zhangyp15 / OccFormer

[ICCV 2023] OccFormer: Dual-path Transformer for Vision-based 3D Semantic Occupancy Prediction
https://arxiv.org/abs/2304.05316
Apache License 2.0
324 stars 22 forks source link

Expected dtype object, got 'numpy.dtype[uint8]' #9

Closed zhangzw12319 closed 1 year ago

zhangzw12319 commented 1 year ago

When I try to run test.py with one GPU, the terminal reports Expected dtype object, got 'numpy.dtype[uint8] in Line 107 of projects/mmdet3d_plugin/datasets/pipelines/loading_nusc_occ.py. But i have checked Line122-123, nb.jit('u1[:,:,:](u1[:,:,:],i8[:,:])', nopython=True, cache=True, parallel=False) and the datatype matches the processed_label and label_voxel_pair. Can you help me solve this issue? Thx!

zhangzw12319 commented 1 year ago

Sorry for disturbances. I've found the reason that numpy>=1.20 is incompatitble with mmdet3d and numba==0.48, which are installed as the default enviornment of OccFormer. The numpy>=1.20 is brought by conda install pytorch==1.10.1 torchvision==0.11.2 torchaudio==0.10.1 cudatoolkit=11.3 -c pytorch. It can be solved by installing numpy=1.19.5 after installing pytorch via conda command. For detailed reason, plz see this issue.