How do you achieve 96.9% for TAUP+Supervised Finetuning?

nikheelpandey / TAUP-PyTorch

Task Agnostic Unsupervised Learning

15 stars 0 forks source link

How do you achieve 96.9% for TAUP+Supervised Finetuning? #5

Open BenjiKCF opened 3 years ago

BenjiKCF commented 3 years ago

I am having trouble to reproduce the result in your table? Can you indicate what's the dataset, batchsize and model for that table? Is that Cifar 10 and resnet18?

nikheelpandey commented 3 years ago

Hi, seems like there is some issue with weight initialization. Even I am having a hard time reproducing the result. BTW, I used a couple of datasets for end-to-end evaluation: STL10 and FashionMNIST. The base model I used was resnet18. I hope that helps.