingface.co/transformers/model_doc/gpt.html#openaigptlmheadmodel ) in such a way that this is possible?? For instance in this case, how can we give a set of sentences and then let the model check which one of the sentences below is the most likely continuation?
["Hello, my dog is cute [CLS]", "Hello, my cat is cute [CLS]"]