Could NOT find CUDA (missing: CUDA_CUDART_LIBRARY) (found version "10.2")

Robotics & Edge Computing Jetson & Embedded Systems Jetson Nano
1

Hi, Team !
I got such error:

-- Could NOT find CUDA (missing: CUDA_CUDART_LIBRARY) (found version "10.2")
CMake Warning at /usr/local/lib/python3.6/dist-packages/torch/share/cmake/Caffe2/public/cuda.cmake:31 (message):
  Caffe2: CUDA cannot be found.  Depending on whether you are building Caffe2
  or a Caffe2 dependent library, the next warning / error will give you more
  info.
Call Stack (most recent call first):
  /usr/local/lib/python3.6/dist-packages/torch/share/cmake/Caffe2/Caffe2Config.cmake:88 (include)
  /usr/local/lib/python3.6/dist-packages/torch/share/cmake/Torch/TorchConfig.cmake:68 (find_package)
  CMakeLists.txt:18 (find_package)


CMake Error at /usr/local/lib/python3.6/dist-packages/torch/share/cmake/Caffe2/Caffe2Config.cmake:90 (message):
  Your installed Caffe2 version uses CUDA but I cannot find the CUDA
  libraries.  Please set the proper CUDA prefixes and / or install CUDA.
Call Stack (most recent call first):
  /usr/local/lib/python3.6/dist-packages/torch/share/cmake/Torch/TorchConfig.cmake:68 (find_package)
  CMakeLists.txt:18 (find_package)


-- Configuring incomplete, errors occurred!
The command '/bin/sh -c git clone https://github.com/pytorch/vision.git --branch v0.15.2 &&      cd vision &&     mkdir build &&     cd build &&     cmake .. &&     make -j${BUILD_CORES} &&     make install &&     cd .. && rm -rf build' returned a non-zero code: 1

I am trying to build docker container. All log I will add below this sentence.

docker build -t jetson-build:latest -f Jetson.Dockerfile --build-arg BUILD_CORES=2 .
Sending build context to Docker daemon  78.25MB
Step 1/18 : ARG BASE_IMAGE=nvcr.io/nvidia/l4t-pytorch:r32.6.1-pth1.9-py3
Step 2/18 : FROM $BASE_IMAGE as base
 ---> 9a936a937ee0
Step 3/18 : RUN apt-get update && apt-get upgrade -yy && apt-get install -yy     wget     git     build-essential     git     clang-3.8     libpng-dev &&     apt remove cmake -y &&     pip3 install --upgrade pip &&     pip3 install cmake --upgrade
 ---> Using cache
 ---> 6b4b9c1e48fe
Step 4/18 : ENV GOLANG_VERSION=1.21.0     GOROOT=/usr/local/go     PATH=$PATH:/usr/local/go/bin
 ---> Using cache
 ---> 94016a7375bc
Step 5/18 : RUN wget https://golang.org/dl/go${GOLANG_VERSION}.linux-arm64.tar.gz &&     tar -C /usr/local -xzf go${GOLANG_VERSION}.linux-arm64.tar.gz &&     rm go${GOLANG_VERSION}.linux-arm64.tar.gz
 ---> Using cache
 ---> 35bced8d7b56
Step 6/18 : ARG PYTHON_VERSION=3.6
 ---> Using cache
 ---> 04c2e56572e8
Step 7/18 : ENV LIBTORCH_PATH=/usr/local/lib/python${PYTHON_VERSION}/dist-packages/torch
 ---> Using cache
 ---> 98639e576c6d
Step 8/18 : ENV LD_LIBRARY_PATH=$LD_LIBRARY_PATH:$LIBTORCH_PATH/lib:/usr/local/lib:/app/src/libs     Torch_DIR=$LIBTORCH_PATH
 ---> Using cache
 ---> c73573681888
Step 9/18 : RUN ln -s ${LIBTORCH_PATH}/include/* /usr/local/include/ &&     ln -s ${LIBTORCH_PATH}/lib/* /usr/local/lib/
 ---> Using cache
 ---> 19d49516a9ee
Step 10/18 : WORKDIR /app
 ---> Using cache
 ---> 7c48a054db6d
Step 11/18 : COPY go.mod go.sum ./
 ---> Using cache
 ---> 1f99a86eced9
Step 12/18 : RUN go mod download
 ---> Using cache
 ---> 755acffaff1f
Step 13/18 : ARG BUILD_CORES=2
 ---> Using cache
 ---> dff8f2c603fb
Step 14/18 : RUN cd /root/go/pkg/mod/github.com/mishabeliy15/gotorch*/cgotorch &&     chmod +x build.sh &&     ./build.sh -j${BUILD_CORES}
 ---> Using cache
 ---> 6698c92ad13f
Step 15/18 : RUN git clone https://github.com/pytorch/vision.git --branch v0.15.2 &&      cd vision &&     mkdir build &&     cd build &&     cmake .. &&     make -j${BUILD_CORES} &&     make install &&     cd .. && rm -rf build
 ---> Running in 9900e3a02d7d
Cloning into 'vision'...
Note: checking out 'fa99a5360fbcd1683311d57a76fcc0e7323a4c1e'.

You are in 'detached HEAD' state. You can look around, make experimental
changes and commit them, and you can discard any commits you make in this
state without impacting any branches by performing another checkout.

If you want to create a new branch to retain commits you create, you may
do so (now or later) by using -b with the checkout command again. Example:

  git checkout -b <new-branch-name>

-- The C compiler identification is GNU 7.5.0
-- The CXX compiler identification is GNU 7.5.0
-- Detecting C compiler ABI info
-- Detecting C compiler ABI info - done
-- Check for working C compiler: /usr/bin/cc - skipped
-- Detecting C compile features
-- Detecting C compile features - done
-- Detecting CXX compiler ABI info
-- Detecting CXX compiler ABI info - done
-- Check for working CXX compiler: /usr/bin/c++ - skipped
-- Detecting CXX compile features
-- Detecting CXX compile features - done
-- Performing Test CMAKE_HAVE_LIBC_PTHREAD
-- Performing Test CMAKE_HAVE_LIBC_PTHREAD - Failed
-- Looking for pthread_create in pthreads
-- Looking for pthread_create in pthreads - not found
-- Looking for pthread_create in pthread
-- Looking for pthread_create in pthread - found
-- Found Threads: TRUE  
-- Could NOT find CUDA (missing: CUDA_CUDART_LIBRARY) (found version "10.2")
CMake Warning at /usr/local/lib/python3.6/dist-packages/torch/share/cmake/Caffe2/public/cuda.cmake:31 (message):
  Caffe2: CUDA cannot be found.  Depending on whether you are building Caffe2
  or a Caffe2 dependent library, the next warning / error will give you more
  info.
Call Stack (most recent call first):
  /usr/local/lib/python3.6/dist-packages/torch/share/cmake/Caffe2/Caffe2Config.cmake:88 (include)
  /usr/local/lib/python3.6/dist-packages/torch/share/cmake/Torch/TorchConfig.cmake:68 (find_package)
  CMakeLists.txt:18 (find_package)


