Can't build TPAT

[heluocs]

Is there something wrong with the build document ？When I follow the document clone the TPAT repository and try to build with command:

mkdir build && cp cmake/config.cmake build

the error occurred:

cp: cannot stat 'cmake/config.cmake': No such file or directory

[buptqq]

Did you clone the repo with 'git clone --recursive' ? If yes, you can find the cmake/config in this directory : 'TPAT/3rdparty/blazerml-tvm'

[heluocs]

It seems clone the TensorRT code and rename the directory as TPAT in the document？not clone the really TPAT repository.

git clone -b master https://github.com/nvidia/TensorRT TPAT && cd TPAT && git submodule update --init --recursive

[buptqq]

Sorry. our mistakes... it should be TPAT directory. git clone --recursive https://github.com/Tencent/TPAT.git we have modified the repo address. thanks

[LegendSun0]

请问，您跑通了么？我安装遇到一些问题，方便留个联系方式，请教一下么？

[buptqq]

请问，您跑通了么？我安装遇到一些问题，方便留个联系方式，请教一下么？ qianqiu@tencent.com

[GeneralJing]

can't build docker image. Sending build context to Docker daemon 401.9kB Step 1/9 : FROM nvcr.io/nvidia/tensorflow:20.06-tf1-py3 ---> 61568efc3e0e Step 2/9 : RUN wget -O "llvm-9.0.1.src.tar.xz" https://github.com/llvm/llvm-project/releases/download/llvmorg-9.0.1/llvm-9.0.1.src.tar.xz && tar -xvf llvm-9.0.1.src.tar.xz && mkdir llvm-9.0.1.src/build && cd llvm-9.0.1.src/build && cmake -G "Unix Makefiles" -DLLVM_TARGETS_TO_BUILD=X86 -DCMAKE_BUILD_TYPE="Release" -DCMAKE_INSTALL_PREFIX="/usr/local/llvm" .. && make -j8 && make install PREFIX="/usr/local/llvm" ---> Using cache ---> 81790a52d1ca Step 3/9 : RUN pip install pycuda onnx nvidia-pyindex && pip install onnx-graphsurgeon onnxruntime tf2onnx xgboost ---> Using cache ---> 7b262fc70ae1 Step 4/9 : RUN git clone --recursive https://github.com/Tencent/TPAT.git /workspace/TPAT && cd /workspace/TPAT/3rdparty/blazerml-tvm && mkdir build && cp cmake/config.cmake build && cd build ---> Running in 6cbc94a98aad Cloning into '/workspace/TPAT'... Submodule '3rdparty/blazerml-tvm' (https://github.com/Tencent/BlazerML-tvm.git) registered for path '3rdparty/blazerml-tvm' Cloning into '/workspace/TPAT/3rdparty/blazerml-tvm'... fatal: unable to access 'https://github.com/Tencent/BlazerML-tvm.git/': Failed to connect to github.com port 443: Connection timed out fatal: clone of 'https://github.com/Tencent/BlazerML-tvm.git' into submodule path '/workspace/TPAT/3rdparty/blazerml-tvm' failed Failed to clone '3rdparty/blazerml-tvm'. Retry scheduled Cloning into '/workspace/TPAT/3rdparty/blazerml-tvm'... fatal: unable to access 'https://github.com/Tencent/BlazerML-tvm.git/': Failed to connect to github.com port 443: Connection timed out fatal: clone of 'https://github.com/Tencent/BlazerML-tvm.git' into submodule path '/workspace/TPAT/3rdparty/blazerml-tvm' failed Failed to clone '3rdparty/blazerml-tvm' a second time, aborting The command '/bin/sh -c git clone --recursive https://github.com/Tencent/TPAT.git /workspace/TPAT && cd /workspace/TPAT/3rdparty/blazerml-tvm && mkdir build && cp cmake/config.cmake build && cd build' returned a non-zero code: 1

[buptqq]

can't build docker image. Sending build context to Docker daemon 401.9kB Step 1/9 : FROM nvcr.io/nvidia/tensorflow:20.06-tf1-py3 ---> 61568efc3e0e Step 2/9 : RUN wget -O "llvm-9.0.1.src.tar.xz" https://github.com/llvm/llvm-project/releases/download/llvmorg-9.0.1/llvm-9.0.1.src.tar.xz && tar -xvf llvm-9.0.1.src.tar.xz && mkdir llvm-9.0.1.src/build && cd llvm-9.0.1.src/build && cmake -G "Unix Makefiles" -DLLVM_TARGETS_TO_BUILD=X86 -DCMAKE_BUILD_TYPE="Release" -DCMAKE_INSTALL_PREFIX="/usr/local/llvm" .. && make -j8 && make install PREFIX="/usr/local/llvm" ---> Using cache ---> 81790a52d1ca Step 3/9 : RUN pip install pycuda onnx nvidia-pyindex && pip install onnx-graphsurgeon onnxruntime tf2onnx xgboost ---> Using cache ---> 7b262fc70ae1 Step 4/9 : RUN git clone --recursive https://github.com/Tencent/TPAT.git /workspace/TPAT && cd /workspace/TPAT/3rdparty/blazerml-tvm && mkdir build && cp cmake/config.cmake build && cd build ---> Running in 6cbc94a98aad Cloning into '/workspace/TPAT'... Submodule '3rdparty/blazerml-tvm' (https://github.com/Tencent/BlazerML-tvm.git) registered for path '3rdparty/blazerml-tvm' Cloning into '/workspace/TPAT/3rdparty/blazerml-tvm'... fatal: unable to access 'https://github.com/Tencent/BlazerML-tvm.git/': Failed to connect to github.com port 443: Connection timed out fatal: clone of 'https://github.com/Tencent/BlazerML-tvm.git' into submodule path '/workspace/TPAT/3rdparty/blazerml-tvm' failed Failed to clone '3rdparty/blazerml-tvm'. Retry scheduled Cloning into '/workspace/TPAT/3rdparty/blazerml-tvm'... fatal: unable to access 'https://github.com/Tencent/BlazerML-tvm.git/': Failed to connect to github.com port 443: Connection timed out fatal: clone of 'https://github.com/Tencent/BlazerML-tvm.git' into submodule path '/workspace/TPAT/3rdparty/blazerml-tvm' failed Failed to clone '3rdparty/blazerml-tvm' a second time, aborting The command '/bin/sh -c git clone --recursive https://github.com/Tencent/TPAT.git /workspace/TPAT && cd /workspace/TPAT/3rdparty/blazerml-tvm && mkdir build && cp cmake/config.cmake build && cd build' returned a non-zero code: 1

