mirror of
https://github.com/PaddlePaddle/FastDeploy.git
synced 2025-10-06 09:07:10 +08:00
595ca69251
* add doc for vdl serving * add doc for vdl serving * add doc for vdl serving * fix link * fix link * fix gif size * fix gif size * add english version * fix links * fix links * update format * update docs * update docs * update docs * update docs * update docs * update docs --------- Co-authored-by: heliqi <1101791222@qq.com>
559 lines
20 KiB
Markdown
559 lines
20 KiB
Markdown
中文 ｜ [English](../EN/client-en.md)
|
||
# 客户端访问说明
|
||
本文以访问使用fastdeployserver部署的yolov5模型为例，讲述客户端如何请求服务端进行推理服务。关于如何使用fastdeployserver部署yolov5模型，可以参考文档[yolov5服务化部署](../../../examples/vision/detection/yolov5/serving)
|
||
|
||
## 基本原理介绍
|
||
fastdeployserver实现了由[kserve](https://github.com/kserve/kserve)提出的为机器学习模型推理服务而设计的[Predict Protocol协议](https://github.com/kserve/kserve/blob/master/docs/predict-api/v2/required_api.md) API，该API既简单易用同时又支持高性能部署的使用场景，目前提供基于HTTP和GRPC两种网络协议的访问方式。
|
||
|
||
|
||
当fastdeployserver启动后，默认情况下，8000端口用于响应HTTP请求，8001端口用于响应GRPC请求。需要请求的资源通常有两种：
|
||
|
||
### **模型的元信息（metadata)**
|
||
|
||
**HTTP**
|
||
|
||
访问方式： GET `v2/models/${MODEL_NAME}[/versions/${MODEL_VERSION}]`
|
||
|
||
使用GET请求该url路径可以获取参与服务的模型的元信息，其中`${MODEL_NAME}`表示模型的名字，`${MODEL_VERSION}`表示模型的版本。服务器会把模型的元信息以json格式返回，返回的格式为一个字典，以`$metadata_model_response`表示返回的对象，各字段和内容形式表示如下：
|
||
|
||
```json
|
||
$metadata_model_response =
|
||
{
|
||
"name" : $string,
|
||
"versions" : [ $string, ... ] #optional,
|
||
"platform" : $string,
|
||
"inputs" : [ $metadata_tensor, ... ],
|
||
"outputs" : [ $metadata_tensor, ... ]
|
||
}
|
||
|
||
$metadata_tensor =
|
||
{
|
||
"name" : $string,
|
||
"datatype" : $string,
|
||
"shape" : [ $number, ... ]
|
||
}
|
||
```
|
||
|
||
**GRPC**
|
||
|
||
模型服务的GRPC定义为
|
||
|
||
```text
|
||
service GRPCInferenceService
|
||
{
|
||
// Check liveness of the inference server.
|
||
rpc ServerLive(ServerLiveRequest) returns (ServerLiveResponse) {}
|
||
|
||
// Check readiness of the inference server.
|
||
rpc ServerReady(ServerReadyRequest) returns (ServerReadyResponse) {}
|
||
|
||
// Check readiness of a model in the inference server.
|
||
rpc ModelReady(ModelReadyRequest) returns (ModelReadyResponse) {}
|
||
|
||
// Get server metadata.
|
||
rpc ServerMetadata(ServerMetadataRequest) returns (ServerMetadataResponse) {}
|
||
|
||
// Get model metadata.
|
||
rpc ModelMetadata(ModelMetadataRequest) returns (ModelMetadataResponse) {}
|
||
|
||
// Perform inference using a specific model.
|
||
rpc ModelInfer(ModelInferRequest) returns (ModelInferResponse) {}
|
||
}
|
||
```
|
||
|
||
访问方式：使用GRPC客户端调用模型服务GRPC接口中定义的ModelMetadata方法。
|
||
|
||
接口中请求的ModelMetadataRequest message和返回的ServerMetadataResponse message结构如下，可以看到和上面的HTTP里使用的json结构基本相同。
|
||
|
||
```text
|
||
message ModelMetadataRequest
|
||
{
|
||
// The name of the model.
