train_api_request和complete_gpt3区别在哪

[steveluo01]

    for m, missing_batch in enumerate(batches['sentence']):
        if split == "train":
            missing_response = train_api_request(missing_batch, l=0, model_name=args.model_name_or_path, num_log_probs=1, echo=True)
        else: 
            missing_response = complete_gpt3(missing_batch, l=0, model_name=args.model_name_or_path, num_log_probs=1, echo=True)

train_api_request和complete_gpt3区别在哪

[shizhediao]

感谢关注！ train_api_request的目的是为了计算API调用次数，于是在complete_gpt3的外层套了一个counter。

---

Thanks for attention! The purpose of train_api_request is to count the number of API calls, so a counter is set in the outer layer of complete_gpt3.