Open findalexli opened 1 year ago
Hi, did all four of your one-on-one experiments work well
For sleep eeg to epilepsy, I also fix a basic model gradient kept accumulating bug. After that, my result on epilepsy are still bad with the default number of samples. I emailed the author and was advices to increasing the fine tuning number, and increasing it to beyond. 200 worked
I briefly tried other experiment and ran into issues, which I didn’t invest more time fixing.
Yours, Alex On Apr 10, 2023 at 11:05 PM -0700, yuyunannan @.***>, wrote:
Hi, did all four of your one-on-one experiments work well — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you authored the thread.Message ID: @.***>
As I am attempting to go through the codebase, I noticed that both the finetuning and testing accuracy metrics are not computed correctly
There are couple main issue
This is my working script under which I commented out the 'best performance' and changed to global evaluation.