|
||
string name = 1;
|
||
|
||
// The version of the model to check for readiness. If not given the
|
||
// server will choose a version based on the model and internal policy.
|
||
string version = 2;
|
||
}
|
||
|
||
message ModelMetadataResponse
|
||
{
|
||
// Metadata for a tensor.
|
||
message TensorMetadata
|
||
{
|
||
// The tensor name.
|
||
string name = 1;
|
||
|
||
// The tensor data type.
|
||
string datatype = 2;
|
||
|
||
// The tensor shape. A variable-size dimension is represented
|
||
// by a -1 value.
|
||
repeated int64 shape = 3;
|
||
}
|
||
|
||
// The model name.
|
||
string name = 1;
|
||
|
||
// The versions of the model available on the server.
|
||
repeated string versions = 2;
|
||
|
||
// The model's platform. See Platforms.
|
||
string platform = 3;
|
||
|
||
// The model's inputs.
|
||
repeated TensorMetadata inputs = 4;
|
||
|
||
// The model's outputs.
|
||
repeated TensorMetadata outputs = 5;
|
||
}
|
||
```
|
||
|
||
|
||
### **推理服务**
|
||
|
||
**HTTP**
|
||
|
||
访问方式：POST `v2/models/${MODEL_NAME}[/versions/${MODEL_VERSION}]/infer`
|
||
|
||
使用POST请求该url路径可以请求模型的推理服务，获取推理结果。POST请求中的数据同样以json格式上传，以$inference_request表示上传的对象，各字段和内容形式表示如下：
|
||
```json
|
||
$inference_request =
|
||
{
|
||
"id" : $string #optional,
|
||
"parameters" : $parameters #optional,
|
||
"inputs" : [ $request_input, ... ],
|
||
"outputs" : [ $request_output, ... ] #optional
|
||
}
|
||
|
||
$request_input =
|
||
{
|
||
"name" : $string,
|
||
"shape" : [ $number, ... ],
|
||
"datatype" : $string,
|
||
"parameters" : $parameters #optional,
|
||
"data" : $tensor_data
|
||
}
|
||
|
||
$request_output =
|
||
{
|
||
"name" : $string,
|
||
"parameters" : $parameters #optional,
|
||
}
|
||
|
||
$parameters =
|
||
{
|
||
$parameter, ...
|
||
}
|
||
|
||
$parameter = $string : $string | $number | $boolean
|
||
```
|
||
其中$tensor_data表示一维或多维数组，如果是一维数据，必须按照行主序的方式进行排列tensor中的数据。
|
||
服务器推理完成后，返回结果数据，以$inference_response表示返回的对象，各字段和内容形式表示如下：
|
||
|
||
```json
|
||
$inference_response =
|
||
{
|
||
"model_name" : $string,
|
||
"model_version" : $string #optional,
|
||
"id" : $string,
|
||
"parameters" : $parameters #optional,
|
||
"outputs" : [ $response_output, ... ]
|
||
}
|
||
|
||
$response_output =
|
||
{
|
||
"name" : $string,
|
||
"shape" : [ $number, ... ],
|
||
"datatype" : $string,
|
||
"parameters" : $parameters #optional,
|
||
"data" : $tensor_data
|
||
}
|
||
```
|
||
|
||
**GRPC**
|
||
|
||
访问方式：使用GRPC客户端调用模型服务GRPC接口中定义的ModelInfer方法。
|
||
|
||
接口中请求的ModelInferRequest message和返回的ModelInferResponse message结构如下，更完整的结构定义可以参考kserve Predict Protocol [GRPC部分](https://github.com/kserve/kserve/blob/master/docs/predict-api/v2/required_api.md#grpc)
|
||
|
||
```text
|
||
message ModelInferRequest
|
||
{
|
||
// An input tensor for an inference request.
|
||
message InferInputTensor
|
||
{
|
||
// The tensor name.
|
||
string name = 1;
|
||
|
||
// The tensor data type.
|
||
string datatype = 2;
|
||
|
||
// The tensor shape.
|
||
repeated int64 shape = 3;
|
||
|
||
// Optional inference input tensor parameters.
|
||
map<string, InferParameter> parameters = 4;
|
||
|
||
// The tensor contents using a data-type format. This field must
|
||
// not be specified if "raw" tensor contents are being used for
|
||
// the inference request.