CMake Error at /usr/local/lib/python3.6/dist-packages/torch/share/cmake/Caffe2/Caffe2Config.cmake:90 (message):
  Your installed Caffe2 version uses CUDA but I cannot find the CUDA
  libraries.  Please set the proper CUDA prefixes and / or install CUDA.
Call Stack (most recent call first):
  /usr/local/lib/python3.6/dist-packages/torch/share/cmake/Torch/TorchConfig.cmake:68 (find_package)
  CMakeLists.txt:18 (find_package)


-- Configuring incomplete, errors occurred!
The command '/bin/sh -c git clone https://github.com/pytorch/vision.git --branch v0.15.2 &&      cd vision &&     mkdir build &&     cd build &&     cmake .. &&     make -j${BUILD_CORES} &&     make install &&     cd .. && rm -rf build' returned a non-zero code: 1

Full Docker file:

ARG BASE_IMAGE=nvcr.io/nvidia/l4t-pytorch:r32.6.1-pth1.9-py3

FROM $BASE_IMAGE as base

RUN apt-get update && apt-get upgrade -yy && apt-get install -yy \
    wget \
    git \
    build-essential \
    git \
    clang-3.8 \
    libpng-dev && \
    apt remove cmake -y && \
    pip3 install --upgrade pip && \
    pip3 install cmake --upgrade
#    libgeos-dev \ ToDo: need to install later
#    libproj-dev


# Donwload and install Go
ENV GOLANG_VERSION=1.21.0 \
    GOROOT=/usr/local/go \
    PATH=$PATH:/usr/local/go/bin

RUN wget https://golang.org/dl/go${GOLANG_VERSION}.linux-arm64.tar.gz && \
    tar -C /usr/local -xzf go${GOLANG_VERSION}.linux-arm64.tar.gz && \
    rm go${GOLANG_VERSION}.linux-arm64.tar.gz

# Set ENV var for libtorch
ARG PYTHON_VERSION=3.6
ENV LIBTORCH_PATH=/usr/local/lib/python${PYTHON_VERSION}/dist-packages/torch \
    CUDA_HOME=/usr/local/cuda-10.2
ENV LD_LIBRARY_PATH=$LD_LIBRARY_PATH:$CUDA_HOME/lib64:$LIBTORCH_PATH/lib:/usr/local/lib:/app/src/libs \
    Torch_DIR=$LIBTORCH_PATH \
    PATH=/usr/local/cuda-10.2/bin:$PATH \
    CUDA_BIN_PATH=/usr/local/cuda-10.2/bin

# Create symlinks libtorch to system
RUN ln -s ${LIBTORCH_PATH}/include/* /usr/local/include/ && \
    ln -s ${LIBTORCH_PATH}/lib/* /usr/local/lib/ && \
    ln -s /usr/local/cuda-10.2 /usr/local/cuda


WORKDIR /app

COPY go.mod go.sum ./
RUN go mod download

ARG BUILD_CORES=2
# Build gotorch
RUN cd /root/go/pkg/mod/github.com/mishabeliy15/gotorch*/cgotorch && \
    chmod +x build.sh && \
    ./build.sh -j${BUILD_CORES}

# Download and build libtorchvision
RUN git clone https://github.com/pytorch/vision.git --branch v0.15.2 &&  \
    cd vision && \
    mkdir build && \
    cd build && \
    cmake .. && \
    make -j${BUILD_CORES} && \
    make install && \
    cd .. && rm -rf build

FROM base as dev

WORKDIR /app/src
CMD ["bash"]

Torch version:

pip3 show torch
Name: torch
Version: 1.13.0a0+git7c98e70
Summary: Tensors and Dynamic neural networks in Python with strong GPU acceleration
Home-page: https://pytorch.org/
Author: PyTorch Team
Author-email: packages@pytorch.org
License: BSD-3
Location: /usr/local/lib/python3.8/dist-packages
Requires: typing-extensions
Required-by: torchaudio, torchvision

Cuda version:

nvcc --version
nvcc: NVIDIA (R) Cuda compiler driver
Copyright (c) 2005-2021 NVIDIA Corporation
Built on Sun_Feb_28_22:34:44_PST_2021
Cuda compilation tools, release 10.2, V10.2.300
Build cuda_10.2_r440.TC440_70.29663091_0
3

Hi @s.pinchuk, have you set your default docker runtime to ‘nvidia’? See here: https://github.com/dusty-nv/jetson-containers/blob/master/docs/setup.md#docker-default-runtime

If you’re on JetPack 4 and need the CUDA/cuDNN/TensorRT headers/libraries while building container, you need to set that so that the NVIDIA Container Runtime gets used during docker build operations. This way, CUDA/cuDNN/TensorRT will be mounted into your build context (on JetPack 4, those CUDA libraries get mounted from the host)

