Official repository of OFA (ICML 2022). Paper: OFA: Unifying Architectures, Tasks, and Modalities Through a Simple Sequence-to-Sequence Learning Framework
Is there a way to do few shots with OFA for VQA? I saw there is a zero shot function but I saw nothing about doing few shots in the code or the paper. So, was wondering if maybe i've missed something or is there a reason for not doing few shots?
We haven't tried it. As to fewshot, fewshot by finetuning is possible, but fewshot by in-context learning as Flamingo does is still not possible due to our model and training design.
Is there a way to do few shots with OFA for VQA? I saw there is a zero shot function but I saw nothing about doing few shots in the code or the paper. So, was wondering if maybe i've missed something or is there a reason for not doing few shots?