Hello everyone, I want to use the local code library (with many code files) for GPT to fine-tune. How should I construct this training data, and how should I split the code files and feed them to the GPT model.
Then when I use the fine-tuned GPT model to generate code, the model can use my local code library.
for example:
I have a code library that implements an interface for issuing rewards to users
public class RewardService {
/**
* 发放奖励
* @param uid 用户uid
* @param rewardId 奖励id
* @param num 奖励数量
* @param dueTime 到期时间
*/
public void sendReward(long uid, int rewardId, int num, Date dueTime) {
......
}
}
When use the fine-tuned GPT input prompts “Help me generate code, send 2 numbers 123 reward to user 456”, GPT model can use my local code library.