1 Like
5

I set default runtime for NVIDIA:

sudo docker info | grep 'Default Runtime'
 Default Runtime: nvidia

But I have already got the same error:

make build-jetson-docker 
docker build -t jetson-build:latest -f Jetson.Dockerfile .
Sending build context to Docker daemon  78.25MB
Step 1/18 : ARG BASE_IMAGE=nvcr.io/nvidia/l4t-pytorch:r32.6.1-pth1.9-py3
Step 2/18 : FROM $BASE_IMAGE as base
 ---> 9a936a937ee0
Step 3/18 : RUN apt-get update && apt-get upgrade -yy && apt-get install -yy     wget     git     build-essential     git     clang-3.8     libpng-dev &&     apt remove cmake -y &&     pip3 install --upgrade pip &&     pip3 install cmake --upgrade
 ---> Using cache
 ---> 6b4b9c1e48fe
Step 4/18 : ENV GOLANG_VERSION=1.21.0     GOROOT=/usr/local/go     PATH=$PATH:/usr/local/go/bin
 ---> Using cache
 ---> 94016a7375bc
Step 5/18 : RUN wget https://golang.org/dl/go${GOLANG_VERSION}.linux-arm64.tar.gz &&     tar -C /usr/local -xzf go${GOLANG_VERSION}.linux-arm64.tar.gz &&     rm go${GOLANG_VERSION}.linux-arm64.tar.gz
 ---> Using cache
 ---> 35bced8d7b56
Step 6/18 : ARG PYTHON_VERSION=3.6
 ---> Using cache
 ---> 04c2e56572e8
Step 7/18 : ENV LIBTORCH_PATH=/usr/local/lib/python${PYTHON_VERSION}/dist-packages/torch
 ---> Using cache
 ---> 98639e576c6d
Step 8/18 : ENV LD_LIBRARY_PATH=$LD_LIBRARY_PATH:$LIBTORCH_PATH/lib:/usr/local/lib:/app/src/libs     Torch_DIR=$LIBTORCH_PATH
 ---> Using cache
 ---> c73573681888
Step 9/18 : RUN ln -s ${LIBTORCH_PATH}/include/* /usr/local/include/ &&     ln -s ${LIBTORCH_PATH}/lib/* /usr/local/lib/
 ---> Using cache
 ---> 19d49516a9ee
Step 10/18 : WORKDIR /app
 ---> Using cache
 ---> 7c48a054db6d
Step 11/18 : COPY go.mod go.sum ./
 ---> Using cache
 ---> 1f99a86eced9
Step 12/18 : RUN go mod download
 ---> Using cache
 ---> 755acffaff1f
Step 13/18 : ARG BUILD_CORES=2
 ---> Using cache
 ---> dff8f2c603fb
Step 14/18 : RUN cd /root/go/pkg/mod/github.com/mishabeliy15/gotorch*/cgotorch &&     chmod +x build.sh &&     ./build.sh -j${BUILD_CORES}
 ---> Using cache
 ---> 6698c92ad13f
Step 15/18 : RUN git clone https://github.com/pytorch/vision.git --branch v0.15.2 &&      cd vision &&     mkdir build &&     cd build &&     cmake .. &&     make -j${BUILD_CORES} &&     make install &&     cd .. && rm -rf build
 ---> Running in a00bf4f36a00
Cloning into 'vision'...
Note: checking out 'fa99a5360fbcd1683311d57a76fcc0e7323a4c1e'.

You are in 'detached HEAD' state. You can look around, make experimental
changes and commit them, and you can discard any commits you make in this
state without impacting any branches by performing another checkout.

If you want to create a new branch to retain commits you create, you may
do so (now or later) by using -b with the checkout command again. Example:

  git checkout -b <new-branch-name>

-- The C compiler identification is GNU 7.5.0
-- The CXX compiler identification is GNU 7.5.0
-- Detecting C compiler ABI info
-- Detecting C compiler ABI info - done
-- Check for working C compiler: /usr/bin/cc - skipped
-- Detecting C compile features
-- Detecting C compile features - done
-- Detecting CXX compiler ABI info
-- Detecting CXX compiler ABI info - done
-- Check for working CXX compiler: /usr/bin/c++ - skipped
-- Detecting CXX compile features
-- Detecting CXX compile features - done
-- Performing Test CMAKE_HAVE_LIBC_PTHREAD
-- Performing Test CMAKE_HAVE_LIBC_PTHREAD - Failed
-- Looking for pthread_create in pthreads
-- Looking for pthread_create in pthreads - not found
-- Looking for pthread_create in pthread
-- Looking for pthread_create in pthread - found
-- Found Threads: TRUE  
-- Could NOT find CUDA (missing: CUDA_CUDART_LIBRARY) (found version "10.2")
CMake Warning at /usr/local/lib/python3.6/dist-packages/torch/share/cmake/Caffe2/public/cuda.cmake:31 (message):
  Caffe2: CUDA cannot be found.  Depending on whether you are building Caffe2
  or a Caffe2 dependent library, the next warning / error will give you more
  info.
Call Stack (most recent call first):
  /usr/local/lib/python3.6/dist-packages/torch/share/cmake/Caffe2/Caffe2Config.cmake:88 (include)
  /usr/local/lib/python3.6/dist-packages/torch/share/cmake/Torch/TorchConfig.cmake:68 (find_package)
  CMakeLists.txt:18 (find_package)


CMake Error at /usr/local/lib/python3.6/dist-packages/torch/share/cmake/Caffe2/Caffe2Config.cmake:90 (message):
  Your installed Caffe2 version uses CUDA but I cannot find the CUDA
  libraries.  Please set the proper CUDA prefixes and / or install CUDA.
Call Stack (most recent call first):
  /usr/local/lib/python3.6/dist-packages/torch/share/cmake/Torch/TorchConfig.cmake:68 (find_package)
  CMakeLists.txt:18 (find_package)


-- Configuring incomplete, errors occurred!
The command '/bin/sh -c git clone https://github.com/pytorch/vision.git --branch v0.15.2 &&      cd vision &&     mkdir build &&     cd build &&     cmake .. &&     make -j${BUILD_CORES} &&     make install &&     cd .. && rm -rf build' returned a non-zero code: 1
make: *** [Makefile:111: build-jetson-docker] Error 1
6

@s.pinchuk do you actually have the full CUDA Toolkit installed on your device, outside of container?

# Set ENV var for libtorch
ARG PYTHON_VERSION=3.6
ENV LIBTORCH_PATH=/usr/local/lib/python${PYTHON_VERSION}/dist-packages/torch \
    CUDA_HOME=/usr/local/cuda-10.2
ENV LD_LIBRARY_PATH=$LD_LIBRARY_PATH:$CUDA_HOME/lib64:$LIBTORCH_PATH/lib:/usr/local/lib:/app/src/libs \
    Torch_DIR=$LIBTORCH_PATH \
    PATH=/usr/local/cuda-10.2/bin:$PATH \
    CUDA_BIN_PATH=/usr/local/cuda-10.2/bin

# Create symlinks libtorch to system
RUN ln -s ${LIBTORCH_PATH}/include/* /usr/local/include/ && \
    ln -s ${LIBTORCH_PATH}/lib/* /usr/local/lib/ && \
    ln -s /usr/local/cuda-10.2 /usr/local/cuda

Try removing ln -s /usr/local/cuda-10.2 /usr/local/cuda → you are overwriting the symlink to /usr/local/cuda that the NVIDIA Container Runtime mounts. Also, try referencing /usr/local/cuda in your paths instead of /usr/local/cuda-10.2. There are lots of custom steps you are performing here that I would try removing until it works.

