Bug calculating group accuracy in ./ulits/toolkit.py

[raziqueA]

when calculating group accuracy in ./ulits/toolkit.py file, the accuracy for the last group in each task is calculated incorrectly.

For example CIFAR100, which has classes [0, 1, 2, ..., 99] and Base 0 Increment 10 case, the issue arises when the code attempts to compute the accuracy for each task. In the first task with a base of 0 and an increment of 10, the code only calculates the accuracy for the first 9 classes ([0, 1, 2, ..., 8]), while it should be calculating the accuracy for all 10 classes ([0, 1, 2, ..., 9]). In subsequent tasks, for example, task 2 with 20 classes ([0, 1, 2, ..., 19]), the code correctly computes the accuracy for the first 10 classes (task 1) but then inaccurately computes the accuracy for task 2 by only considering classes from 10 to 18, missing the last class (19). similarly happens for all the tasks

i have added a pull request too.

[simomagi]

Yes, I have noticed the same. By fixing as suggested by @raziqueA , the accuracy computation should be fine.

[caoshuai888]

这是来自QQ邮箱的假期自动回复邮件。   您好，我最近正在休假中，无法亲自回复您的邮件。我将在假期结束后，尽快给您回复。