Giters

facebookresearch / mobile-vision

Mobile vision models and code

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

GPU memory in searching

wangq95 opened this issue · comments

commented

Hi, may I ask a question about the GPU memory occupation during training? Is that smaller than DARTS as only one block was actived for each layer? But the $mi$ in Eq.5 is relexed by Gumbel Softmax so that every $mi$ is greater than zero, so every block should be calculated, how it different from DARTS? Thank you.