7

Also I found such interesting issue Could NOT find CUDA (missing: CUDA_CUDART_LIBRARY) (found version "8.0") · Issue #317 · nicehash/nheqminer · GitHub
But I can`t use it, because in docker container we don’t have such file, by path /usr/local/cuda/lib64/libcudart.so

root@a0d109cc52a5:/usr/local/cuda/lib64/stubs# ls
libcublas.so    libcufft.so   libcusolver.so  libnppial.so   libnppidei.so  libnppim.so   libnppitc.so   libnvrtc.so
libcublasLt.so  libcufftw.so  libcusparse.so  libnppicc.so   libnppif.so    libnppist.so  libnpps.so
libcuda.so      libcurand.so  libnppc.so      libnppicom.so  libnppig.so    libnppisu.so  libnvgraph.so
root@a0d109cc52a5:/usr/local/cuda/lib64/stubs# cd ../
root@a0d109cc52a5:/usr/local/cuda/lib64# ls
libcudadevrt.a  libcudart_static.a  stubs
root@a0d109cc52a5:/usr/local/cuda/lib64#
8

I did such actions and Dockerfile now :

cat Jetson.Dockerfile 
ARG BASE_IMAGE=nvcr.io/nvidia/l4t-pytorch:r32.6.1-pth1.9-py3

FROM $BASE_IMAGE as base

RUN apt-get update && apt-get upgrade -yy && apt-get install -yy \
    wget \
    git \
    build-essential \
    git \
    clang-3.8 \
    libpng-dev && \
    apt remove cmake -y && \
    pip3 install --upgrade pip && \
    pip3 install cmake --upgrade
#    libgeos-dev \ ToDo: need to install later
#    libproj-dev


# Donwload and install Go
ENV GOLANG_VERSION=1.21.0 \
    GOROOT=/usr/local/go \
    PATH=$PATH:/usr/local/go/bin

RUN wget https://golang.org/dl/go${GOLANG_VERSION}.linux-arm64.tar.gz && \
    tar -C /usr/local -xzf go${GOLANG_VERSION}.linux-arm64.tar.gz && \
    rm go${GOLANG_VERSION}.linux-arm64.tar.gz

# Set ENV var for libtorch
ARG PYTHON_VERSION=3.6
ENV LIBTORCH_PATH=/usr/local/lib/python${PYTHON_VERSION}/dist-packages/torch \
    CUDA_HOME=/usr/local/cuda-10.2
ENV LD_LIBRARY_PATH=$LD_LIBRARY_PATH:$CUDA_HOME/lib64:$LIBTORCH_PATH/lib:/usr/local/lib:/app/src/libs \
    Torch_DIR=$LIBTORCH_PATH \
    PATH=/usr/local/cuda-10.2/bin:$PATH \
    CUDA_BIN_PATH=/usr/local/cuda-10.2/bin

# Create symlinks libtorch to system
RUN ln -s ${LIBTORCH_PATH}/include/* /usr/local/include/ && \
    ln -s ${LIBTORCH_PATH}/lib/* /usr/local/lib/ 
#&& \
#    ln -s /usr/local/cuda-10.2 /usr/local/cuda


WORKDIR /app

COPY go.mod go.sum ./
RUN go mod download

ARG BUILD_CORES=2
# Build gotorch
RUN cd /root/go/pkg/mod/github.com/mishabeliy15/gotorch*/cgotorch && \
    chmod +x build.sh 
#&& \
#    ./build.sh -j${BUILD_CORES}

# Download and build libtorchvision
RUN git clone https://github.com/pytorch/vision.git --branch v0.15.2 &&  \
    cd vision && \
    mkdir build 
# && \
 #   cd build && \
 #   cmake .. && \
 #   make -j${BUILD_CORES} && \
 #   make install && \
 #   cd .. && rm -rf build

FROM base as dev

WORKDIR /app/src
CMD ["bash"]
jetson@nano:~/p

But I have the same error, when we build torch in docker container:

root@a0d109cc52a5:~/go/pkg/mod/github.com/mishabeliy15/gotorch@v0.0.0-20230822142026-b89c6923d2d0# cd cgotorch/
root@a0d109cc52a5:~/go/pkg/mod/github.com/mishabeliy15/gotorch@v0.0.0-20230822142026-b89c6923d2d0/cgotorch# ./build.sh
+ set -e
++ uname
+ OS=Linux
+++ dirname ./build.sh
++ realpath .
+ BASE_DIR=/root/go/pkg/mod/github.com/mishabeliy15/gotorch@v0.0.0-20230822142026-b89c6923d2d0/cgotorch
+ BUILD_DIR=/root/go/pkg/mod/github.com/mishabeliy15/gotorch@v0.0.0-20230822142026-b89c6923d2d0/cgotorch/build
+ rm -rf /root/go/pkg/mod/github.com/mishabeliy15/gotorch@v0.0.0-20230822142026-b89c6923d2d0/cgotorch/build
+ mkdir -p /root/go/pkg/mod/github.com/mishabeliy15/gotorch@v0.0.0-20230822142026-b89c6923d2d0/cgotorch/build
+ cd /root/go/pkg/mod/github.com/mishabeliy15/gotorch@v0.0.0-20230822142026-b89c6923d2d0/cgotorch/build
+ cmake ..
-- The C compiler identification is GNU 7.5.0
-- The CXX compiler identification is GNU 7.5.0
-- Detecting C compiler ABI info
-- Detecting C compiler ABI info - done
-- Check for working C compiler: /usr/bin/cc - skipped
-- Detecting C compile features
-- Detecting C compile features - done
-- Detecting CXX compiler ABI info
-- Detecting CXX compiler ABI info - done
-- Check for working CXX compiler: /usr/bin/c++ - skipped
-- Detecting CXX compile features
-- Detecting CXX compile features - done
CMake Warning (dev) at CMakeLists.txt:8 (find_package):
  Policy CMP0146 is not set: The FindCUDA module is removed.  Run "cmake
  --help-policy CMP0146" for policy details.  Use the cmake_policy command to
  set the policy and suppress this warning.

