The repeat time parameter in the dataset configuration file

[ChenJian7578]

Is there any benefit to model fine-tuning when the repeat_time parameter in the data set configuration file is greater than 1

[yuecao0119]

Thank you very much for your question.
Regarding the repeat_time parameter, my understanding is that on the one hand, the amount of high-quality data is relatively small, and on the other hand, the amount of data in different fields is different. By setting repeat_time, the number of active data samples and the data mixing ratio can be adjusted to make better use of high-quality data. In addition, increasing the repeat_time of a small amount of data to increase its proportion in the entire data can allow the model to quickly adapt to this specific data.
I hope this can help you.

[ChenJian7578]

Can it be understood that when I have only one custom dataset, the repeat time argument is equivalent to the nummber_reain_epoch argument?

[czczup]

Can it be understood that when I have only one custom dataset, the repeat time argument is equivalent to the nummber_reain_epoch argument?

Yes, you are right.