ValueError: The provided metric AccuracyNovel for keeping the best model is not computed by the evaluation routine. about fewshotwithoutforgetting HOT 4 OPEN

This issue is caused by different versions of pytorch. Higher versions of Pytorch will return a tensor and lower version will return the type of float for this line:
https://github.com/gidariss/FewShotWithoutForgetting/blob/master/algorithms/FewShot.py#L20
Then in the update method(https://github.com/gidariss/FewShotWithoutForgetting/blob/master/utils.py#L154), AccuracyNovel will be ignored.

from fewshotwithoutforgetting.

change back to pytorch version==0.4

from fewshotwithoutforgetting.

This issue is caused by different versions of pytorch. Higher versions of Pytorch will return a tensor and lower version will return the type of float for this line:
https://github.com/gidariss/FewShotWithoutForgetting/blob/master/algorithms/FewShot.py#L20
Then in the update method(https://github.com/gidariss/FewShotWithoutForgetting/blob/master/utils.py#L154), AccuracyNovel will be ignored.

nice!!!

from fewshotwithoutforgetting.

This issue is caused by different versions of pytorch. Higher versions of Pytorch will return a tensor and lower version will return the type of float for this line: https://github.com/gidariss/FewShotWithoutForgetting/blob/master/algorithms/FewShot.py#L20 Then in the update method(https://github.com/gidariss/FewShotWithoutForgetting/blob/master/utils.py#L154), AccuracyNovel will be ignored.

Hi, bro, i use the same pytorch version like u, but i still get the error ,could you help me ?

from fewshotwithoutforgetting.