This warning is for project developers.  Use -Wno-dev to suppress it.

-- Performing Test CMAKE_HAVE_LIBC_PTHREAD
-- Performing Test CMAKE_HAVE_LIBC_PTHREAD - Failed
-- Looking for pthread_create in pthreads
-- Looking for pthread_create in pthreads - not found
-- Looking for pthread_create in pthread
-- Looking for pthread_create in pthread - found
-- Found Threads: TRUE  
-- Could NOT find CUDA (missing: CUDA_CUDART_LIBRARY) (found version "10.2")
CMake Warning at /usr/local/lib/python3.6/dist-packages/torch/share/cmake/Caffe2/public/cuda.cmake:31 (message):
  Caffe2: CUDA cannot be found.  Depending on whether you are building Caffe2
  or a Caffe2 dependent library, the next warning / error will give you more
  info.
Call Stack (most recent call first):
  /usr/local/lib/python3.6/dist-packages/torch/share/cmake/Caffe2/Caffe2Config.cmake:88 (include)
  /usr/local/lib/python3.6/dist-packages/torch/share/cmake/Torch/TorchConfig.cmake:68 (find_package)
  CMakeLists.txt:15 (find_package)


CMake Error at /usr/local/lib/python3.6/dist-packages/torch/share/cmake/Caffe2/Caffe2Config.cmake:90 (message):
  Your installed Caffe2 version uses CUDA but I cannot find the CUDA
  libraries.  Please set the proper CUDA prefixes and / or install CUDA.
Call Stack (most recent call first):
  /usr/local/lib/python3.6/dist-packages/torch/share/cmake/Torch/TorchConfig.cmake:68 (find_package)
  CMakeLists.txt:15 (find_package)


-- Configuring incomplete, errors occurred!
root@a0d109cc52a5:~/go/pkg/mod/github.com/mishabeliy15/gotorch@v0.0.0-20230822142026-b89c6923d2d0/cgotorch# cd /app/sr
bash: cd: /app/sr: No such file or directory
root@a0d109cc52a5:~/go/pkg/mod/github.com/mishabeliy15/gotorch@v0.0.0-20230822142026-b89c6923d2d0/cgotorch# cd /app/src
root@a0d109cc52a5:/app/src# cd /go/pkg/mod/github.com/mishabeliy15/gotorch@v0.0.0-20230822142026-b89c6923d2d0/cgotorch#
bash: cd: /go/pkg/mod/github.com/mishabeliy15/gotorch@v0.0.0-20230822142026-b89c6923d2d0/cgotorch#: No such file or directory
root@a0d109cc52a5:/app/src# cd ~/go/pkg/mod/github.com/mishabeliy15/gotorch@v0.0.0-20230822142026-b89c6923d2d0/cgotorch#
bash: cd: /root/go/pkg/mod/github.com/mishabeliy15/gotorch@v0.0.0-20230822142026-b89c6923d2d0/cgotorch#: No such file or directory

Also I tried to remove package build and create it clean one more time, but result is the same.
Steps of commands, which I did:

rm -rf build
mkdir build
cd build
cmake -DCUDA_DIR=/usr/local/cuda ..

Logs in this case:

root@a0d109cc52a5:~/go/pkg/mod/github.com/mishabeliy15/gotorch@v0.0.0-20230822142026-b89c6923d2d0/cgotorch# ls
CMakeLists.txt  cgotorch.h  device.cc      functional.h  ivalue.cc  memory.h  pickle.cc  tensor.h  torchdef.h
build           cuda.cc     device.h       init.cc       ivalue.h   optim.cc  pickle.h   torch.cc  torchscript.cc
build.sh        cuda.h      functional.cc  init.h        memory.cc  optim.h   tensor.cc  torch.h   torchscript.h
root@a0d109cc52a5:~/go/pkg/mod/github.com/mishabeliy15/gotorch@v0.0.0-20230822142026-b89c6923d2d0/cgotorch# ls
CMakeLists.txt  cgotorch.h  device.cc      functional.h  ivalue.cc  memory.h  pickle.cc  tensor.h  torchdef.h
build           cuda.cc     device.h       init.cc       ivalue.h   optim.cc  pickle.h   torch.cc  torchscript.cc
build.sh        cuda.h      functional.cc  init.h        memory.cc  optim.h   tensor.cc  torch.h   torchscript.h
root@a0d109cc52a5:~/go/pkg/mod/github.com/mishabeliy15/gotorch@v0.0.0-20230822142026-b89c6923d2d0/cgotorch# rm -rf build
root@a0d109cc52a5:~/go/pkg/mod/github.com/mishabeliy15/gotorch@v0.0.0-20230822142026-b89c6923d2d0/cgotorch# mkdir build
root@a0d109cc52a5:~/go/pkg/mod/github.com/mishabeliy15/gotorch@v0.0.0-20230822142026-b89c6923d2d0/cgotorch# cd build
root@a0d109cc52a5:~/go/pkg/mod/github.com/mishabeliy15/gotorch@v0.0.0-20230822142026-b89c6923d2d0/cgotorch/build# cmake -DCUDA_DIR=/usr/local/cuda ..
-- The C compiler identification is GNU 7.5.0
-- The CXX compiler identification is GNU 7.5.0
-- Detecting C compiler ABI info
-- Detecting C compiler ABI info - done
-- Check for working C compiler: /usr/bin/cc - skipped
-- Detecting C compile features
-- Detecting C compile features - done
-- Detecting CXX compiler ABI info
-- Detecting CXX compiler ABI info - done
-- Check for working CXX compiler: /usr/bin/c++ - skipped
-- Detecting CXX compile features
-- Detecting CXX compile features - done
CMake Warning (dev) at CMakeLists.txt:8 (find_package):
  Policy CMP0146 is not set: The FindCUDA module is removed.  Run "cmake
  --help-policy CMP0146" for policy details.  Use the cmake_policy command to
  set the policy and suppress this warning.

This warning is for project developers.  Use -Wno-dev to suppress it.

