释放双眼，带上耳机，听听看~！
本文分享了在Triton中使用PyTorch进行图片分类的开发实践，包括环境准备、模型准备和转换，以及使用tritonclient库进行推理服务。
前面给大家分享了模型推理服务化框架Triton保姆式教程系列的快速入门和 架构解析，本文进行Triton开发实践的分享。
环境准备
准备镜像及依赖库安装
首先，拉取推理系统的服务端镜像(tritonserver)。
# tritonserver服务端镜像
# nvcr.io/nvidia/tritonserver:<yy.mm>-py3
docker pull nvcr.io/nvidia/tritonserver:23.04-py3
然后，拉取Pytorch官方镜像作为推理系统的客户端同时进行一些预处理操作（当然也可以直接拉取tritonserver客户端SDK镜像）。
docker pull pytorch/pytorch:2.0.0-cuda11.7-cudnn8-devel
# tritonserver客户端SDK镜像
# nvcr.io/nvidia/tritonserver:<yy.mm>-py3-sdk
# docker pull nvcr.io/nvidia/tritonserver:23.04-py3-sdk
接下来，基于官方Pytorch镜像创建一个容器客户端。
docker run -dt --name pytorch200_cu117_dev --restart=always --gpus all
--network=host
--shm-size 4G
-v /home/gdong/workspace:/workspace
-w /workspace
pytorch/pytorch:2.0.0-cuda11.7-cudnn8-devel
/bin/bash
进入容器。
docker exec -it pytorch200_cu117_dev bash
安装datasets、transformer以及tritonclient库。
pip install datasets transformers -i https://pypi.tuna.tsinghua.edu.cn/simple --trusted-host pypi.tuna.tsinghua.edu.cn
pip install tritonclient[all] -i https://pypi.tuna.tsinghua.edu.cn/simple --trusted-host pypi.tuna.tsinghua.edu.cn
准备模型
本文将基于 PyTorch 后端使用 resnet50 模型来进行图片分类，因此，需预先下载 resnet50 模型，然后将其转换为torchscript格式。具体代码（resnet50_convert_torchscript.py ）如下所示：
import torch
import torchvision.models as models
resnet50 = models.resnet50(pretrained=True)
resnet50.eval()
image = torch.randn(1, 3, 244, 244)
resnet50_traced = torch.jit.trace(resnet50, image)
resnet50(image)
# resnet50_traced.save('https://b2.7b2.com/workspace/model/resnet50/model.pt')
torch.jit.save(resnet50_traced, "https://b2.7b2.com/workspace/model/resnet50/model.pt")
最后，拉取Triton Server 代码库。
git clone -b r23.04 https://github.com/triton-inference-server/server.git
一些常见后端backend的配置都在server/docs/examples目录下。
tree docs/examples -L 2
docs/examples
|-- README.md
|-- fetch_models.sh
|-- jetson
| |-- README.md
| `-- concurrency_and_dynamic_batching
`-- model_repository
|-- densenet_onnx
|-- inception_graphdef
|-- simple
|-- simple_dyna_sequence
|-- simple_identity
|-- simple_int8
|-- simple_sequence
`-- simple_string
11 directories, 3 files
拉取Triton Tutorials库，该仓库中包含Triton的教程和样例，本文使用Quick_Deploy/PyTorch下部署一个Pytorch模型进行讲解。
git clone https://github.com/triton-inference-server/tutorials.git
ok，目前为止，前期的准备工作已经好了，下面进行具体的开发实践。
开发实践
创建一个模型仓库
首先，构建一个模型仓库，仓库的目录结构如下所示：
model_repository/
`-- resnet50
|-- 1
| `-- model.pt
`-- config.pbtxt
其中， config.pbtxt 是模型配置文件； 1表示模型版本号； resnet50表示模型名，需要与config.pbtxt文件中的name字段保存一致；model.pt为模型权重（即上面转换后的模型权重）。
config.pbtxt具体内容如下所示：
name: "resnet50"
platform: "pytorch_libtorch"
max_batch_size : 0
input [
{
name: "input__0"
data_type: TYPE_FP32
dims: [ 3, 224, 224 ]
reshape { shape: [ 1, 3, 224, 224 ] }
}
]
output [
{
name: "output__0"
data_type: TYPE_FP32
dims: [ 1, 1000 ,1, 1]
reshape { shape: [ 1, 1000 ] }
}
]
重要字段说明如下：
- name：模型名
- platform：用于指定模型对应的后端（backend），比如：pytorch_libtorch、onnxruntime_onnx、tensorrt_plan等
- max_batch_size：模型推理在batch模式下支持的最大batch数
- input：模型输入属性配置。
- output：模型输出属性配置。
模型仓库构建好之后，接下来启动Triton推理服务端。
