search

NVlabs / ODISE

Official PyTorch implementation of ODISE: Open-Vocabulary Panoptic Segmentation with Text-to-Image Diffusion Models [CVPR 2023 Highlight]
https://arxiv.org/abs/2303.04803
Other
845 stars 45 forks source link

Running equipment and running time #4

Closed 123456klk1 closed 1 year ago

123456klk1 commented 1 year ago

Hello, I have learned about your work. I feel really cool. I would like to ask, what is the running equipment and running time?

KinGeorge commented 1 year ago

In Getting Started with ODISE , it says that ODISE trains in 6 days on 32 NVIDIA V100 GPUs. But in the orignial paper's implementation details, it said that We use 16 V100 on coco_50e takes 2 days, so I'm a bit confused

shalinidemello commented 1 year ago

Experimental results in section 4.3 take takes 2 days on 16 V100 machines. To reproduce the results in section 4.2 takes 6 days on 32 V100 GPUs.