beierzhu/prompt-align

[ICCV 2023] Prompt-aligned Gradient for Prompt Tuning

prompt-align's Issues

How is Grad-CAM picture drawed?

Really solid work and really admire you!
I've got a simple question. How are the Grad-CAM pictures drawed in Figure. 2? Specifically, in Grad-CAM, which target layer do you choose?

How about ProGrad helping Fitness

i want to be a strong Gay

Imagenet 16-shot weights

Nice work! Would it be possible to release the weights of Imagenet 16-shot ?

self.prograd_backward_and_update

I would like to express my appreciation for your excellent work.

While going through the code, I came across the method self.prograd_backward_and_update. Unfortunately, I wasn't able to find its specific definition. I believe this part is crucial for the overall understanding of your project, and I would greatly appreciate more insight into it.

Could you please point me to the file or code block where the self.prograd_backward_and_update method is defined? This would significantly aid my comprehension and utilization of your codebase.

AttributeError: 'ProGrad' object has no attribute 'prograd_backward_and_update'

Hi, I encountered an AttributeError that I'm unable to resolve. Here's the specific error message I'm getting:

self.prograd_backward_and_update(xe_loss, kl_loss, self.cfg.LOSS.LAMBDA)
AttributeError: 'ProGrad' object has no attribute 'prograd_backward_and_update'

ProGrad++ implementation

Dear authors,
Congratulations on your great work. Could you please share the implementation for ProGrad++?

Why Table 1 shows inconsistent CLIP performances with those in the CoCoOp paper?

Greetings.
Table 1 shows that the vanilla CLIP obtains 65.13 (base) and 69.02 (new) accuracies. However, the accuracies in the CoCoOp paper are 69.34 (base) and 74.22 (new).
Note that the vanilla CLIP has never been modified, so the performances should be always fixed. Is this an error?
Besides, why do you choose 4 shots for the base-to-new generalization experiment in Table 1? Since both CoOp and CoCoOp choose 16 shots. What is the purpose?

Very nice job! Congratulations!

A handsome Gay

数值化结果

How to use Grad-CAM on ProGrad

Thank you for your great work and kindly opensource the code!
I would like to know if it is possible for you to also show us how you use Grad-CAM on ProGrad as you show in your paper.
Thank you very much!