启动推理服务
启动推理服务启动服务的方法有两种：一种是用 docker 启动并执行命令，一种是进入 docker 中然后手动调用命令。
本文直接使用 Docker 启动并执行命令，命令如下所示：
docker run --gpus all --rm
-p 8000:8000
-p 8001:8001
-p 8002:8002
-v ${PWD}/model_repository:/models
nvcr.io/nvidia/tritonserver:23.04-py3
tritonserver --model-repository=/models
参数说明：
- -p：宿主机与容器内端口映射
- -v：将宿主机存储挂载进容器，这里将模型仓库挂载进容器
- –model-repository：指定Triton服务模型仓库的地址
启动过程：
docker run --gpus all --rm
> -p 8000:8000
> -p 8001:8001
> -p 8002:8002
> -v ${PWD}/model_repository:/models
> nvcr.io/nvidia/tritonserver:23.04-py3
> tritonserver --model-repository=/models
=============================
== Triton Inference Server ==
=============================
NVIDIA Release 23.04 (build 58408265)
Triton Server Version 2.33.0
Copyright (c) 2018-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
Various files include modifications (c) NVIDIA CORPORATION & AFFILIATES. All rights reserved.
This container image and its contents are governed by the NVIDIA Deep Learning Container License.
By pulling and using the container, you accept the terms and conditions of this license:
https://developer.nvidia.com/ngc/nvidia-deep-learning-container-license
NOTE: CUDA Forward Compatibility mode ENABLED.
Using CUDA 12.1 driver version 530.30.02 with kernel driver version 525.105.17.
See https://docs.nvidia.com/deploy/cuda-compatibility/ for details.
I0529 16:17:07.342510 1 pinned_memory_manager.cc:240] Pinned memory pool is created at '0x7fa1fc000000' with size 268435456
I0529 16:17:07.356934 1 cuda_memory_manager.cc:105] CUDA memory pool is created on device 0 with size 67108864
I0529 16:17:07.356947 1 cuda_memory_manager.cc:105] CUDA memory pool is created on device 1 with size 67108864
I0529 16:17:07.356951 1 cuda_memory_manager.cc:105] CUDA memory pool is created on device 2 with size 67108864
I0529 16:17:07.356956 1 cuda_memory_manager.cc:105] CUDA memory pool is created on device 3 with size 67108864
I0529 16:17:09.509579 1 model_lifecycle.cc:459] loading: resnet50:1
I0529 16:17:48.920259 1 libtorch.cc:2008] TRITONBACKEND_Initialize: pytorch
I0529 16:17:48.920338 1 libtorch.cc:2018] Triton TRITONBACKEND API version: 1.12
I0529 16:17:48.920368 1 libtorch.cc:2024] 'pytorch' TRITONBACKEND API version: 1.12
I0529 16:17:50.186661 1 libtorch.cc:2057] TRITONBACKEND_ModelInitialize: resnet50 (version 1)
W0529 16:17:50.188218 1 libtorch.cc:284] skipping model configuration auto-complete for 'resnet50': not supported for pytorch backend
I0529 16:17:50.188786 1 libtorch.cc:313] Optimized execution is enabled for model instance 'resnet50'
I0529 16:17:50.188800 1 libtorch.cc:332] Cache Cleaning is disabled for model instance 'resnet50'
I0529 16:17:50.188806 1 libtorch.cc:349] Inference Mode is disabled for model instance 'resnet50'
I0529 16:17:50.188812 1 libtorch.cc:443] NvFuser is not specified for model instance 'resnet50'
I0529 16:17:50.188965 1 libtorch.cc:2101] TRITONBACKEND_ModelInstanceInitialize: resnet50 (GPU device 0)
I0529 16:17:51.451587 1 libtorch.cc:2101] TRITONBACKEND_ModelInstanceInitialize: resnet50 (GPU device 1)