|
||
InferTensorContents contents = 5;
|
||
}
|
||
|
||
// An output tensor requested for an inference request.
|
||
message InferRequestedOutputTensor
|
||
{
|
||
// The tensor name.
|
||
string name = 1;
|
||
|
||
// Optional requested output tensor parameters.
|
||
map<string, InferParameter> parameters = 2;
|
||
}
|
||
|
||
// The name of the model to use for inferencing.
|
||
string model_name = 1;
|
||
|
||
// The version of the model to use for inference. If not given the
|
||
// server will choose a version based on the model and internal policy.
|
||
string model_version = 2;
|
||
|
||
// Optional identifier for the request. If specified will be
|
||
// returned in the response.
|
||
string id = 3;
|
||
|
||
// Optional inference parameters.
|
||
map<string, InferParameter> parameters = 4;
|
||
|
||
// The input tensors for the inference.
|
||
repeated InferInputTensor inputs = 5;
|
||
|
||
// The requested output tensors for the inference. Optional, if not
|
||
// specified all outputs produced by the model will be returned.
|
||
repeated InferRequestedOutputTensor outputs = 6;
|
||
|
||
// The data contained in an input tensor can be represented in "raw"
|
||
// bytes form or in the repeated type that matches the tensor's data
|
||
// type. To use the raw representation 'raw_input_contents' must be
|
||
// initialized with data for each tensor in the same order as
|
||
// 'inputs'. For each tensor, the size of this content must match
|
||
// what is expected by the tensor's shape and data type. The raw
|
||
// data must be the flattened, one-dimensional, row-major order of
|
||
// the tensor elements without any stride or padding between the
|
||
// elements. Note that the FP16 and BF16 data types must be represented as
|
||
// raw content as there is no specific data type for a 16-bit float type.
|
||
//
|
||
// If this field is specified then InferInputTensor::contents must
|
||
// not be specified for any input tensor.
|
||
repeated bytes raw_input_contents = 7;
|
||
}
|
||
|
||
message ModelInferResponse
|
||
{
|
||
// An output tensor returned for an inference request.
|
||
message InferOutputTensor
|
||
{
|
||
// The tensor name.
|
||
string name = 1;
|
||
|
||
// The tensor data type.
|
||
string datatype = 2;
|
||
|
||
// The tensor shape.
|
||
repeated int64 shape = 3;
|
||
|
||
// Optional output tensor parameters.
|
||
map<string, InferParameter> parameters = 4;
|
||
|
||
// The tensor contents using a data-type format. This field must
|
||
// not be specified if "raw" tensor contents are being used for
|
||
// the inference response.
|
||
InferTensorContents contents = 5;
|
||
}
|
||
|
||
// The name of the model used for inference.
|
||
string model_name = 1;
|
||
|
||
// The version of the model used for inference.
|
||
string model_version = 2;
|
||
|
||
// The id of the inference request if one was specified.
|
||
string id = 3;
|
||
|
||
// Optional inference response parameters.
|
||
map<string, InferParameter> parameters = 4;
|
||
|
||
// The output tensors holding inference results.
|
||
repeated InferOutputTensor outputs = 5;
|
||
|
||
// The data contained in an output tensor can be represented in
|
||
// "raw" bytes form or in the repeated type that matches the
|
||
// tensor's data type. To use the raw representation 'raw_output_contents'
|
||
// must be initialized with data for each tensor in the same order as
|
||
// 'outputs'. For each tensor, the size of this content must match
|
||
// what is expected by the tensor's shape and data type. The raw
|
||
// data must be the flattened, one-dimensional, row-major order of
|
||
// the tensor elements without any stride or padding between the
|
||
// elements. Note that the FP16 and BF16 data types must be represented as
|
||
// raw content as there is no specific data type for a 16-bit float type.
|
||
//
|
||
// If this field is specified then InferOutputTensor::contents must
|
||
// not be specified for any output tensor.
