Question on the few-shot setting

Question

Question on the few-shot setting

Closed this issue a year ago · 1 comments

This is very interesting work for sure.

All the existing experiment results seem to be only on the "few-shot" settings, from 16-256 training examples, will this approach can also be applied to those scenarios with sufficient training examples? If not, what could be the possible limitations and do you have any plans working on that?

Answer 1 · 2023-07-20T11:30:18.000Z

Sorry for the late reply and thank you for your attention.
The reason we only conduct “few-shot" setting is due to GPU resource constraints. We will include more data volume analysis in our future version. We believe it can also be applied to sufficient training examples, but it also needs a lot GPU consumption.