-
Notifications
You must be signed in to change notification settings - Fork 138
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
memory_size,memory_per_class #2
Comments
If the number of pictures in the second increment is much smaller than that in the first increment, how should I set these two parameters |
When fixed memory is set to false, memory size is used, which means that we keep memory size images for all the old classes, which means that for each of the k old categories, we keep memory size/k(floor) images. On the contrary, when fixed memory is set to true, memory_per_class will be used, which means that we will keep memory_per_class images for each category. In other words, its memory size increases with the number of categories. |
In most cases, to be fair, when training cifar100 and imagenet100, we set memory size to 2000, and when training imagenet1000, we set memory size to 20000. |
I probably understand. Thank you for your explanation |
Thanks for your interests. 😊 |
Hello, thank you very much for the code you provided. Can you tell me what the parameter memory_size and memory_per_class represents?
The text was updated successfully, but these errors were encountered: