RuntimeError: CUDA error: no kernel image is available for execution on the device

flyingalpaca · February 5, 2021, 6:49am

Hello,

I’m trying to set up a machine learning model but have the following error showing up
RuntimeError: CUDA error: no kernel image is available for execution on the device

The system I am using is:
Ubuntu 18.04
Cuda toolkit 10.0
Nvidia driver 460
2 GPUs, both are GeForce RTX 3090

I think the problem may also be due to the driver as when I open the “Additional Driver”, I see the following.

The person who set up this computer tried to install nvidia-driver-455 but nothing shows up.

Does anyone know how to solve this?

Thank you

generix · February 5, 2021, 11:09pm

Please check if you have the correct cuda version installed for what you’re trying to run.
Please post the output of the deviceQuery demo that comes with cuda.

flyingalpaca · February 9, 2021, 12:02am

Hello,

Below is the output to deviceQuery

./deviceQuery Starting...

 CUDA Device Query (Runtime API) version (CUDART static linking)

Detected 2 CUDA Capable device(s)

Device 0: "GeForce RTX 3090"
  CUDA Driver Version / Runtime Version          11.2 / 10.0
  CUDA Capability Major/Minor version number:    8.6
  Total amount of global memory:                 24268 MBytes (25447170048 bytes)
MapSMtoCores for SM 8.6 is undefined.  Default to use 64 Cores/SM
MapSMtoCores for SM 8.6 is undefined.  Default to use 64 Cores/SM
  (82) Multiprocessors, ( 64) CUDA Cores/MP:     5248 CUDA Cores
  GPU Max Clock rate:                            1800 MHz (1.80 GHz)
  Memory Clock rate:                             9751 Mhz
  Memory Bus Width:                              384-bit
  L2 Cache Size:                                 6291456 bytes
  Maximum Texture Dimension Size (x,y,z)         1D=(131072), 2D=(131072, 65536), 3D=(16384, 16384, 16384)
  Maximum Layered 1D Texture Size, (num) layers  1D=(32768), 2048 layers
  Maximum Layered 2D Texture Size, (num) layers  2D=(32768, 32768), 2048 layers
  Total amount of constant memory:               65536 bytes
  Total amount of shared memory per block:       49152 bytes
  Total number of registers available per block: 65536
  Warp size:                                     32
  Maximum number of threads per multiprocessor:  1536
  Maximum number of threads per block:           1024
  Max dimension size of a thread block (x,y,z): (1024, 1024, 64)
  Max dimension size of a grid size    (x,y,z): (2147483647, 65535, 65535)
  Maximum memory pitch:                          2147483647 bytes
  Texture alignment:                             512 bytes
  Concurrent copy and kernel execution:          Yes with 2 copy engine(s)
  Run time limit on kernels:                     Yes
  Integrated GPU sharing Host Memory:            No
  Support host page-locked memory mapping:       Yes
  Alignment requirement for Surfaces:            Yes
  Device has ECC support:                        Disabled
  Device supports Unified Addressing (UVA):      Yes
  Device supports Compute Preemption:            Yes
  Supports Cooperative Kernel Launch:            Yes
  Supports MultiDevice Co-op Kernel Launch:      Yes
  Device PCI Domain ID / Bus ID / location ID:   0 / 161 / 0
  Compute Mode:
     < Default (multiple host threads can use ::cudaSetDevice() with device simultaneously) >

Device 1: "GeForce RTX 3090"
  CUDA Driver Version / Runtime Version          11.2 / 10.0
  CUDA Capability Major/Minor version number:    8.6
  Total amount of global memory:                 24265 MBytes (25443893248 bytes)
MapSMtoCores for SM 8.6 is undefined.  Default to use 64 Cores/SM
MapSMtoCores for SM 8.6 is undefined.  Default to use 64 Cores/SM
  (82) Multiprocessors, ( 64) CUDA Cores/MP:     5248 CUDA Cores
  GPU Max Clock rate:                            1800 MHz (1.80 GHz)
  Memory Clock rate:                             9751 Mhz
  Memory Bus Width:                              384-bit
  L2 Cache Size:                                 6291456 bytes
  Maximum Texture Dimension Size (x,y,z)         1D=(131072), 2D=(131072, 65536), 3D=(16384, 16384, 16384)
  Maximum Layered 1D Texture Size, (num) layers  1D=(32768), 2048 layers
  Maximum Layered 2D Texture Size, (num) layers  2D=(32768, 32768), 2048 layers
  Total amount of constant memory:               65536 bytes
  Total amount of shared memory per block:       49152 bytes
  Total number of registers available per block: 65536
  Warp size:                                     32
  Maximum number of threads per multiprocessor:  1536
  Maximum number of threads per block:           1024
  Max dimension size of a thread block (x,y,z): (1024, 1024, 64)
  Max dimension size of a grid size    (x,y,z): (2147483647, 65535, 65535)
  Maximum memory pitch:                          2147483647 bytes
  Texture alignment:                             512 bytes
  Concurrent copy and kernel execution:          Yes with 2 copy engine(s)
  Run time limit on kernels:                     Yes
  Integrated GPU sharing Host Memory:            No
  Support host page-locked memory mapping:       Yes
  Alignment requirement for Surfaces:            Yes
  Device has ECC support:                        Disabled
  Device supports Unified Addressing (UVA):      Yes
  Device supports Compute Preemption:            Yes
  Supports Cooperative Kernel Launch:            Yes
  Supports MultiDevice Co-op Kernel Launch:      Yes
  Device PCI Domain ID / Bus ID / location ID:   0 / 193 / 0
  Compute Mode:
     < Default (multiple host threads can use ::cudaSetDevice() with device simultaneously) >
> Peer access from GeForce RTX 3090 (GPU0) -> GeForce RTX 3090 (GPU1) : No
> Peer access from GeForce RTX 3090 (GPU1) -> GeForce RTX 3090 (GPU0) : No

deviceQuery, CUDA Driver = CUDART, CUDA Driver Version = 11.2, CUDA Runtime Version = 10.0, NumDevs = 2

I’m very new to GPU based computing so I’ve followed the instructions given from mmskeleton/GETTING_STARTED.md at master · open-mmlab/mmskeleton · GitHub
and just setup the environment with cuda toolkit 10.0.

generix · February 9, 2021, 8:36am

I suppose the pytorch version you’re using doesn’t support the Ampere RTX 3090:
https://github.com/pytorch/pytorch/issues/49161
So you will have to use pytorch 1.7.1+cuda11

flyingalpaca · February 17, 2021, 10:59am

Hello,

Unfortunately, the code I’m trying to run can only run with pytorch=1.2.0
As a result, I’ll need to have cuda 10.0

I found this nvidia document which I believe will allow me to use cuda 10.0 on the RTX3090 architecture
https://docs.nvidia.com/cuda/ampere-compatibility-guide/index.html

So I’ve been trying to run this instruction:

/usr/local/cuda-10.0/bin/nvcc \
  -gencode=arch=compute_52,code=sm_52 \
  -gencode=arch=compute_60,code=sm_60 \
  -gencode=arch=compute_61,code=sm_61 \
  -gencode=arch=compute_70,code=sm_70 \
  -gencode=arch=compute_75,code=sm_75 \
  -gencode=arch=compute_75,code=compute_75 \
  -O2 -o mykernel.o -c mykernel.cu \

but I am getting this error

gcc: error: mykernel.cu: No such file or directory
gcc: warning: '-x c++' after last input file has no effect
gcc: fatal error: no input file
compilation terminated.

Am I doing this correctly?

Thank you

generix · February 17, 2021, 12:07pm

Again, pytorch 1.2 is incompatible to Ampere, so cuda doesn’t matter.

flyingalpaca · February 21, 2021, 7:52pm

Hello,

Thank you for your answer.
I’ve now installed CUDA 11.0 with pytorch 1.7.0 but I’m getting errors when trying to install the model.

FAILED: /home/lab/mmskeleton/build/temp.linux-x86_64-3.7/mmskeleton/ops/nms/nms_kernel.o 
/usr/local/cuda-11.0/bin/nvcc -DWITH_CUDA -Immskeleton/ops/nms -I/home/lab/anaconda3/envs/mmskeleton/lib/python3.7/site-packages/numpy/core/include -I/home/lab/anaconda3/envs/mmskeleton/lib/python3.7/site-packages/torch/include -I/home/lab/anaconda3/envs/mmskeleton/lib/python3.7/site-packages/torch/include/torch/csrc/api/include -I/home/lab/anaconda3/envs/mmskeleton/lib/python3.7/site-packages/torch/include/TH -I/home/lab/anaconda3/envs/mmskeleton/lib/python3.7/site-packages/torch/include/THC -I/usr/local/cuda-11.0/include -I/home/lab/anaconda3/envs/mmskeleton/include/python3.7m -c -c /home/lab/mmskeleton/mmskeleton/ops/nms/nms_kernel.cu -o /home/lab/mmskeleton/build/temp.linux-x86_64-3.7/mmskeleton/ops/nms/nms_kernel.o -D__CUDA_NO_HALF_OPERATORS__ -D__CUDA_NO_HALF_CONVERSIONS__ -D__CUDA_NO_HALF2_OPERATORS__ --expt-relaxed-constexpr --compiler-options ''"'"'-fPIC'"'"'' -D__CUDA_NO_HALF_OPERATORS__ -D__CUDA_NO_HALF_CONVERSIONS__ -D__CUDA_NO_HALF2_OPERATORS__ -DTORCH_API_INCLUDE_EXTENSION_H -DTORCH_EXTENSION_NAME=gpu_nms -D_GLIBCXX_USE_CXX11_ABI=0 -gencode=arch=compute_86,code=sm_86 -std=c++14
nvcc fatal   : Unsupported gpu architecture 'compute_86'

