model.prediction does not match model.evaluation loss error

[Manal4]

Hi
Thanks for the tutorial.
I applied the same code using different data, it worked perfectly, I added the mes, mae metrics in addition to the loss fuction, but when I tried to calculate the individual MSEs, it seems that the error from model.prediction() does not match the error from the model.evaluation().
I just add this line in the comparison function:

                             `  mean_squared_error(signal_true,signal_pred) `

I also divide it by 1000 but the total of mean squared errors from different signals does not match the mse from the model.evaluation().
thanks.

[Manal4]

I saw @bstriner reply #keras-team/keras#5140 (comment) regarding a similar issue in #keras-team/keras#5140, but I couldn't fix it, it seems to me that shapes/data types are correct.

[Hvass-Labs]

I don't know what you are referring to, but I have just updated many of the tutorials to support TensorFlow 2, so you can try again with the new version.

[STEVE-ELF]

Thanks for your letters.

…

------------------&nbsp;原始邮件&nbsp;------------------ 发件人:&nbsp;"Hvass-Labs"<[e-mail]&gt;; 发送时间:&nbsp;2020年4月2日(星期四) 晚上8:52 收件人:&nbsp;"Hvass-Labs/TensorFlow-Tutorials"<[e-mail]&gt;; 抄送:&nbsp;"Subscribed"<[e-mail]&gt;; 主题:&nbsp;Re: [Hvass-Labs/TensorFlow-Tutorials] model.prediction does not match model.evaluation loss error (#119) I don't know what you are referring to, but I have just updated many of the tutorials to support TensorFlow 2, so you can try again with the new version. — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub, or unsubscribe.