Thomas Willberger

Results 2 comments of Thomas Willberger

I tried the scheduler from the paper, turns out to be totally instable. I'd also be curious to know the reason. The architecture seems identical. ![image](https://github.com/thuanz123/enhancing-transformers/assets/7074079/ec9c8e5b-e1fc-4518-b6f5-13da084945ce)

Would be great to see this PR merged!