pytorch 1.7.0 cuda 11.0 setup Error

[panfengsu]

I need to setup mmdetection3D in the environment for pytorch1.7.0 cuda11.0.I modify Dockerfile:
ARG PYTORCH="1.7.0"
ARG CUDA="11.0"
ARG CUDNN="8"
FROM pytorch/pytorch:${PYTORCH}-cuda${CUDA}-cudnn${CUDNN}-devel
ENV TORCH_CUDA_ARCH_LIST="6.0 6.1 7.0+PTX"
ENV TORCH_NVCC_FLAGS="-Xfatbin -compress-all"
ENV CMAKE_PREFIX_PATH="$(dirname $(which conda))/../"
RUN apt-get update && apt-get install -y ffmpeg libsm6 libxext6 git ninja-build libglib2.0-0 libsm6 libxrender-dev libxext6
&& apt-get clean
&& rm -rf /var/lib/apt/lists/*

Install MMCV

RUN pip install mmcv-full==latest+torch1.7.0+cu110 -f https://openmmlab.oss-accelerate.aliyuncs.com/mmcv/dist/index.html

But I get error:
ERROR: Command errored out with exit status 1:
command: /opt/conda/bin/python -c 'import sys, setuptools, tokenize; sys.argv[0] = '"'"'/mmdetection3d/setup.py'"'"'; file='"'"'/mmdetection3d/setup.py'"'"';f=getattr(tokenize, '"'"'open'"'"', open)(file);code=f.read().replace('"'"'\r\n'"'"', '"'"'\n'"'"');f.close();exec(compile(code, file, '"'"'exec'"'"'))' develop --no-deps

/mmdetection3d/mmdet3d/ops/voxel/src/scatter_points_cuda.cu(242): error: class "at::Tensor" has no member "max_values"
/mmdetection3d/mmdet3d/ops/voxel/src/scatter_points_cuda.cu(249): error: type name is not allowed
/mmdetection3d/mmdet3d/ops/voxel/src/scatter_points_cuda.cu(249): error: expected an expression

Do you test setup mmdetection in the the environment for pytorch1.7.0 cuda11.0?Can you give some suggestions on this issue？Thanks

[ConnorMcShane]

I am having the same issue.

[Tai-Wang]

Please refer to #362 .

[panfengsu]

@Tai-Wang I compile tag v0.10.0 and do not get error, thanks.

[Tai-Wang]

@Tai-Wang I compile tag v0.10.0 and do not get error, thanks.

Welcome. So it seems like a compatibility problem introduced by PRs #318 and #326 . Thanks for your feedback.
Close this issue due to duplication.

[superchenyan]

I need to setup mmdetection3D in the environment for pytorch1.7.0 cuda11.0.I modify Dockerfile:
ARG PYTORCH="1.7.0"
ARG CUDA="11.0"
ARG CUDNN="8"
FROM pytorch/pytorch:${PYTORCH}-cuda${CUDA}-cudnn${CUDNN}-devel
ENV TORCH_CUDA_ARCH_LIST="6.0 6.1 7.0+PTX"
ENV TORCH_NVCC_FLAGS="-Xfatbin -compress-all"
ENV CMAKE_PREFIX_PATH="$(dirname $(which conda))/../"
RUN apt-get update && apt-get install -y ffmpeg libsm6 libxext6 git ninja-build libglib2.0-0 libsm6 libxrender-dev libxext6
&& apt-get clean
&& rm -rf /var/lib/apt/lists/*

Install MMCV

RUN pip install mmcv-full==latest+torch1.7.0+cu110 -f https://openmmlab.oss-accelerate.aliyuncs.com/mmcv/dist/index.html

But I get error:
ERROR: Command errored out with exit status 1:
command: /opt/conda/bin/python -c 'import sys, setuptools, tokenize; sys.argv[0] = '"'"'/mmdetection3d/setup.py'"'"'; file='"'"'/mmdetection3d/setup.py'"'"';f=getattr(tokenize, '"'"'open'"'"', open)(file);code=f.read().replace('"'"'\r\n'"'"', '"'"'\n'"'"');f.close();exec(compile(code, file, '"'"'exec'"'"'))' develop --no-deps

/mmdetection3d/mmdet3d/ops/voxel/src/scatter_points_cuda.cu(242): error: class "at::Tensor" has no member "max_values"
/mmdetection3d/mmdet3d/ops/voxel/src/scatter_points_cuda.cu(249): error: type name is not allowed
/mmdetection3d/mmdet3d/ops/voxel/src/scatter_points_cuda.cu(249): error: expected an expression

Do you test setup mmdetection in the the environment for pytorch1.7.0 cuda11.0?Can you give some suggestions on this issue？Thanks

I meet the same problem, cuda 11.0, torch1.7.1, when install mmcv-full, I use the followed commend:
MMCV_WITH_OPS=1 MMCV_CUDA_ARGS='-gencode=arch=compute_80,code=sm_80' pip install -e
it sloved, the mmcv hasi sucessfully installed.
Does it works for mmdetction3d? I don't know how to set '-gencode=arch=compute_80,code=sm_80' when install mmdetection3d