Efficiency on H100 GPU

[liuxu77]

Hi, thanks for sharing this great work.

I tested this code on ml-1m dataset on a single Nvidia H100 GPU card and it took about 90 minutes to complete 100 epochs. I wonder if this is the expected speed of the code? Thanks.

[jiaqizhai]

Hi - thanks for your interest in our work! We published code for Section 4.1.1 then later Section 4.2 but as you observed, the two pieces are not fully integrated in the current repo. It should be (hopefully) simple to add some glue code.