I think one reason could be that the CUDA version I get with nvcc -V and ./deviceQuery are different

nvcc -V output

nvcc: NVIDIA (R) Cuda compiler driver
Copyright (c) 2005-2020 NVIDIA Corporation
Built on Wed_Jul_22_19:09:09_PDT_2020
Cuda compilation tools, release 11.0, V11.0.221
Build cuda_11.0_bu.TC445_37.28845127_0

./deviceQuery output

./deviceQuery Starting...

 CUDA Device Query (Runtime API) version (CUDART static linking)

Detected 2 CUDA Capable device(s)

Device 0: "GeForce RTX 3090"
  CUDA Driver Version / Runtime Version          11.2 / 10.0
  CUDA Capability Major/Minor version number:    8.6
  Total amount of global memory:                 24268 MBytes (25447170048 bytes)
MapSMtoCores for SM 8.6 is undefined.  Default to use 64 Cores/SM
MapSMtoCores for SM 8.6 is undefined.  Default to use 64 Cores/SM
  (82) Multiprocessors, ( 64) CUDA Cores/MP:     5248 CUDA Cores
  GPU Max Clock rate:                            1800 MHz (1.80 GHz)
  Memory Clock rate:                             9751 Mhz
  Memory Bus Width:                              384-bit
  L2 Cache Size:                                 6291456 bytes
  Maximum Texture Dimension Size (x,y,z)         1D=(131072), 2D=(131072, 65536), 3D=(16384, 16384, 16384)
  Maximum Layered 1D Texture Size, (num) layers  1D=(32768), 2048 layers
  Maximum Layered 2D Texture Size, (num) layers  2D=(32768, 32768), 2048 layers
  Total amount of constant memory:               65536 bytes
  Total amount of shared memory per block:       49152 bytes
  Total number of registers available per block: 65536
  Warp size:                                     32
  Maximum number of threads per multiprocessor:  1536
  Maximum number of threads per block:           1024
  Max dimension size of a thread block (x,y,z): (1024, 1024, 64)
  Max dimension size of a grid size    (x,y,z): (2147483647, 65535, 65535)
  Maximum memory pitch:                          2147483647 bytes
  Texture alignment:                             512 bytes
  Concurrent copy and kernel execution:          Yes with 2 copy engine(s)
  Run time limit on kernels:                     Yes
  Integrated GPU sharing Host Memory:            No
  Support host page-locked memory mapping:       Yes
  Alignment requirement for Surfaces:            Yes
  Device has ECC support:                        Disabled
  Device supports Unified Addressing (UVA):      Yes
  Device supports Compute Preemption:            Yes
  Supports Cooperative Kernel Launch:            Yes
  Supports MultiDevice Co-op Kernel Launch:      Yes
  Device PCI Domain ID / Bus ID / location ID:   0 / 161 / 0
  Compute Mode:
     < Default (multiple host threads can use ::cudaSetDevice() with device simultaneously) >