-- Performing Test CMAKE_HAVE_LIBC_PTHREAD
-- Performing Test CMAKE_HAVE_LIBC_PTHREAD - Failed
-- Looking for pthread_create in pthreads
-- Looking for pthread_create in pthreads - not found
-- Looking for pthread_create in pthread
-- Looking for pthread_create in pthread - found
-- Found Threads: TRUE  
-- Could NOT find CUDA (missing: CUDA_CUDART_LIBRARY) (found version "10.2")
CMake Warning at /usr/local/lib/python3.6/dist-packages/torch/share/cmake/Caffe2/public/cuda.cmake:31 (message):
  Caffe2: CUDA cannot be found.  Depending on whether you are building Caffe2
  or a Caffe2 dependent library, the next warning / error will give you more
  info.
Call Stack (most recent call first):
  /usr/local/lib/python3.6/dist-packages/torch/share/cmake/Caffe2/Caffe2Config.cmake:88 (include)
  /usr/local/lib/python3.6/dist-packages/torch/share/cmake/Torch/TorchConfig.cmake:68 (find_package)
  CMakeLists.txt:15 (find_package)


CMake Error at /usr/local/lib/python3.6/dist-packages/torch/share/cmake/Caffe2/Caffe2Config.cmake:90 (message):
  Your installed Caffe2 version uses CUDA but I cannot find the CUDA
  libraries.  Please set the proper CUDA prefixes and / or install CUDA.
Call Stack (most recent call first):
  /usr/local/lib/python3.6/dist-packages/torch/share/cmake/Torch/TorchConfig.cmake:68 (find_package)
  CMakeLists.txt:15 (find_package)


-- Configuring incomplete, errors occurred!
root@a0d109cc52a5:~/go/pkg/mod/github.com/mishabeliy15/gotorch@v0.0.0-20230822142026-b89c6923d2d0/cgotorch/build# cmake -DCUDA_DIR=/usr/local/cuda-10-2 ..
CMake Warning (dev) at CMakeLists.txt:8 (find_package):
  Policy CMP0146 is not set: The FindCUDA module is removed.  Run "cmake
  --help-policy CMP0146" for policy details.  Use the cmake_policy command to
  set the policy and suppress this warning.

This warning is for project developers.  Use -Wno-dev to suppress it.

-- Could NOT find CUDA (missing: CUDA_CUDART_LIBRARY) (found version "10.2")
CMake Warning at /usr/local/lib/python3.6/dist-packages/torch/share/cmake/Caffe2/public/cuda.cmake:31 (message):
  Caffe2: CUDA cannot be found.  Depending on whether you are building Caffe2
  or a Caffe2 dependent library, the next warning / error will give you more
  info.
Call Stack (most recent call first):
  /usr/local/lib/python3.6/dist-packages/torch/share/cmake/Caffe2/Caffe2Config.cmake:88 (include)
  /usr/local/lib/python3.6/dist-packages/torch/share/cmake/Torch/TorchConfig.cmake:68 (find_package)
  CMakeLists.txt:15 (find_package)


CMake Error at /usr/local/lib/python3.6/dist-packages/torch/share/cmake/Caffe2/Caffe2Config.cmake:90 (message):
  Your installed Caffe2 version uses CUDA but I cannot find the CUDA
  libraries.  Please set the proper CUDA prefixes and / or install CUDA.
Call Stack (most recent call first):
  /usr/local/lib/python3.6/dist-packages/torch/share/cmake/Torch/TorchConfig.cmake:68 (find_package)
  CMakeLists.txt:15 (find_package)


-- Configuring incomplete, errors occurred!
root@a0d109cc52a5:~/go/pkg/mod/github.com/mishabeliy15/gotorch@v0.0.0-20230822142026-b89c6923d2d0/cgotorch/build# cmake -DCUDA_DIR=/usr/local/cuda-11-4 ..
CMake Warning (dev) at CMakeLists.txt:8 (find_package):
  Policy CMP0146 is not set: The FindCUDA module is removed.  Run "cmake
  --help-policy CMP0146" for policy details.  Use the cmake_policy command to
  set the policy and suppress this warning.

This warning is for project developers.  Use -Wno-dev to suppress it.

-- Could NOT find CUDA (missing: CUDA_CUDART_LIBRARY) (found version "10.2")
CMake Warning at /usr/local/lib/python3.6/dist-packages/torch/share/cmake/Caffe2/public/cuda.cmake:31 (message):
  Caffe2: CUDA cannot be found.  Depending on whether you are building Caffe2
  or a Caffe2 dependent library, the next warning / error will give you more
  info.
Call Stack (most recent call first):
  /usr/local/lib/python3.6/dist-packages/torch/share/cmake/Caffe2/Caffe2Config.cmake:88 (include)
  /usr/local/lib/python3.6/dist-packages/torch/share/cmake/Torch/TorchConfig.cmake:68 (find_package)
  CMakeLists.txt:15 (find_package)


CMake Error at /usr/local/lib/python3.6/dist-packages/torch/share/cmake/Caffe2/Caffe2Config.cmake:90 (message):
  Your installed Caffe2 version uses CUDA but I cannot find the CUDA
  libraries.  Please set the proper CUDA prefixes and / or install CUDA.
Call Stack (most recent call first):
  /usr/local/lib/python3.6/dist-packages/torch/share/cmake/Torch/TorchConfig.cmake:68 (find_package)
  CMakeLists.txt:15 (find_package)


-- Configuring incomplete, errors occurred!
9

You are still setting the CUDA paths yourself and that should not be necessary.

Are you able to build/run CUDA samples (like deviceQuery and vectorAdd) in nvcr.io/nvidia/l4t-base:r32.6.1 ?

sudo docker run --runtime nvidia -it --rm nvcr.io/nvidia/l4t-base:r32.6.1
apt-get update && apt-get install -y build-essential
cp -r /usr/local/cuda/samples /cuda_samples
cd /cuda_samples/1_Utilities/deviceQuery
make
./deviceQuery
cd /cuda_samples/0_Simple/vectorAdd
make
./vectorAdd

This should confirm that your NVIDIA Container Runtime, CUDA Toolkit installation, and GPU are working ok.

10 
root@e580222041cd:/# cp -r /usr/local/cuda/samples /cuda_samples
cp: cannot stat '/usr/local/cuda/samples': No such file or directory
root@e580222041cd:/# cd /cuda_samples/1_Utilities/deviceQuery
bash: cd: /cuda_samples/1_Utilities/deviceQuery: No such file or directory
11

@s.pinchuk outside of container, do you have the CUDA Toolkit under /usr/local/cuda and the samples in /usr/local/cuda/samples on your device?