I0529 16:17:53.696469 1 libtorch.cc:2101] TRITONBACKEND_ModelInstanceInitialize: resnet50 (GPU device 2)
I0529 16:17:55.208326 1 libtorch.cc:2101] TRITONBACKEND_ModelInstanceInitialize: resnet50 (GPU device 3)
I0529 16:17:57.150685 1 model_lifecycle.cc:694] successfully loaded 'resnet50' version 1
I0529 16:17:57.150981 1 server.cc:583]
+------------------+------+
| Repository Agent | Path |
+------------------+------+
+------------------+------+
I0529 16:17:57.151158 1 server.cc:610]
+---------+---------------------------------------------------------+---------------------------------------------------------------------------------------------------------------------------------------------------------------+
| Backend | Path | Config |
+---------+---------------------------------------------------------+---------------------------------------------------------------------------------------------------------------------------------------------------------------+
| pytorch | /opt/tritonserver/backends/pytorch/libtriton_pytorch.so | {"cmdline":{"auto-complete-config":"true","backend-directory":"/opt/tritonserver/backends","min-compute-capability":"6.000000","default-max-batch-size":"4"}} |
+---------+---------------------------------------------------------+---------------------------------------------------------------------------------------------------------------------------------------------------------------+
I0529 16:17:57.151242 1 server.cc:653]
+----------+---------+--------+
| Model | Version | Status |
+----------+---------+--------+
| resnet50 | 1 | READY |
+----------+---------+--------+
I0529 16:17:57.279760 1 metrics.cc:808] Collecting metrics for GPU 0: NVIDIA A800 80GB PCIe
I0529 16:17:57.279796 1 metrics.cc:808] Collecting metrics for GPU 1: NVIDIA A800 80GB PCIe
I0529 16:17:57.279805 1 metrics.cc:808] Collecting metrics for GPU 2: NVIDIA A800 80GB PCIe
I0529 16:17:57.279812 1 metrics.cc:808] Collecting metrics for GPU 3: NVIDIA A800 80GB PCIe
I0529 16:17:57.280398 1 metrics.cc:701] Collecting CPU metrics
I0529 16:17:57.280604 1 tritonserver.cc:2387]
+----------------------------------+-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
| Option | Value |
+----------------------------------+-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
| server_id | triton |
| server_version | 2.33.0 |
| server_extensions | classification sequence model_repository model_repository(unload_dependents) schedule_policy model_configuration system_shared_memory cuda_shared_memory binary_tensor_data parameters statistics trace logging |
| model_repository_path[0] | /models |
| model_control_mode | MODE_NONE |
| strict_model_config | 0 |
| rate_limit | OFF |
| pinned_memory_pool_byte_size | 268435456 |
| cuda_memory_pool_byte_size{0} | 67108864 |
| cuda_memory_pool_byte_size{1} | 67108864 |
| cuda_memory_pool_byte_size{2} | 67108864 |
| cuda_memory_pool_byte_size{3} | 67108864 |
| min_supported_compute_capability | 6.0 |
