QLYYLQ

just install another torch version is ok, i thought. ```python python -c "import torch; print(torch.__version__); print(torch.version.cuda)" ``` install pytorch with same cuda version for `nvcc -V` Then you can go...

as shown in code: dataset/voc.py. You should set a own Segmentation(like VOCSegmentation in voc.py) which can use __getitem__ to visit total dataset, then you can write an incremental dataset, like...

Maybe you can try this, i solved an non-reason install problem (doing correct steps in apex's readme) using this: ```shell git checkout origin/22.02-parallel-state # or another vesion python setup.py install...

> git clone https://github.com/NVIDIA/apex.git cd apex git tag # if not wanted version git branch -r git checkout origin/22.02-parallel-state # or another vesion python setup.py install --cuda_ext --cpp_ext pip3 install...

哥们你cuda版本多少，我这边10.8的时候换分支安上了。 you can try this after clone: ```shell git checkout origin/22.02-parallel-state # or another vesion python setup.py install --cuda_ext --cpp_ext pip3 install -v --no-cache-dir ./ ```