不使用 Docker 构建 Triton 服务器并在 Google Colab 平台上部署 HuggingFace 模型

Build Triton server without docker and deploy HuggingFace models on Google Colab platform

Environment
Building Triton server
Deploying HuggingFace models
客户端
推荐阅读
参考

Environment

根据Triton 环境对应表，Colab 环境缺少 tensorrt-8.6.1，cudnn9-cuda-12，triton-server 版本应该选择 r23.10。
在这里插入图片描述

apt update && apt install -y --no-install-recommends \
    ca-certificates autoconf automake build-essential docker.io git libre2-dev libssl-dev libtool libboost-dev \
    libcurl4-openssl-dev libb64-dev patchelf python3-dev python3-pip python3-setuptools rapidjson-dev scons \
    software-properties-common unzip wget zlib1g-dev libarchive-dev pkg-config uuid-dev libnuma-dev curl \
    libboost-all-dev datacenter-gpu-manager cudnn9-cuda-12

pip3 install --upgrade pip && pip3 install --upgrade wheel setuptools tritonclient[all] diffusers>=0.27.0 transformers accelerate safetensors optimum["onnxruntime"]

upgrade boost

wget https://boostorg.jfrog.io/artifactory/main/release/1.84.0/source/boost_1_84_0.tar.gz
tar -zxvf boost_1_84_0.tar.gz 
cd boost_1_84_0
chmod -R 777 .
./bootstrap.sh --with-libraries=all --with-toolset=gcc
./b2 -j20 toolset=gcc
./b2 install

install libarchive

wget https://github.com/libarchive/libarchive/releases/download/v3.6.2/libarchive-3.6.2.tar.gz
tar -zxvf libarchive-3.6.2.tar.gz 
cd libarchive-3.6.2
./configure
make
sudo make install

install tensorrt-8.6.1

# 方法一
wget https://developer.nvidia.com/downloads/compute/machine-learning/tensorrt/secure/8.6.1/tars/TensorRT-8.6.1.6.Linux.x86_64-gnu.cuda-12.0.tar.gz
tar -xvf TensorRT-8.6.1.6.Linux.x86_64-gnu.cuda-12.0.tar.gz 
sudo mv TensorRT-8.6.1.6/ /usr/local/
vim ~/.bashrc 
export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/usr/local/TensorRT-8.6.1.6/lib
source ~/.bashrc 

# 方法二
wget https://developer.nvidia.com/downloads/compute/machine-learning/tensorrt/secure/8.6.1/local_repos/nv-tensorrt-local-repo-ubuntu2204-8.6.1-cuda-12.0_1.0-1_amd64.deb
sudo cp /var/nv-tensorrt-local-repo-ubuntu2204-8.6.1-cuda-12.0/nv-tensorrt-local-42B2FC56-keyring.gpg /usr/share/keyrings/
sudo dpkg -i nv-tensorrt-local-repo-ubuntu2204-8.6.1-cuda-12.0_1.0-1_amd64.deb

Building Triton server

编译 Triton

git clone -b r23.10 https://github.com/triton