|
||
repeated bytes raw_output_contents = 6;
|
||
}
|
||
```
|
||
|
||
|
||
## 客户端工具
|
||
|
||
了解了fastdeployserver服务提供的接口之后，可以使用HTTP客户端工具来请求HTTP服务器，或者是使用GRPC客户端工具请求GRPC服务器。默认情况下，fastdeployserver启动后，8000端口用于响应HTTP请求，8001端口用于响应GRPC请求。此外，还可以使用[VisualDL提供的客户端进行可视化请求](#使用fastdeploy-client进行可视化请求)。
|
||
|
||
### 使用HTTP客户端
|
||
|
||
这里分别介绍如何使用tritonclient和requests库来访问fastdeployserver的HTTP服务，第一种工具是专门为模型服务做的客户端，对请求和响应进行了封装，方便开发者使用。而第二种工具通用的http客户端工具，使用该工具进行访问可以帮助开发者更好地理解上述原理描述中的数据结构。
|
||
|
||
一. 使用tritonclient访问服务
|
||
|
||
安装tritonclient\[http\]
|
||
|
||
```bash
|
||
pip install tritonclient[http]
|
||
```
|
||
|
||
1.获取yolov5的模型元数据
|
||
```python
|
||
import tritonclient.http as httpclient # 导入httpclient
|
||
server_addr = 'localhost:8000' # 这里写fastdeployserver服务器的实际地址
|
||
client = httpclient.InferenceServerClient(server_addr) # 创建client
|
||
model_metadata = client.get_model_metadata(
|
||
model_name='yolov5', model_version='1') # 请求yolov5模型的元数据
|
||
```
|
||
可以打印看一下模型的输入和输出有哪些
|
||
```python
|
||
print(model_metadata.inputs)
|
||
```
|
||
|
||
```text
|
||
[{'name': 'INPUT', 'datatype': 'UINT8', 'shape': [-1, -1, -1, 3]}]
|
||
```
|
||
|
||
```python
|
||
print(model_metadata.outputs)
|
||
```
|
||
|
||
```text
|
||
[{'name': 'detction_result', 'datatype': 'BYTES', 'shape': [-1, -1]}]
|
||
```
|
||
|
||
2.请求推理服务
|
||
|
||
根据模型的inputs和outputs构造数据，然后请求推理
|
||
|
||
```python
|
||
# 假设图像数据的文件名为000000014439.jpg
|
||
import cv2
|
||
image = cv2.imread('000000014439.jpg')
|
||
image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)[None]
|
||
|
||
inputs = []
|
||
infer_input = httpclient.InferInput('INPUT', image.shape, 'UINT8') # 构造输入
|
||
infer_input.set_data_from_numpy(image) # 载入输入数据
|
||
inputs.append(infer_input)
|
||
outputs = []
|
||
infer_output = httpclient.InferRequestedOutput('detction_result') # 构造输出
|
||
outputs.append(infer_output)
|
||
response = client.infer(
|
||
'yolov5', inputs, model_version='1', outputs=outputs) # 请求推理
|
||
response_outputs = response.as_numpy('detction_result') # 根据输出变量名获取结果
|
||
```
|
||
|
||
二. 使用requests访问服务
|
||
|
||
安装requests
|
||
```bash
|
||
pip install requests
|
||
```
|
||
1.获取yolov5的模型元数据
|
||
|
||
```python
|
||
import requests
|
||
url = 'http://localhost:8000/v2/models/yolov5/versions/1' # 根据上述章节中"模型的元信息"的获取接口构造url
|
||
response = requests.get(url)
|
||
response = response.json() # 返回数据为json，以json格式解析
|
||
```
|
||
打印一下返回的模型元数据
|
||
```python
|
||
print(response)
|
||
```
|
||
```text
|
||
{'name': 'yolov5', 'versions': ['1'], 'platform': 'ensemble', 'inputs': [{'name': 'INPUT', 'datatype': 'UINT8', 'shape': [-1, -1, -1, 3]}], 'outputs': [{'name': 'detction_result', 'datatype': 'BYTES', 'shape': [-1, -1]}]}