Device 1: "GeForce RTX 3090"
  CUDA Driver Version / Runtime Version          11.2 / 10.0
  CUDA Capability Major/Minor version number:    8.6
  Total amount of global memory:                 24265 MBytes (25443893248 bytes)
MapSMtoCores for SM 8.6 is undefined.  Default to use 64 Cores/SM
MapSMtoCores for SM 8.6 is undefined.  Default to use 64 Cores/SM
  (82) Multiprocessors, ( 64) CUDA Cores/MP:     5248 CUDA Cores
  GPU Max Clock rate:                            1800 MHz (1.80 GHz)
  Memory Clock rate:                             9751 Mhz
  Memory Bus Width:                              384-bit
  L2 Cache Size:                                 6291456 bytes
  Maximum Texture Dimension Size (x,y,z)         1D=(131072), 2D=(131072, 65536), 3D=(16384, 16384, 16384)
  Maximum Layered 1D Texture Size, (num) layers  1D=(32768), 2048 layers
  Maximum Layered 2D Texture Size, (num) layers  2D=(32768, 32768), 2048 layers
  Total amount of constant memory:               65536 bytes
  Total amount of shared memory per block:       49152 bytes
  Total number of registers available per block: 65536
  Warp size:                                     32
  Maximum number of threads per multiprocessor:  1536
  Maximum number of threads per block:           1024
  Max dimension size of a thread block (x,y,z): (1024, 1024, 64)
  Max dimension size of a grid size    (x,y,z): (2147483647, 65535, 65535)
  Maximum memory pitch:                          2147483647 bytes
  Texture alignment:                             512 bytes
  Concurrent copy and kernel execution:          Yes with 2 copy engine(s)
  Run time limit on kernels:                     Yes
  Integrated GPU sharing Host Memory:            No
  Support host page-locked memory mapping:       Yes
  Alignment requirement for Surfaces:            Yes
  Device has ECC support:                        Disabled
  Device supports Unified Addressing (UVA):      Yes
  Device supports Compute Preemption:            Yes
  Supports Cooperative Kernel Launch:            Yes
  Supports MultiDevice Co-op Kernel Launch:      Yes
  Device PCI Domain ID / Bus ID / location ID:   0 / 193 / 0
  Compute Mode:
     < Default (multiple host threads can use ::cudaSetDevice() with device simultaneously) >
> Peer access from GeForce RTX 3090 (GPU0) -> GeForce RTX 3090 (GPU1) : No
> Peer access from GeForce RTX 3090 (GPU1) -> GeForce RTX 3090 (GPU0) : No

deviceQuery, CUDA Driver = CUDART, CUDA Driver Version = 11.2, CUDA Runtime Version = 10.0, NumDevs = 2
Result = PASS

Can this be fixed?
Thank you very much

generix · February 21, 2021, 8:09pm

You also need to uninstall cuda 10 and install cuda 11 instead.

flyingalpaca · February 21, 2021, 8:15pm

Hello,

I uninstalled CUDA 10.0 using sudo apt-get --purge remove cuda-10.0 before I installed cuda-11.0 so I’m quite confused as to why it still says it’s there.
I’ve also updated the path setup to 11.0
After uninstalling, I also checked that there is no longer the cuda-10.0 file in the /usr/local/ directory

Is there something I missed during the uninstalling or installing process?

Thank you very much

generix · February 21, 2021, 8:18pm

That’s odd indeed, nvcc reports the correct version. Please post the output of
dpkg -l |grep cuda

