About experiment setting

[jhyukjang]

Thanks for sharing the codes.

I have a minor question.

In the experiments, does the model converges enough for 150 epochs and there is no significant performance improvement if you train the model more? If not, is there any reason you trained your model for 150 epochs?

[tjpulkl]

font{
    line-height: 1.6;
}
ul,ol{
    padding-left: 20px;
    list-style-position: inside;
}

    I have not trained the model more than 150 epochs. I just follow the previous papers that use the same epochs.
    I dont know if there is significant performance impovement when train the model more epochs. I just select the sameepochs with other papers for the convenience of fair comparison with previous performances.

On 7/12/2022 19:50，Jaehyuk ***@***.***> wrote： 

Thanks for sharing the codes. I have a minor question. In the experiments, does the model converges enough for 150 epochs and there is no significant performance improvement if you train the model more? If not, is there any reason you trained your model for 150 epochs?

—Reply to this email directly, view it on GitHub, or unsubscribe.You are receiving this because you are subscribed to this thread.Message ID: @.***>

[jhyukjang]

Thanks for your reply! I have one more question. Then did you choose the model of the final epoch like 149 for the performance report, not the best model for all 150 epochs?