did fastertransformer support version nvcr.io/nvidia/tritonserver:21.07-py3

[changleilei]

Description

branch :https://github.com/triton-inference-server/fastertransformer_backend.git main
docker version : nvcr.io/nvidia/tritonserver:21.07-py3
GPU: Tesla V100

when i flow the steps to build fastertransformer backend, an error happend:

451 BackendModel(TRITONBACKEND_Model* triton_model); 
/workspace/build/_deps/repo-backend-src/include/triton/backend/backend_model.h:45:3: note: candidate expects 1 argument, 2 provided 
[100%] Linking cxX executable../../../../../bin/multi_gpu_gpt_triton_example
[100%] Built target multi_gpu_gpt_triton_example
make[2]:***[CMakeFiles/triton-fastertransformer-backend.dir/build.make:82:CMakeFiles/triton-fastertransformer-backend.dir/src/libfastertransforner.cc.o] Error 1 make[1]:***[CMakeFiles/Makefile2:1457:CMakeFiles/triton-fastertransformer-backend.dir/all] Error 2 make[1]: *** Waiting for unfinished jobs....
/workspace/build/_deps/repo-ft-src/src/fastertransformer/utils/nccl_utils.h:In function'bool test_context_sharing(const strings, const strings) [with T = float]'
/workspace/build/_deps/repo-ft-src/src/fastertransformer/utils/ncclutils.h:72:144:warning:'pipeline_para.fastertransformer::NcclParam:inccl_uid’may be used uninitiali ed in this function [-Wmaybe-uninitialized]
721 NcclParam(NcclParam const& param) 
/workspace/bui1d/_deps/repo-ft-src/src/fastertransformer/utils/nccutils.h:72:144:warning:'tensor_para.fastertransformer::NcclParam::nccl_uid_’ may be used uninitialize in this function [-Wmaybe-uninitialized]
721 NcclParam(NcclParam const& param): 
A
/
[100%] Linking CXX executable ../../../../bin/test_context_decoder_layer
[100%] Built target test_context_decoder_layer
[100%] Linking CXX executable ../../..../bin/test_sampling
[100%] Built target test_sampling
make: *** [Makefile:149: all] Error 2

Reproduced Steps

Reproduced steps:
1. 
git clone https://github.com/triton-inference-server/fastertransformer_backend.git
cd fastertransformer_backend
export WORKSPACE=$(pwd)
export CONTAINER_VERSION=21.07
export TRITON_DOCKER_IMAGE=triton_with_ft:${CONTAINER_VERSION}

2. docker run -it \
    -v ${WORKSPACE}:/workspace \
    --name ft_backend_builder \
    nvcr.io/nvidia/tritonserver:21.07-py3  bash
3. rm /opt/tritonserver/lib/cmake/FasterTransformer/ -rf # Remove original library
cd fastertransformer_backend

4. apt-key del 7fa2af80 && \
    apt-key adv --fetch-keys https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2004/x86_64/3bf863cc.pub && \
    apt-key adv --fetch-keys https://developer.download.nvidia.com/compute/machine-learning/repos/ubuntu2004/x86_64/7fa2af80.pub && \
    apt-get update && \
    apt-get install -y --no-install-recommends openssh-server zsh tmux mosh locales-all clangd sudo \
        zip unzip wget build-essential autoconf autogen gdb python3.8 python3-pip python3-dev rapidjson-dev \
        xz-utils zstd libz-dev && \
    pip3 install torch==1.9.1+cu111 -f https://download.pytorch.org/whl/torch_stable.html  && \
    pip3 install --extra-index-url https://pypi.ngc.nvidia.com regex fire tritonclient[all] && \
    pip3 install transformers huggingface_hub tokenizers SentencePiece sacrebleu && \
    apt-get clean && \
    rm -rf /var/lib/apt/lists/*

export CMAKE_VERSION=3.18 && \
export CMAKE_BUILD=3.18.4 && \
    wget -nv https://cmake.org/files/v${CMAKE_VERSION}/cmake-${CMAKE_BUILD}.tar.gz && \
    tar -xf cmake-${CMAKE_BUILD}.tar.gz && \
    cd cmake-${CMAKE_BUILD} && \
    ./bootstrap --parallel=$(grep -c ^processor /proc/cpuinfo) -- -DCMAKE_USE_OPENSSL=OFF && \
    make -j"$(grep -c ^processor /proc/cpuinfo)" install && \
    cd /workspace/build/ && \
    rm -rf /workspace/build/cmake-${CMAKE_BUILD}
5. mkdir build -p && cd build && \
    cmake \
      -D CMAKE_EXPORT_COMPILE_COMMANDS=1 \
      -D CMAKE_BUILD_TYPE=Release \
      -D CMAKE_INSTALL_PREFIX=/opt/tritonserver \
      -D TRITON_COMMON_REPO_TAG="r${NVIDIA_TRITON_SERVER_VERSION}" \
      -D TRITON_CORE_REPO_TAG="r${NVIDIA_TRITON_SERVER_VERSION}" \
      -D TRITON_BACKEND_REPO_TAG="r${NVIDIA_TRITON_SERVER_VERSION}" \
      -D 
      .. && \
    make -j"$(grep -c ^processor /proc/cpuinfo)" install

[byshiue]

Latest ft backend requires newer tritonserver docker image because several new features are only supported in newer triton.

We adjust using the triton docker we recommend in the document.

[changleilei]

Thank you !!