HDF5 group paths do not need to be OS generic

drprojects / superpoint_transformer

Official PyTorch implementation of Superpoint Transformer introduced in [ICCV'23] "Efficient 3D Semantic Segmentation with Superpoint Transformer" and SuperCluster introduced in [3DV'24 Oral] "Scalable 3D Panoptic Segmentation As Superpoint Graph Clustering"

MIT License

560 stars 72 forks source link

HDF5 group paths do not need to be OS generic #37

Closed rjanvier closed 10 months ago

rjanvier commented 10 months ago

HDF5 group paths are currently built using os.path.join. It results in using \\ separator on Windows OSes but HDF5 group paths separator is / no matter the underlying OS is in use and thus it creates incorrect hdf5 files on Windows. This PR intends to correct this issue by forcing the use of / separator everywhere.

rjanvier commented 10 months ago

I have something working on Windows on my side... And also something full CPU (timings are ok for inference) but it's intricated with other change I made for our specific needs here. So expect more commits in the comming days :)