flyingalpaca · February 21, 2021, 8:21pm

ii  cuda-11-0                                  11.0.3-1                                         amd64        CUDA 11.0 meta-package
ii  cuda-command-line-tools-11-0               11.0.3-1                                         amd64        CUDA command-line tools
ii  cuda-compiler-11-0                         11.0.3-1                                         amd64        CUDA compiler
rc  cuda-cublas-10-0                           10.0.130-1                                       amd64        CUBLAS native runtime libraries
rc  cuda-cudart-10-0                           10.0.130-1                                       amd64        CUDA Runtime native Libraries
ii  cuda-cudart-11-0                           11.0.221-1                                       amd64        CUDA Runtime native Libraries
rc  cuda-cudart-dev-10-0                       10.0.130-1                                       amd64        CUDA Runtime native dev links, headers
ii  cuda-cudart-dev-11-0                       11.0.221-1                                       amd64        CUDA Runtime native dev links, headers
rc  cuda-cufft-10-0                            10.0.130-1                                       amd64        CUFFT native runtime libraries
ii  cuda-cuobjdump-11-0                        11.0.221-1                                       amd64        CUDA cuobjdump
rc  cuda-cupti-10-0                            10.0.130-1                                       amd64        CUDA profiling tools interface.
ii  cuda-cupti-11-0                            11.0.221-1                                       amd64        CUDA profiling tools runtime libs.
ii  cuda-cupti-dev-11-0                        11.0.221-1                                       amd64        CUDA profiling tools interface.
rc  cuda-curand-10-0                           10.0.130-1                                       amd64        CURAND native runtime libraries
rc  cuda-cusolver-10-0                         10.0.130-1                                       amd64        CUDA solver native runtime libraries
rc  cuda-cusparse-10-0                         10.0.130-1                                       amd64        CUSPARSE native runtime libraries
ii  cuda-demo-suite-11-0                       11.0.167-1                                       amd64        Demo suite for CUDA
ii  cuda-documentation-11-0                    11.0.228-1                                       amd64        CUDA documentation
ii  cuda-driver-dev-11-0                       11.0.221-1                                       amd64        CUDA Driver native dev stub library
ii  cuda-drivers                               460.32.03-1                                      amd64        CUDA Driver meta-package, branch-agnostic
ii  cuda-drivers-460                           460.32.03-1                                      amd64        CUDA Driver meta-package, branch-specific
ii  cuda-gdb-11-0                              11.0.221-1                                       amd64        CUDA-GDB
ii  cuda-libraries-11-0                        11.0.3-1                                         amd64        CUDA Libraries 11.0 meta-package
ii  cuda-libraries-dev-11-0                    11.0.3-1                                         amd64        CUDA Libraries 11.0 development meta-package
ii  cuda-memcheck-11-0                         11.0.221-1                                       amd64        CUDA-MEMCHECK
rc  cuda-npp-10-0                              10.0.130-1                                       amd64        NPP native runtime libraries
ii  cuda-nsight-11-0                           11.0.221-1                                       amd64        CUDA nsight
rc  cuda-nsight-compute-10-0                   10.0.130-1                                       amd64        NVIDIA Nsight Compute
ii  cuda-nsight-compute-11-0                   11.0.3-1                                         amd64        NVIDIA Nsight Compute
ii  cuda-nsight-systems-11-0                   11.0.3-1                                         amd64        NVIDIA Nsight Systems
rc  cuda-nvcc-10-0                             10.0.130-1                                       amd64        CUDA nvcc
ii  cuda-nvcc-11-0                             11.0.221-1                                       amd64        CUDA nvcc
ii  cuda-nvdisasm-11-0                         11.0.221-1                                       amd64        CUDA disassembler
rc  cuda-nvgraph-10-0                          10.0.130-1                                       amd64        NVGRAPH native runtime libraries
ii  cuda-nvml-dev-11-0                         11.0.167-1                                       amd64        NVML native dev links, headers
rc  cuda-nvprof-10-0                           10.0.130-1                                       amd64        CUDA Profiler tools
ii  cuda-nvprof-11-0                           11.0.221-1                                       amd64        CUDA Profiler tools
ii  cuda-nvprune-11-0                          11.0.221-1                                       amd64        CUDA nvprune
rc  cuda-nvrtc-10-0                            10.0.130-1                                       amd64        NVRTC native runtime libraries
ii  cuda-nvrtc-11-0                            11.0.221-1                                       amd64        NVRTC native runtime libraries
ii  cuda-nvrtc-dev-11-0                        11.0.221-1                                       amd64        NVRTC native dev links, headers
rc  cuda-nvtx-10-0                             10.0.130-1                                       amd64        NVIDIA Tools Extension
ii  cuda-nvtx-11-0                             11.0.167-1                                       amd64        NVIDIA Tools Extension
ii  cuda-nvvp-11-0                             11.0.221-1                                       amd64        CUDA Profiler tools
ii  cuda-repo-ubuntu1804                       10.2.89-1                                        amd64        cuda repository configuration files
ii  cuda-runtime-11-0                          11.0.3-1                                         amd64        CUDA Runtime 11.0 meta-package
ii  cuda-samples-11-0                          11.0.221-1                                       amd64        CUDA example applications
ii  cuda-sanitizer-11-0                        11.0.221-1                                       amd64        CUDA Sanitizer
rc  cuda-toolkit-10-0                          10.0.130-1                                       amd64        CUDA Toolkit 10.0 meta-package
ii  cuda-toolkit-11-0                          11.0.3-1                                         amd64        CUDA Toolkit 11.0 meta-package
ii  cuda-tools-11-0                            11.0.3-1                                         amd64        CUDA Tools meta-package
rc  cuda-visual-tools-10-0                     10.0.130-1                                       amd64        CUDA visual tools
ii  cuda-visual-tools-11-0                     11.0.3-1                                         amd64        CUDA visual tools

generix · February 21, 2021, 8:23pm

Seems you only removed the meta-package, leaving the sub-package installed. Please try

sudo apt remove cuda*10-0
sudo apt remove libcuda*10-0

flyingalpaca · February 21, 2021, 8:28pm

I tried running those lines but it didn’t seem to have removed anything

sudo apt remove cuda*10-0 outputted the following:

