Why does the gpt3 model is not able to learn context with samples like seq2seq model based sample? Can’t we train with seq2seq inputs and as questions in prompt?
