search

YanchaoYang / FDA

Fourier Domain Adaptation for Semantic Segmentation
491 stars 79 forks source link

Train from scratch requirements #15

Open Nadavc220 opened 3 years ago

Nadavc220 commented 3 years ago

Hi, Thanks for this contribution. What are the requirements for training the model from scratch? being more specific: 1) What gpu should be used? 2) how much gpu memory is needed? 3) which pytorch version? 4) which python version? 5) which cuda version?

I there anything crucial needed for reproducing the results of the paper? from my experience minor changes in the training environment can make a big difference.

Thanks =)

victorvini08 commented 3 years ago

Heyy mann!....We had the same doubts like yours especially in the part of generation of pseudo labels. We are using a tesla T4 gpu but the get pseudo label script is getting killed after processing half images.....Could you find any solution??

Nadavc220 commented 3 years ago

Hi, Unfortunately, I did not get to the point of running the pseudo label script yet. I guess I will get there eventually and will try o tackle your problem. If you get to solve this I will be happy if you can find the time to post it. I will do the same if I get there before you.

KaiiZhang commented 3 years ago

Heyy mann!....We had the same doubts like yours especially in the part of generation of pseudo labels. We are using a tesla T4 gpu but the get pseudo label script is getting killed after processing half images.....Could you find any solution??

It is caused by insufficient memory. Because all prediction results of the entire dataset need to be stored in memory for sorting. In my implementation, I used 97G of memory.