Reading package lists... Done
Building dependency tree       
Reading state information... Done
Note, selecting 'cuda-cusparse-cross-qnx-10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-npp-cross-qnx-10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-nvprune-10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-nvgraph-10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-cublas-dev-10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-gdb-10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-nsight-compute--10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-visual-tools-10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-curand-dev-10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-cupti-10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-curand-cross-qnx-10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-libraries-dev-10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-cublas-cross-qnx-10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-nvjpeg-10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-nvrtc-dev-10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-minimal-build-10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-cufft-cross-qnx-10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-libraries-10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-runtime-10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-cusolver-10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-cusparse-dev-10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-nvrtc-10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-nvjpeg-dev-10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-cufft-10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-cusparse-10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-npp-10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-tools-10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-nvtx-10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-cudart-10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-license-10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-nvdisasm-10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-samples-10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-documentation-10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-cusolver-cross-qnx-10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-compat-10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-nvcc-10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-nvvp-10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-cross-qnx-10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-toolkit-10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-nvgraph-dev-10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-cudart-dev-10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-nvml-dev-10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-gpu-library-advisor-10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-cublas-10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-core-10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-cusolver-dev-10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-driver-cross-qnx-10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-compiler-10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-demo-suite-10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-npp-dev-10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-nsight-10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-curand-10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-command-line-tools-10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-cufft-dev-10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-driver-dev-10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-misc-headers-cross-qnx-10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-memcheck-10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-cudart-cross-qnx-10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-gdb-src-10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-misc-headers-10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-cuobjdump-10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-nsight-compute-10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-nvprof-10-0' for glob 'cuda*10-0'
Package 'cuda-cudart-cross-qnx-10-0' is not installed, so not removed
Package 'cuda-driver-cross-qnx-10-0' is not installed, so not removed
Package 'cuda-misc-headers-cross-qnx-10-0' is not installed, so not removed
Package 'cuda-cusolver-cross-qnx-10-0' is not installed, so not removed
Package 'cuda-cublas-cross-qnx-10-0' is not installed, so not removed
Package 'cuda-cufft-cross-qnx-10-0' is not installed, so not removed
Package 'cuda-curand-cross-qnx-10-0' is not installed, so not removed
Package 'cuda-cusparse-cross-qnx-10-0' is not installed, so not removed
Package 'cuda-npp-cross-qnx-10-0' is not installed, so not removed
Package 'cuda-10-0' is not installed, so not removed
Package 'cuda-command-line-tools-10-0' is not installed, so not removed
Package 'cuda-compat-10-0' is not installed, so not removed
Package 'cuda-compiler-10-0' is not installed, so not removed
Package 'cuda-core-10-0' is not installed, so not removed
Package 'cuda-cross-qnx-10-0' is not installed, so not removed
Package 'cuda-cublas-10-0' is not installed, so not removed
Package 'cuda-cublas-dev-10-0' is not installed, so not removed
Package 'cuda-cudart-10-0' is not installed, so not removed
Package 'cuda-cudart-dev-10-0' is not installed, so not removed
Package 'cuda-cufft-10-0' is not installed, so not removed
Package 'cuda-cufft-dev-10-0' is not installed, so not removed
Package 'cuda-cuobjdump-10-0' is not installed, so not removed
Package 'cuda-cupti-10-0' is not installed, so not removed
Package 'cuda-curand-10-0' is not installed, so not removed
Package 'cuda-curand-dev-10-0' is not installed, so not removed
Package 'cuda-cusolver-10-0' is not installed, so not removed
Package 'cuda-cusolver-dev-10-0' is not installed, so not removed
Package 'cuda-cusparse-10-0' is not installed, so not removed
Package 'cuda-cusparse-dev-10-0' is not installed, so not removed
Package 'cuda-demo-suite-10-0' is not installed, so not removed
Package 'cuda-documentation-10-0' is not installed, so not removed
Package 'cuda-driver-dev-10-0' is not installed, so not removed
Package 'cuda-gdb-10-0' is not installed, so not removed
Package 'cuda-gdb-src-10-0' is not installed, so not removed
Package 'cuda-gpu-library-advisor-10-0' is not installed, so not removed
Package 'cuda-libraries-10-0' is not installed, so not removed
Package 'cuda-libraries-dev-10-0' is not installed, so not removed
Package 'cuda-license-10-0' is not installed, so not removed
Package 'cuda-memcheck-10-0' is not installed, so not removed
Package 'cuda-minimal-build-10-0' is not installed, so not removed
Package 'cuda-misc-headers-10-0' is not installed, so not removed
Package 'cuda-npp-10-0' is not installed, so not removed
Package 'cuda-npp-dev-10-0' is not installed, so not removed
Package 'cuda-nsight-10-0' is not installed, so not removed
Package 'cuda-nsight-compute--10-0' is not installed, so not removed
Package 'cuda-nsight-compute-10-0' is not installed, so not removed
Package 'cuda-nvcc-10-0' is not installed, so not removed
Package 'cuda-nvdisasm-10-0' is not installed, so not removed
Package 'cuda-nvgraph-10-0' is not installed, so not removed
Package 'cuda-nvgraph-dev-10-0' is not installed, so not removed
Package 'cuda-nvjpeg-10-0' is not installed, so not removed
Package 'cuda-nvjpeg-dev-10-0' is not installed, so not removed
Package 'cuda-nvml-dev-10-0' is not installed, so not removed
Package 'cuda-nvprof-10-0' is not installed, so not removed
Package 'cuda-nvprune-10-0' is not installed, so not removed
Package 'cuda-nvrtc-10-0' is not installed, so not removed
Package 'cuda-nvrtc-dev-10-0' is not installed, so not removed
Package 'cuda-nvtx-10-0' is not installed, so not removed
Package 'cuda-nvvp-10-0' is not installed, so not removed
Package 'cuda-runtime-10-0' is not installed, so not removed
Package 'cuda-samples-10-0' is not installed, so not removed
Package 'cuda-toolkit-10-0' is not installed, so not removed
Package 'cuda-tools-10-0' is not installed, so not removed
Package 'cuda-visual-tools-10-0' is not installed, so not removed
0 upgraded, 0 newly installed, 0 to remove and 0 not upgraded.

