zllxot / CORE

ICCV2023 - CORE: Cooperative Reconstruction for Multi-Agent Perception
38 stars 3 forks source link

代码运行有问题的请看这个 #7

Open fengxueguiren opened 6 months ago

fengxueguiren commented 6 months ago

如果代码提示不能导入AttBEVBackboneCORE的错误, 请将代码的配置文件修改为point_pillar_intermediate_core.yaml而不是point_pillar_core.yaml, 当batch_size=1时,还需要修改CORE/opencood/models/fuse_modules/core_attentive_collaboration.py中的代码:

att = self.convatt(q, v, att_map)

att = self.convatt(q.unsqueeze(0), v.unsqueeze(0), att_map)
Michael-Chao commented 6 days ago

point_pillar_intermediate_core.yaml这个配置文件是不是没有办法多GPU运行?voxelnet_intermediate_fusion.yaml试的这个是可以的