|
||
```
|
||
2.请求推理服务
|
||
|
||
根据模型的inputs和outputs构造数据，然后请求推理。
|
||
```python
|
||
url = 'http://localhost:8000/v2/models/yolov5/versions/1/infer' # 根据上述章节中"推理服务"的接口构造url
|
||
# 假设图像数据的文件名为000000014439.jpg
|
||
import cv2
|
||
image = cv2.imread('000000014439.jpg')
|
||
image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)[None]
|
||
|
||
payload = {
|
||
"inputs" : [
|
||
{
|
||
"name" : "INPUT",
|
||
"shape" : image.shape,
|
||
"datatype" : "UINT8",
|
||
"data" : image.tolist()
|
||
}
|
||
],
|
||
"outputs" : [
|
||
{
|
||
"name" : "detction_result"
|
||
}
|
||
]
|
||
}
|
||
response = requests.post(url, data=json.dumps(payload))
|
||
response = response.json() # 返回数据为json，以json格式解析后即为推理后返回的结果
|
||
```
|
||
|
||
### 使用GRPC客户端
|
||
|
||
安装tritonclient\[grpc\]
|
||
```bash
|
||
pip install tritonclient[grpc]
|
||
```
|
||
tritonclient\[grpc\]提供了使用GRPC的客户端，并且对GRPC的交互进行了封装，使得不用手动和服务端建立连接，也不用去直接使用grpc的stub去调用服务器的接口，而是封装后给开发者提供了和tritonclient HTTP客户端一样的接口进行使用。
|
||
|
||
1.获取yolov5的模型元数据
|
||
```python
|
||
import tritonclient.grpc as grpcclient # 导入grpc客户端
|
||
server_addr = 'localhost:8001' # 这里写fastdeployserver grpc服务器的实际地址
|
||
client = grpcclient.InferenceServerClient(server_addr) # 创建client
|
||
model_metadata = client.get_model_metadata(
|
||
model_name='yolov5', model_version='1') # 请求yolov5模型的元数据
|
||
```
|
||
2.请求推理服务
|
||
根据返回的model_metadata来构造请求数据。首先看一下模型的输入和输出有哪些
|
||
```python
|
||
print(model_metadata.inputs)
|
||
```
|
||
```text
|
||
[name: "INPUT"
|
||
datatype: "UINT8"
|
||
shape: -1
|
||
shape: -1
|
||
shape: -1
|
||
shape: 3
|
||
]
|
||
```
|
||
|
||
```python
|
||
print(model_metadata.outputs)
|
||
```
|
||
|
||
```text
|
||
[name: "detction_result"
|
||
datatype: "BYTES"
|
||
shape: -1
|
||
shape: -1
|
||
]
|
||
```
|
||
|
||
根据模型的inputs和outputs构造数据，然后请求推理
|
||
```python
|
||
# 假设图像数据的文件名为000000014439.jpg
|
||
import cv2
|
||
image = cv2.imread('000000014439.jpg')
|
||
image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)[None]
|
||
|
||
inputs = []
|
||
infer_input = grpcclient.InferInput('INPUT', image.shape, 'UINT8') # 构造输入
|
||
infer_input.set_data_from_numpy(image) # 载入输入数据
|
||
inputs.append(infer_input)
|
||
outputs = []
|
||
infer_output = grpcclient.InferRequestedOutput('detction_result') # 构造输出
|
||
outputs.append(infer_output)
|
||
response = client.infer(
|
||
'yolov5', inputs, model_version='1', outputs=outputs) # 请求推理
|
||
response_outputs = response.as_numpy('detction_result') # 根据输出变量名获取结果
|
||
```
|
||
|
||
### 使用fastdeploy client进行可视化请求
|
||
|
||
VisualDL提供客户端组件用于快速的访问fastdeployserver服务，帮助开发者进行预测请求和结果的可视化，对部署的服务进行快速验证。该页面基于gradio进行开发，向推理服务发送http请求并获取结果。