and sudo apt remove libcuda*10-0 outputted the following:

Reading package lists... Done
Building dependency tree       
Reading state information... Done
E: Unable to locate package libcuda*10-0
E: Couldn't find any package by glob 'libcuda*10-0'
E: Couldn't find any package by regex 'libcuda*10-0'

generix · February 21, 2021, 8:39pm

Does
sudo apt autoremove
remove the cuda 10 packages?

flyingalpaca · February 21, 2021, 8:47pm

Nope, it just outputs

Building dependency tree       
Reading state information... Done
0 upgraded, 0 newly installed, 0 to remove and 0 not upgraded.

generix · February 21, 2021, 9:08pm

Please try

sudo apt remove cuda*10-0 libcublas*10*

flyingalpaca · February 21, 2021, 9:24pm

sudo apt remove cuda*10-0 libcublas*10*outputs:

Reading package lists... Done
Building dependency tree       
Reading state information... Done
Note, selecting 'cuda-cusparse-cross-qnx-10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-npp-cross-qnx-10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-nvprune-10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-nvgraph-10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-cublas-dev-10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-gdb-10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-nsight-compute--10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-visual-tools-10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-curand-dev-10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-cupti-10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-curand-cross-qnx-10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-libraries-dev-10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-cublas-cross-qnx-10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-nvjpeg-10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-nvrtc-dev-10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-minimal-build-10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-cufft-cross-qnx-10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-libraries-10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-runtime-10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-cusolver-10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-cusparse-dev-10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-nvrtc-10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-nvjpeg-dev-10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-cufft-10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-cusparse-10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-npp-10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-tools-10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-nvtx-10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-cudart-10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-license-10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-nvdisasm-10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-samples-10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-documentation-10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-cusolver-cross-qnx-10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-compat-10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-nvcc-10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-nvvp-10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-cross-qnx-10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-toolkit-10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-nvgraph-dev-10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-cudart-dev-10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-nvml-dev-10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-gpu-library-advisor-10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-cublas-10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-core-10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-cusolver-dev-10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-driver-cross-qnx-10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-compiler-10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-demo-suite-10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-npp-dev-10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-nsight-10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-curand-10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-command-line-tools-10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-cufft-dev-10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-driver-dev-10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-misc-headers-cross-qnx-10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-memcheck-10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-cudart-cross-qnx-10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-gdb-src-10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-misc-headers-10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-cuobjdump-10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-nsight-compute-10-0' for glob 'cuda*10-0'
