yizhou-wang/RODNet

Number of GPUs and Train time

AmrinKareem opened this issue · 1 comments

Hello team,

Could you please provide information on how many GPUs were used to train the RODNet and how long it took to train the network? Thanks in advance!

The authors of the paper trained RODNet using a single V100 GPU.

ref: https://arxiv.org/pdf/2102.05150.pdf
Moreover, we train the RODNet on an NVIDIA Quadro GV100 GPU. It takes about 4 days to train the RODNet (vanilla), 8 days to train the RODNet (HG), and 10 days to train the RODNet (HG) with temporal inception convolution layers.