search

lhoyer / HRDA

[ECCV22] Official Implementation of HRDA: Context-Aware High-Resolution Domain-Adaptive Semantic Segmentation
Other
231 stars 32 forks source link

DDP train #7

Closed wzlyyb closed 2 years ago

wzlyyb commented 2 years ago

Hello,ihoyer: When I use ddpwrapper to train your model, the final result only achieve 71.57 mIoU. Why the performance will fill ?Can you give me some suggestion ?

lhoyer commented 2 years ago

We only used single GPU training in this project. Therefore, I don't know why you observe this behavior.

wzlyyb commented 2 years ago

thank you very much