Dear authors,

I am trying to the link you provided to access the pretraining code you used, but clicking on the link gives me the Page not found error.

Would you mind double-checking whether this link is still valid? Thank you!

Dear Authors,
Many thanks for the great work. While going through the paper I could not understand how you compute the final accuracy on the test samples. Is the accuracy computed in an online manner (evaluate the model on the current batch as it trains on) or offline manner (evaluate the whole test set after training is finished)?

Thanks.

You should update the model under model.train(), and then test it under model.eval().
I'm sorry if I have misunderstood.

Thank you for your good work. Can you release the training code for the pre-trained model on CIFAR10/100 (the checkpoint of pre-trained Resnet-50 can be downloaded (214MB)…………)

There is a discrepancy about test-time adaptation in this code that has me wondering.

When adaptation operation runs on the test set, TTT and Tent perform only one epoch instead of hundreds of epochs. As I understand it, this code performs multiple epochs of adaptation to the network on the test set, which often does not make sense in practice in my opinion.

Hello
Thank you for your great work!
Because load_state_dict is kind of deepcopy which gives the model an independant memory space. I am wondering if the weight of models (net, head, ssh) still sync after net.load_state_dict(net_dict) and head.load_state_dict(head_dict) in function load_resnet50() in test_helper.py.

Can you release the pre-train model of VisDA-C and corresponding test-time training code?

I'm confused of how the covariance is calculated in the code, the result of covariance function is different from what np.cov gives. I'm sorry if I have misunderstood.