Open wusuoweima opened 3 years ago
We use lr 1e-6, 50 epochs for CUB-200-2011 (100 classes) and Caltech-101 (Fig.6). The parameters of other datasets are shown in 'Implementation Details' part of the paper.
Which 100 classes do you choose for CUB-200? First 100? For CUB-200 and Caltech do you pre-train on the first task? For Caltech do you pre-train on Imagenet?
We follow the LwM paper and use the first 100 for CUB-200 in Fig6. For CUB and Caltech we use the pretrained Imagenet.
Could you please provide more run_demo.sh about other datasets(CUB-200-2011, Flowers-102, Caltech-101, ImageNet-Subset) ? @yulu0724