|
||
<p align="center">
|
||
<img src="https://user-images.githubusercontent.com/22424850/211204267-8e044f32-1008-46a7-828a-d7c27ac5754a.gif" width="100%"/>
|
||
</p>
|
||
|
||
#### 使用方式
|
||
|
||
安装VisualDL（版本需>=2.5.0)
|
||
|
||
```shell
|
||
python3 -m pip install visualdl
|
||
```
|
||
|
||
使用命令
|
||
|
||
```shell
|
||
visualdl --host 0.0.0.0 --port 8080
|
||
```
|
||
接着在浏览器打开`http://127.0.0.1:8080`（如果浏览器和启动visualdl的机器不同，请替换为启动visualdl机器的ip），即可以看到FastDeploy Client的功能选项卡。
|
||
|
||
#### 功能说明
|
||
|
||
FastDeploy服务的客户端页面主要分四部分，第一部分为fastdeployserver服务的参数设置区，包括http服务的ip, 端口号，性能服务的端口号，要请求推理服务的模型名称和版本号。第二部分为模型的输入和输出区，帮助开发者对预测请求和返回结果进行可视化。第三部分为服务的性能统计区，用于展示服务当前的性能指标。第四部分用来显示各个操作的执行状态。
|
||
|
||
<p align="center">
|
||
<img src="https://user-images.githubusercontent.com/22424850/211206389-9932d606-ea71-4f05-87eb-dcc21c5eeec9.png" width="100%"/>
|
||
</p>
|
||
|
||
- fastdeployserver服务的参数设置区
|
||
|
||
设置需要请求的服务器的ip地址和端口信息，以及要执行推理的模型名称和版本号。后续的请求都会发往所设置的地址。
|
||
|
||
|
||
- 模型的输入和输出区
|
||
|
||
当前提供两种访问服务的方式，第一种是"组件形式"，这种方式会通过服务器直接获取模型的输入和输出，并且通过gradio的组件来进行表示，每个输入和输出变量配有一个文本组件和一个图像组件，开发者根据变量的实际类型来选择对应组件进行使用，比如变量是图像数据则用图像组件输入，是文本则用文本组件进行输入。返回的数据也会通过组件进行可视化呈现。由于不同任务可视化的解析操作不同，因此上方提供了任务类型的选择，当不指定任务类型的时候，输出变量只用文本组件进行显示，显示服务器返回的原始数据。当指定了任务类型，将会对返回的数据进行解析，并且使用图像组件来可视化呈现出来。
|
||
<p align="center">
|
||
<img src="https://user-images.githubusercontent.com/22424850/211207902-4717011d-8ae2-4105-9508-ab164896f177.gif" width="100%"/>
|
||
</p>
|
||
|
||
|
||
第二种是"原始形式"，这种方式相当于一个原始的http客户端，输入框中输入http请求的原始负载（一个json格式的http请求body），输出框中显示服务器返回的原始负载（返回的json格式的结果）。例如对于[paddledetection](../../../examples/vision/detection/paddledetection/serving/README_CN.md)的serving的请求内容格式如下，完整数据可以参考[数据文件](../../../examples/vision/detection/paddledetection/serving/ppdet_request.json)。
|
||
|
||
```json
|
||
{
|
||
"inputs": [
|
||
{
|
||
"name": "INPUT",
|
||
"shape": [1, 404, 640, 3],
|
||
"datatype": "UINT8",
|
||
"data": 具体的图像数据
|
||
},
|
||
"outputs": [
|
||
{
|
||
"name": "DET_RESULT"
|
||
}
|
||
]
|
||
}
|
||
```
|
||
|
||
<p align="center">
|
||
<img src="https://user-images.githubusercontent.com/22424850/211208731-381222bb-8fbe-45fa-bf78-4a3e2c7f6f04.gif" width="100%"/>
|
||
</p>
|
||
|
||
|
||
- 服务的性能统计区
|
||
|
||
用来显示通过fastdeployserver性能服务的端口请求回来的性能指标信息，包括各个模型响应请求的执行统计、延迟统计，以及环境中计算设备的利用率。
|
||
|
||
<p align="center">
|
||
<img src="https://user-images.githubusercontent.com/22424850/211208071-e772ed55-9a3d-4a21-9bca-3d80f444ca64.gif" width="100%"/>
|
||
</p>
|
||
- 执行状态显示区
|
||
|
||
显示在客户端界面执行的所有操作的情况，当开发者点击"获取模型输入输出"、"提交请求"和"更新统计数据"按钮的时候，如果发生异常，执行状态显示区会给出异常发生的原因，如果执行成功也会有相应提醒，以便开发者知道执行情况。
|