nvlabs / afno-transformer Goto Github PK

Hi, first of all thanks for a very interesting paper.
Could you clarify in more details the results in table 6. AFNO [Non-Adaptive Weights]?
What you mean by Adaptive weights? No MLP? I can see in this repo exist a BFNO module was it used in this experiment?

There seems to be no code for experimenting with Cityscapes segmentation and Few-shot segmentation.
Can you release the code for the two experiments?

@jtguibas, @MortezaMardani
Dear All,
I have already applied FNO for a case study related to the Darcy equation using the data generated by myself (input of 30 & 30 ----> output of 30 & 30. Now, I am supposed to enhance its performance. Do you think the use of AFNO is sensible in this sense?
Thanks so much.