I think your issues may be related to either you not having CUDA Toolkit installed on your device, or your NVIDIA Container Runtime misconfigured. The SD card images comes with all of this installed and already working, so before sinking a lot of time debugging it, I would recommend flashing a fresh SD card and seeing if it works for you then.

12

Yes I have outside container all packages:

jetson@nano:/usr/local/cuda$ ls -a
.  ..  bin  doc  EULA.txt  extras  include  lib64  nvml  nvvm  nvvmx  samples  share  targets  tools  version.json  version.txt
jetson@nano:/usr/local/cuda$ cd samples/
jetson@nano:/usr/local/cuda/samples$ ls -a
.  ..  0_Simple  1_Utilities  2_Graphics  3_Imaging  4_Finance  5_Simulations  6_Advanced  7_CUDALibraries  common  EULA.txt  Makefile
jetson@nano:/usr/local/cuda/samples$
13

OK, something is going wrong with the mounts that the NVIDIA Container Runtime does - do you have these CSV files?

ls -ll /etc/nvidia-container-runtime/host-files-for-container.d/
total 32
-rw-r--r-- 1 root root    26 May 23  2021 cuda.csv
-rw-r--r-- 1 root root  4250 Jul 13  2021 cudnn.csv
-rw-r--r-- 1 root root 12240 Feb  2  2022 l4t.csv
-rw-r--r-- 1 root root  1590 Jan 14  2022 tensorrt.csv
-rw-r--r-- 1 root root   325 Aug 11  2020 visionworks.csv

You could try reinstalling these packages, or more likely try a fresh SD card and test that your able to build/run the CUDA samples then:

$ apt-cache search nvidia-container
libnvidia-container-tools - NVIDIA container runtime library (command-line tools)
libnvidia-container0 - NVIDIA container runtime library
libnvidia-container1 - NVIDIA container runtime library
nvidia-container-csv-cuda - Jetpack CUDA CSV file
nvidia-container-csv-cudnn - Jetpack CUDNN CSV file
nvidia-container-csv-tensorrt - Jetpack TensorRT CSV file
nvidia-container-csv-visionworks - Jetpack VisionWorks CSV file
nvidia-container-runtime - NVIDIA container runtime
nvidia-container-toolkit - NVIDIA container runtime hook
nvidia-container - NVIDIA Container Meta Package
14

Yes, I have them:

jetson@nano:/usr/local/cuda/samples$ ls -ll /etc/nvidia-container-runtime/host-files-for-container.d/
total 32
-rw-r--r-- 1 root root    26 mei 24  2021 cuda.csv
-rw-r--r-- 1 root root  4250 jul 13  2021 cudnn.csv
-rw-r--r-- 1 root root 11828 jul 26  2021 l4t.csv
-rw-r--r-- 1 root root  1521 jun 30  2021 tensorrt.csv
-rw-r--r-- 1 root root   325 aug 11  2020 visionworks.csv

If we speak about second command, all is fine:

 apt-cache search nvidia-container
libnvidia-container1 - NVIDIA container runtime library
libnvidia-container1-dbg - NVIDIA container runtime library (debugging symbols)
libnvidia-container-dev - NVIDIA container runtime library (development files)
libnvidia-container-tools - NVIDIA container runtime library (command-line tools)
nvidia-container-runtime - NVIDIA container runtime
nvidia-container-toolkit - NVIDIA Container toolkit
nvidia-container-toolkit-base - NVIDIA Container Toolkit Base
libnvidia-container0 - NVIDIA container runtime library
nvidia-container-csv-cuda - Jetpack CUDA CSV file
nvidia-container-csv-cudnn - Jetpack CUDNN CSV file
nvidia-container-csv-tensorrt - Jetpack TensorRT CSV file
nvidia-container-csv-visionworks - Jetpack VisionWorks CSV file
nvidia-container - NVIDIA Container Meta Package

I have already reinstalled CUDA twice before I created this thread with questions.

15

Perhaps it is then related to your re-installations of CUDA. What I was recommending was re-flashing your SD card, because these Docker<->CUDA mounting issues on JetPack 4 can be tricky to debug why it’s not working.

16

STEP_1. We want to see does pytorch works in docker container:

docker run -it --rm --runtime nvidia  nvcr.io/nvidia/l4t-pytorch:r32.6.1-pth1.9-py3
Unable to find image 'nvcr.io/nvidia/l4t-pytorch:r32.6.1-pth1.9-py3' locally
r32.6.1-pth1.9-py3: Pulling from nvidia/l4t-pytorch
e47a8a86d66c: Pull complete 
bdce3430dad6: Pull complete 
c26a6b81c746: Pull complete 
a70635e646de: Pull complete 
24cbe60e3161: Pull complete 
c7f64cc97a39: Pull complete 
8777adb92eda: Pull complete 
542a24b3572f: Pull complete 
3d5ade2b8849: Pull complete 
3e865584f789: Pull complete 
4811af6cacf1: Pull complete 
db645757aac7: Pull complete 
0105dc23dc93: Pull complete 
c8aff3baf097: Pull complete 
5f50b3e38480: Pull complete 
b5418d85b074: Pull complete 
78d4b308041f: Pull complete 
44f9a409dddf: Pull complete 
dc19965f7530: Pull complete 
73f02f5e17e7: Pull complete 
de4b3c1acb9b: Pull complete 
6c2b9910111b: Pull complete 
1cb44dee771f: Pull complete 
bf00de567184: Pull complete 
b73a0bf964ae: Pull complete 
579ce47ae238: Pull complete 
2262c28018a2: Pull complete 
177925b6dcf7: Pull complete 
Digest: sha256:247550bf8554d2f4669aa9fe0e9d319ff2c3aaf67275e6259346d6e34186a623
Status: Downloaded newer image for nvcr.io/nvidia/l4t-pytorch:r32.6.1-pth1.9-py3
root@229a1ace600d:/# python3         
Python 3.6.9 (default, Jan 26 2021, 15:33:00) 
[GCC 8.4.0] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> import torch
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/usr/local/lib/python3.6/dist-packages/torch/__init__.py", line 196, in <module>
    _load_global_deps()
  File "/usr/local/lib/python3.6/dist-packages/torch/__init__.py", line 149, in _load_global_deps
    ctypes.CDLL(lib_path, mode=ctypes.RTLD_GLOBAL)
  File "/usr/lib/python3.6/ctypes/__init__.py", line 348, in __init__
    self._handle = _dlopen(self._name, mode)
OSError: libcurand.so.10: cannot open shared object file: No such file or directory
>>>

