about non-differentiable top-k

[xiewende]

Nice Works!

Being very interested in your work，But I found that your top-k ranking removes the differentiable top-k method and uses softmax instead, what is the reason for this ?

[fnzhan]

Hi, I have a deadline recently and will include the code of this part early next month.

[xiewende]

Looking forward to your work!

[fnzhan]

Hi, I update the implementation of top-k including ot_topk and differentiable_topk in models/networks/ranking_attention.py. You can also directly test the top-k function in util/topk_test.py.