Can you check your environment of Git? Mirror will clone submodule of TPAT:'https://github.com/Tencent/BlazerML-tvm.git. Tips: Try to unset proxy of git?

[GeneralJing]

可能是docker镜像内部设置的问题，我在容器外面把仓库clone下来了，然后在启动的容器内部去执行了dockfile里面的剩余命令就可以了，看文档Plugin Compiler Env这块： And export TensorRT/include to Environment Variables : CPLUS_INCLUDE_PATH and C_INCLUDE_PATH： 这块是怎么设置的？我在容器内部执行dpkg -l | grep TensorRT 看到有如下输出： ii graphsurgeon-tf 7.1.2-1+cuda11.0 amd64 GraphSurgeon for TensorRT package ii libnvinfer-bin 7.1.2-1+cuda11.0 amd64 TensorRT binaries ii libnvinfer-dev 7.1.2-1+cuda11.0 amd64 TensorRT development libraries and headers ii libnvinfer-plugin-dev 7.1.2-1+cuda11.0 amd64 TensorRT plugin libraries ii libnvinfer-plugin7 7.1.2-1+cuda11.0 amd64 TensorRT plugin libraries ii libnvinfer7 7.1.2-1+cuda11.0 amd64 TensorRT runtime libraries ii libnvonnxparsers-dev 7.1.2-1+cuda11.0 amd64 TensorRT ONNX libraries ii libnvonnxparsers7 7.1.2-1+cuda11.0 amd64 TensorRT ONNX libraries ii libnvparsers-dev 7.1.2-1+cuda11.0 amd64 TensorRT parsers libraries ii libnvparsers7 7.1.2-1+cuda11.0 amd64 TensorRT parsers libraries ii python3-libnvinfer 7.1.2-1+cuda11.0 amd64 Python 3 bindings for TensorRT ii python3-libnvinfer-dev 7.1.2-1+cuda11.0 amd64 Python 3 development package for TensorRT ii uff-converter-tf 7.1.2-1+cuda11.0 amd64 UFF converter for TensorRT package 容器内部是已经安装好了tensorrt了吧？那这两个变量是怎么设置的？

[buptqq]

可能是docker镜像内部设置的问题，我在容器外面把仓库clone下来了，然后在启动的容器内部去执行了dockfile里面的剩余命令就可以了，看文档Plugin Compiler Env这块： And export TensorRT/include to Environment Variables : CPLUS_INCLUDE_PATH and C_INCLUDE_PATH： 这块是怎么设置的？我在容器内部执行dpkg -l | grep TensorRT 看到有如下输出： ii graphsurgeon-tf 7.1.2-1+cuda11.0 amd64 GraphSurgeon for TensorRT package ii libnvinfer-bin 7.1.2-1+cuda11.0 amd64 TensorRT binaries ii libnvinfer-dev 7.1.2-1+cuda11.0 amd64 TensorRT development libraries and headers ii libnvinfer-plugin-dev 7.1.2-1+cuda11.0 amd64 TensorRT plugin libraries ii libnvinfer-plugin7 7.1.2-1+cuda11.0 amd64 TensorRT plugin libraries ii libnvinfer7 7.1.2-1+cuda11.0 amd64 TensorRT runtime libraries ii libnvonnxparsers-dev 7.1.2-1+cuda11.0 amd64 TensorRT ONNX libraries ii libnvonnxparsers7 7.1.2-1+cuda11.0 amd64 TensorRT ONNX libraries ii libnvparsers-dev 7.1.2-1+cuda11.0 amd64 TensorRT parsers libraries ii libnvparsers7 7.1.2-1+cuda11.0 amd64 TensorRT parsers libraries ii python3-libnvinfer 7.1.2-1+cuda11.0 amd64 Python 3 bindings for TensorRT ii python3-libnvinfer-dev 7.1.2-1+cuda11.0 amd64 Python 3 development package for TensorRT ii uff-converter-tf 7.1.2-1+cuda11.0 amd64 UFF converter for TensorRT package 容器内部是已经安装好了tensorrt了吧？那这两个变量是怎么设置的？

这块如果用Dockerfile的话，From nvidia的镜像里是有的。如果是自己创建的环境，是为了编译plugin.cu和.h。当然如果你修改了python/trt_plugin/Makefile里的trt_path，那么不用加到C和C++的头文件里也可以的