| strict_readiness | 1 |
| exit_timeout | 30 |
| cache_enabled | 0 |
+----------------------------------+-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
I0529 16:17:57.282917 1 grpc_server.cc:2450] Started GRPCInferenceService at 0.0.0.0:8001
I0529 16:17:57.283185 1 http_server.cc:3555] Started HTTPService at 0.0.0.0:8000
I0529 16:17:57.325114 1 http_server.cc:185] Started Metrics Service at 0.0.0.0:8002
启动完成之后，我们会在日志中看到模型处于READY状态。
默认情况下，triton 会在每张 GPU 卡上面启动一个实例。
+-----------------------------------------------------------------------------+
| Processes: |
| GPU GI CI PID Type Process name GPU Memory |
| ID ID Usage |
|=============================================================================|
| 0 N/A N/A 12054 C tritonserver 1658MiB |
| 1 N/A N/A 12054 C tritonserver 1830MiB |
| 2 N/A N/A 12054 C tritonserver 1830MiB |
| 3 N/A N/A 12054 C tritonserver 1658MiB |
+-----------------------------------------------------------------------------+
接下来，进入推理系统客户端，发送推理请求。
发送推理请求
首先，创建客户端脚本client.py：
import numpy as np
from torchvision import transforms
from PIL import Image
import tritonclient.http as httpclient
from tritonclient.utils import triton_to_np_dtype
# 图片预处理
# preprocessing function
def rn50_preprocess(img_path="img1.jpg"):
img = Image.open(img_path)
preprocess = transforms.Compose([
transforms.Resize(256),
transforms.CenterCrop(224),
transforms.ToTensor(),
transforms.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225]),
])
return preprocess(img).numpy()
transformed_img = rn50_preprocess()
# 设置连接到Triton服务端
# Setting up client
client = httpclient.InferenceServerClient(url="localhost:8000")
# 指定resnet50模型的输入和输出
inputs = httpclient.InferInput("input__0", transformed_img.shape, datatype="FP32")
inputs.set_data_from_numpy(transformed_img, binary_data=True)
# class_count表示获取 TopK 分类预测结果。如果没有设置这个选项，默认值为0，那么将会得到一个 1000 维的向量。
outputs = httpclient.InferRequestedOutput("output__0", binary_data=True, class_count=1000)
# 发送一个推理请求到Triton服务端
# Querying the server
results = client.infer(model_name="resnet50", inputs=[inputs], outputs=[outputs])
inference_output = results.as_numpy('output__0')
print(inference_output[:5])
预先下载好，用于推理请求的图片：
wget -O img1.jpg "https://cos.aicmty.com/wp-content/uploads/2023/11/lxtlfwhkjtritonbmsjcskfsj_48436.jpg"
执行客户端脚本发送请求：
> python client.py
[b'12.474869:90' b'11.527128:92' b'9.659309:14' b'8.408504:136' b'8.216769:11']
输出的格式为<confidence_score>:<classification_index>。
如果模型具有与每个分类索引关联的标签，Triton 也会返回这些标签，具体格式如下所示。
<confidence_score>:<classification_index>:<classification_index_label>
那么config.pbtxt文件中，需要添加label_filename配置，具体如下所示：
output [
{
name: "output__0"
data_type: TYPE_FP32
dims: [ 1, 1000 ,1, 1]
reshape { shape: [ 1, 1000 ] }
label_filename: "labels.txt"
}
]
推理请求结果如下所示：
> python client.py
[b'12.474869:90:LORIKEET' b'11.527128:92:BEE EATER' b'9.659309:14:INDIGO FINCH']
(3,)
结语
本文演示了如何使用Triton封装一个模型推理服务，相关代码放置在Github，希望能够给你带来帮助。
参考文档：
本网站的内容主要来自互联网上的各种资源，仅供参考和信息分享之用，不代表本网站拥有相关版权或知识产权。如您认为内容侵犯您的权益，请联系我们，我们将尽快采取行动，包括删除或更正。