Note, selecting 'cuda-nvprof-10-0' for glob 'cuda*10-0'
Package 'cuda-cudart-cross-qnx-10-0' is not installed, so not removed
Package 'cuda-driver-cross-qnx-10-0' is not installed, so not removed
Package 'cuda-misc-headers-cross-qnx-10-0' is not installed, so not removed
Package 'cuda-cusolver-cross-qnx-10-0' is not installed, so not removed
Package 'cuda-cublas-cross-qnx-10-0' is not installed, so not removed
Package 'cuda-cufft-cross-qnx-10-0' is not installed, so not removed
Package 'cuda-curand-cross-qnx-10-0' is not installed, so not removed
Package 'cuda-cusparse-cross-qnx-10-0' is not installed, so not removed
Package 'cuda-npp-cross-qnx-10-0' is not installed, so not removed
Note, selecting 'libcublas10' for glob 'libcublas*10*'
Package 'cuda-10-0' is not installed, so not removed
Package 'cuda-command-line-tools-10-0' is not installed, so not removed
Package 'cuda-compat-10-0' is not installed, so not removed
Package 'cuda-compiler-10-0' is not installed, so not removed
Package 'cuda-core-10-0' is not installed, so not removed
Package 'cuda-cross-qnx-10-0' is not installed, so not removed
Package 'cuda-cublas-10-0' is not installed, so not removed
Package 'cuda-cublas-dev-10-0' is not installed, so not removed
Package 'cuda-cudart-10-0' is not installed, so not removed
Package 'cuda-cudart-dev-10-0' is not installed, so not removed
Package 'cuda-cufft-10-0' is not installed, so not removed
Package 'cuda-cufft-dev-10-0' is not installed, so not removed
Package 'cuda-cuobjdump-10-0' is not installed, so not removed
Package 'cuda-cupti-10-0' is not installed, so not removed
Package 'cuda-curand-10-0' is not installed, so not removed
Package 'cuda-curand-dev-10-0' is not installed, so not removed
Package 'cuda-cusolver-10-0' is not installed, so not removed
Package 'cuda-cusolver-dev-10-0' is not installed, so not removed
Package 'cuda-cusparse-10-0' is not installed, so not removed
Package 'cuda-cusparse-dev-10-0' is not installed, so not removed
Package 'cuda-demo-suite-10-0' is not installed, so not removed
Package 'cuda-documentation-10-0' is not installed, so not removed
Package 'cuda-driver-dev-10-0' is not installed, so not removed
Package 'cuda-gdb-10-0' is not installed, so not removed
Package 'cuda-gdb-src-10-0' is not installed, so not removed
Package 'cuda-gpu-library-advisor-10-0' is not installed, so not removed
Package 'cuda-libraries-10-0' is not installed, so not removed
Package 'cuda-libraries-dev-10-0' is not installed, so not removed
Package 'cuda-license-10-0' is not installed, so not removed
Package 'cuda-memcheck-10-0' is not installed, so not removed
Package 'cuda-minimal-build-10-0' is not installed, so not removed
Package 'cuda-misc-headers-10-0' is not installed, so not removed
Package 'cuda-npp-10-0' is not installed, so not removed
Package 'cuda-npp-dev-10-0' is not installed, so not removed
Package 'cuda-nsight-10-0' is not installed, so not removed
Package 'cuda-nsight-compute--10-0' is not installed, so not removed
Package 'cuda-nsight-compute-10-0' is not installed, so not removed
Package 'cuda-nvcc-10-0' is not installed, so not removed
Package 'cuda-nvdisasm-10-0' is not installed, so not removed
Package 'cuda-nvgraph-10-0' is not installed, so not removed
Package 'cuda-nvgraph-dev-10-0' is not installed, so not removed
Package 'cuda-nvjpeg-10-0' is not installed, so not removed
Package 'cuda-nvjpeg-dev-10-0' is not installed, so not removed
Package 'cuda-nvml-dev-10-0' is not installed, so not removed
Package 'cuda-nvprof-10-0' is not installed, so not removed
Package 'cuda-nvprune-10-0' is not installed, so not removed
Package 'cuda-nvrtc-10-0' is not installed, so not removed
Package 'cuda-nvrtc-dev-10-0' is not installed, so not removed
Package 'cuda-nvtx-10-0' is not installed, so not removed
Package 'cuda-nvvp-10-0' is not installed, so not removed
Package 'cuda-runtime-10-0' is not installed, so not removed
Package 'cuda-samples-10-0' is not installed, so not removed
Package 'cuda-toolkit-10-0' is not installed, so not removed
Package 'cuda-tools-10-0' is not installed, so not removed
Package 'cuda-visual-tools-10-0' is not installed, so not removed
Package 'libcublas10' is not installed, so not removed
0 upgraded, 0 newly installed, 0 to remove and 0 not upgraded.

generix · February 21, 2021, 9:45pm

No idea what that crappy apt doesn’t like. Please try using aptitude to select the packages one-by-one.

flyingalpaca · February 21, 2021, 9:51pm

Do I just remove all the 10.0 packages? Or are there specific packages I must keep?

generix · February 21, 2021, 9:55pm

Just remove all 10.x versions. Maybe synaptics instead of aptitude is more comfortable if you’re on a desktop.