So it doesn’t work in container

STEP_2. We want to see does pytorch work on Jetson Nano 4GB:

jetson@nano:~$ python3
Python 3.8.10 (default, May 26 2023, 14:05:08) 
[GCC 9.4.0] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> import torch
>>> torch.cuda.is_available()
True
>>>
17

Also I have one more trouble. CUDA doesn’t work on C++ lib torch. But python works:

jetson@nano:~$ python3
Python 3.8.10 (default, May 26 2023, 14:05:08) 
[GCC 9.4.0] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> import torch
>>> torch.cuda.is_available()
True
>>>

STEP_1. Write architecture 53 in CMAKE file for Jetson Nano 4GB version:

cat CMakeLists.txt
cmake_minimum_required(VERSION 3.12)
project(tcuda)

set(CMAKE_CXX_STANDARD 14)
set(CMAKE_CXX_STANDARD_REQUIRED True)

# Check for CUDA if necessary
# find_package(CUDA QUIET)
if(CMAKE_CUDA_COMPILER)
  enable_language(CUDA)
  add_definitions(-DUSE_CUDA)
  set(CUDA_FLAGS ${CUDA_LIBRARIES})
  set(CMAKE_CUDA_ARCHITECTURES 53)
endif()

# Locate LibTorch
find_package(Torch REQUIRED)
include_directories(${TORCH_INCLUDE_DIRS})

# Collect all source files
file(GLOB SOURCES "*.cpp")

add_executable(tcuda ${SOURCES})
# Link against LibTorch
target_link_libraries(tcuda ${TORCH_LIBRARIES})

if(CUDA_FOUND)
  target_link_libraries(tcuda ${CUDA_LIBRARIES})
endif()

STEP_2. Write main.cpp for testing:

at main.cpp 
#include "torch/torch.h"
#include <iostream>


int main() {
    std::cout << "Is Cuda Available: " << torch::cuda::is_available() << std::endl;
    auto device = torch::cuda::is_available() ? torch::kCUDA : torch::kCPU;
    torch::Tensor tensor = torch::rand({2, 3}, device=device);
    std::cout << tensor << std::endl;
    return 0;
}

STEP_3. Run command sudo CUDACXX=/usr/local/cuda/bin/nvcc cmake -DCMAKE_PREFIX_PATH=/usr/local/lib/python3.8/dist-packages/torch ..

sudo  CUDACXX=/usr/local/cuda/bin/nvcc cmake -DCMAKE_PREFIX_PATH=/usr/local/lib/python3.8/dist-packages/torch ..
-- The CUDA compiler identification is NVIDIA 10.2.300
-- Detecting CUDA compiler ABI info
-- Detecting CUDA compiler ABI info - done
-- Check for working CUDA compiler: /usr/local/cuda/bin/nvcc - skipped
-- Detecting CUDA compile features
-- Detecting CUDA compile features - done
-- Found CUDA: /usr/local/cuda (found version "10.2") 
-- Caffe2: CUDA detected: 10.2
-- Caffe2: CUDA nvcc is: /usr/local/cuda/bin/nvcc
-- Caffe2: CUDA toolkit directory: /usr/local/cuda
-- Caffe2: Header version is: 10.2
-- Found CUDNN: /usr/lib/aarch64-linux-gnu/libcudnn.so  
-- Found cuDNN: v8.2.1  (include: /usr/include, library: /usr/lib/aarch64-linux-gnu/libcudnn.so)
-- /usr/local/cuda/lib64/libnvrtc.so shorthash is 7d272a04
-- Autodetected CUDA architecture(s):  5.3
-- Added CUDA NVCC flags for: -gencode;arch=compute_53,code=sm_53
CMake Warning at /usr/local/lib/python3.8/dist-packages/torch/share/cmake/Torch/TorchConfig.cmake:22 (message):
  static library kineto_LIBRARY-NOTFOUND not found.
Call Stack (most recent call first):
  /usr/local/lib/python3.8/dist-packages/torch/share/cmake/Torch/TorchConfig.cmake:127 (append_torchlib_if_found)
  CMakeLists.txt:17 (find_package)


-- Found Torch: /usr/local/lib/python3.8/dist-packages/torch/lib/libtorch.so  
-- Configuring done (9.3s)
-- Generating done (0.0s)
-- Build files have been written to: /home/jetson/go/pkg/mod/github.com/mishabeliy15/gotorch@v0.0.0-20230825120526-26c4d596dbac/test_cuda/build

STEP_4. Build

sudo make -j2
[ 50%] Building CXX object CMakeFiles/tcuda.dir/main.cpp.o
[100%] Linking CXX executable tcuda
[100%] Built target tcuda

STEP_5.
Run tcuda:

./tcuda 
Is Cuda Available: 0
 0.5559  0.4669  0.0445
 0.7814  0.0697  0.8526
[ CPUFloatType{2,3} ]
18

Outside of container, do you have /usr/local/cuda/lib64/libcurand.so.10 ?

$ ls -ll /usr/local/cuda/lib64/libcurand*

lrwxrwxrwx 1 root root       15 Mar  1  2021 /usr/local/cuda/lib64/libcurand.so -> libcurand.so.10
lrwxrwxrwx 1 root root       23 Mar  1  2021 /usr/local/cuda/lib64/libcurand.so.10 -> libcurand.so.10.1.2.300
-rw-r--r-- 1 root root 62698584 Mar  1  2021 /usr/local/cuda/lib64/libcurand.so.10.1.2.300
-rw-r--r-- 1 root root 62767380 Mar  1  2021 /usr/local/cuda/lib64/libcurand_static.a

We’ve established earlier in this thread that something’s funky with your NVIDIA Container Runtime and it doesn’t seem to be properly mounting the CUDA drivers/libraries into the containers correctly. Due to your unknown environment and changes, my recommendation was to reflash the SD card and then confirm that GPU is working in l4t-base container with the CUDA deviceQuery/vectorAdd samples. Have you done that?

STEP_3. Run command sudo CUDACXX=/usr/local/cuda/bin/nvcc cmake -DCMAKE_PREFIX_PATH=/usr/local/lib/python3.8/dist-packages/torch ..

You are on Jetson Nano and JetPack 4 (which uses Python 3.6), but these paths/commands show Python 3.8. I’m unsure which PyTorch wheel that is using or about the tests with libtorch, sorry.

20

This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.