aehrc / cvt2distilgpt2

Improving Chest X-Ray Report Generation by Leveraging Warm-Starting
GNU General Public License v3.0
62 stars 5 forks source link

Training loss consultation #15

Open yihp opened 3 months ago

yihp commented 3 months ago

Hi,
I would like to ask what value the training loss finally converges to. Thank you.

anicolson commented 3 months ago

Hi @yihp,

For which dataset? What values are you getting?

Thanks, A.

yihp commented 3 months ago

I'm testing it on my own dataset now, thanks for your reply!

anicolson commented 3 months ago

Your loss curve can depend on many factors, including your dataset, as I am sure you are aware 😀. If you would like you could share your loss curve. More telling would be your metric scores if you would like to share them as well.