Package Details: python-tensorrt 10.3.0.26-1

Git Clone URL: https://aur.archlinux.org/tensorrt.git (read-only, click to copy)
Package Base: tensorrt
Description: A platform for high-performance deep learning inference on NVIDIA hardware (python bindings and tools)
Upstream URL: https://developer.nvidia.com/tensorrt/
Keywords: ai artificial intelligence nvidia
Licenses: Apache-2.0, LicenseRef-custom
Provides: python-onnx-graphsurgeon, python-polygraphy, python-tensorflow-quantization
Submitter: dbermond
Maintainer: dbermond
Last Packager: dbermond
Votes: 18
Popularity: 1.22
First Submitted: 2018-07-29 16:17 (UTC)
Last Updated: 2024-08-10 14:22 (UTC)

Dependencies (18)

Sources (13)

Latest Comments

« First ‹ Previous 1 2 3 4 5 6 7 8 9 Next › Last »

dbermond commented on 2022-12-10 16:00 (UTC)

@otakutyrant license prohibits the package distribution. Unless Nvidia grants the package distribution within Arch Linux, we cannot do it.

dbermond commented on 2022-12-10 15:56 (UTC)

@Baytars It may be. You can always build the package in a clean chroot to make sure that it builds fine. This is how all official repository packages are built, and I build all my packages in this way.

otakutyrant commented on 2022-12-07 16:14 (UTC)

This package is more unpopular than I thought. It should be in official packages.

Baytars commented on 2022-12-01 08:08 (UTC)

@dbermond Is it related to using virtual environment in Anaconda? The output log showed that, instead of the system Python distribution, the Python path it found was that managed by Anaconda, which I use routinely. So despite the existence of makedepends, PKGBUILD could not find the dependencies in the virtual environment.

dbermond commented on 2022-11-30 21:54 (UTC)

@Baytars I've just checked, and the package is building fine. It seems that something is wrong with your system, as that packages that you suggest to be handled with 'pip install' are already in makedepends. Also, notice that installing things with 'pip install' in a PKGBUILD is generally not acceptable.

Baytars commented on 2022-11-29 16:08 (UTC) (edited on 2022-11-29 17:47 (UTC) by Baytars)

Issue 1: could not find pybind11 missing pybind11_DIR
Fix: pip install pybind11[global]

Issue 2:

Looking for a CUDA compiler failed with the following output:
-- The CUDA compiler identification is unknown
CMake Error at /usr/local/share/cmake-3.23/Modules/CMakeDetermineCUDACompiler.cmake:633 (message):
  Failed to detect a default CUDA architecture.

Fix:

    cmake -B build -S TensorRT \
        ......
        -DCMAKE_CUDA_COMPILER="/opt/cuda/bin/nvcc" \  # add this line
        ......
        -Wno-dev

Issue 3: Specify CUDA_TOOLKIT_ROOT_DIR
Fix:

    cmake -B build -S TensorRT \
        ......
        -DCUDAToolkit_ROOT="/opt/cuda/" \ # add this line
        ......
        -Wno-dev

Issue 4:

CMake Error: The following variables are used in this project, but they are set to NOTFOUND.
Please set them or make sure they are set and tested correctly in the CMake files:
/var/tmp/pamac-build/tensorrt/src/TensorRT/python/PY_CONFIG_INCLUDE
   used as include directory in directory /var/tmp/pamac-build/tensorrt/src/TensorRT/python
/var/tmp/pamac-build/tensorrt/src/TensorRT/python/PY_INCLUDE
   used as include directory in directory /var/tmp/pamac-build/tensorrt/src/TensorRT/python

Fix:

cd TensorRT/python
# add begin
sed -i 's/cmake \.\./cmake \.\. -DPY_INCLUDE=\"<path to Python lib escaping />\" -DPY_CONFIG_INCLUDE=\"<path to Python lib escaping />\"/g' ./build.sh
sed -i 's/\\#/#/g' ./build.sh # awk does not recognize the regular expression \#
# add end
./build.sh

Issue 5: No module named <build, onnx, installer ...>
Fix: pip install <build, onnx, installer ...>
Please fix. Thank you!

dbermond commented on 2022-11-03 23:00 (UTC)

@Twelfthnight I'm aware of the git submodule issue and I'm updating my AUR packages slowly as I can. I updated this package and now it is building fine. Thanks for the interest.

Twelfthnight commented on 2022-11-03 09:34 (UTC)

Hi, The new Git update will now refuse to clone repositories via --local clone optimization if there are symlinks in the object directory. This is due to this CVE

It causes this kind of errors 

Cloning into '.../tensorrt/src/TensorRT/parsers/onnx'...
fatal: transport 'file' not allowed
fatal: clone of '.../tensorrt/src/onnx-tensorrt' into submodule path '.../tensorrt/src/TensorRT/parsers/onnx' failed
Failed to clone 'parsers/onnx'. Retry scheduled
Cloning into '.../tensorrt/src/TensorRT/third_party/cub'...
fatal: transport 'file' not allowed
fatal: clone of '.../tensorrt/src/cub-nvlabs' into submodule path '.../tensorrt/src/TensorRT/third_party/cub' failed
Failed to clone 'third_party/cub'. Retry scheduled
Cloning into '.../tensorrt/src/TensorRT/third_party/protobuf'...
fatal: transport 'file' not allowed
fatal: clone of '../tensorrt/src/protobuf-protocolbuffers' into submodule path '.../tensorrt/src/TensorRT/third_party/protobuf' failed
Failed to clone 'third_party/protobuf'. Retry scheduled
Cloning into '.../tensorrt/src/TensorRT/parsers/onnx'...
fatal: transport 'file' not allowed
fatal: clone of '.../tensorrt/src/onnx-tensorrt' into submodule path '.../tensorrt/src/TensorRT/parsers/onnx' failed
Failed to clone 'parsers/onnx' a second time, aborting

I managed to install tensorrt by setting

git config --global protocol.file.allow always

chengscott commented on 2022-10-14 17:00 (UTC)

@dbermond shall the tensorrt file name be in lowercase? tensorrt-8.4.3.1.linux.x86_64-gnu.cuda-11.6.cudnn8.4.tar.gz

dbermond commented on 2022-10-11 01:05 (UTC)

@acxz Sorry, but AUR helpers are not supported. Up to you to figure out why your favorite helper is not working as you expect.

« First ‹ Previous 1 2 3 4 5 6 7 8 9 Next › Last »